All the signal from my 14,000 tweets in one knowledge base

Distribution Beats Everything

• All roads lead to distribution. A 10x worse product with the right distribution outperforms the best product on the market. Building is the easy part - 99% of vibe-coded software serves an invisible market and earns zero. There is no such thing as a broke person who can drive traffic to any product at will: if you master traffic, deals get thrown at you, and the skill transfers across every product type and niche. “Easy” revenue is relative - an easy $30k/month app is easy only for someone great at distribution.
• Traffic generation and direct response marketing are the two master skills of making money online. If you understand both, you can never be poor: there will always be offers to push and channels to push them in. A marketer does not need service expertise - hundreds of equally skilled providers will happily pay a cut for referred leads. “Sauce” (secret tactics) is useless without these fundamentals.
• Marketing is the ultimate portable skill. Most technical communities (AI/startup Twitter especially) lack business and marketing fundamentals, so a marketer can walk into any industry and outcompete more skilled builders. Smarter, more technical people routinely earn less because they cannot position or market their product.
• If you must choose, learn distribution before you learn to build - ideally via affiliate marketing for software that already sells. The affiliate curriculum: what niche communities value, what content converts, how to get views, influencing through content, email list building, audience building, selling to cold traffic, paid ads, funnels, copywriting, social algorithms, and Google/YouTube SEO. Do not attempt a solo SaaS before you understand these. “Build it and they will come” is a route to being broke.
• Beginners should not build products at all. If you have never sold anything, you should not have your own product. Resell a product you already found amazing for a commission - trying to solve both product AND distribution as a first-timer is setting yourself up for failure; tackle one at a time.
• Marketing 101 method for channel selection: identify a product already converting, figure out the exact problem it solves, research every channel (social platforms, Reddit, Google search) where people with that problem look for solutions, rank those destinations, then position your content or ads there. The shortcut: look at what already works for competitors in your niche and replicate it.
• Acquire distribution assets rather than building traffic from scratch. One operator bought the top IQ-related subreddits ranking for “IQ test” and pinned his paid IQ test inside them (with a free-to-take test whose fee appears only at the results screen - sunk cost converts). The generalizable tactic: own the communities, forums, and pages that already rank for high-intent keywords in your niche.
• The core difference between people who print and people who don’t is offer literacy: knowing what a good offer looks like, whose pain it solves, which corner of the internet holds those people, and the lead-magnet/funnel mechanics to pull them in. Beginners throw things at the wall; it is hard to fail once you deeply understand an offer’s value and the community that needs it.
• Never hard-sell in content. It hurts engagement. Make content as educational as possible around the exact problem, funnel viewers to an owned channel (email list or Telegram) with a lead magnet, and do the aggressive selling there - the owned channel is the real breadwinner.
• 95% of what you need to learn to make money with AI has nothing to do with AI (dated 2025-12). It is business fundamentals; AI is a speed multiplier applied to a sound foundation. Build the offer, market understanding, and distribution first; then use AI to 10x outputs. AI will never be the core of the business.

Picking Markets and Problems

• Solve problems you have personally faced - the principle I come back to more than any other. The majority of my most successful ventures came from productizing solutions I originally discovered for my own problems. Having faced a problem yourself tells you which problems need solving, how to build the solution, and how to market it: you know where sufferers look for solutions and how to speak to their pain because you lived it. Someone 100x less knowledgeable about AI will out-earn you if they solve real problems for people with money while you build novelties for broke users - years of experience in your ideal client’s shoes is the actual moat, which is why many brilliant devs stay broke.
• You have no product ideas because you have no problems. Just start any business and you will hit 100 real problems in the first month. Deliberately dive head-first into rich-people problems: you can only solve problems that are painful to you, which is why indie hackers only build tools for other indie hackers - those are the only problems their network exposes them to.
• “Your network is your net worth” works mechanically: your network determines which problems you’re exposed to. Befriend wealthy people, learn their problems, solve one, productize it. Poor people stay poor partly because they only encounter poor-people problems whose solutions can only be sold to poor people. Even simple competence plus trust (being a great support rep to someone wealthy) opens the door.
• Avoid the flashy models everyone sees promoted (trading, dropshipping, YouTube automation, SMMA, copywriting) - normies only believe a model is lucrative if an influencer with a rented Lamborghini says so, which is exactly why those models have a thousand times more supply than demand. If something looks like THE way to make money, run the other way. Meanwhile companies selling egg cartons do nine figures. Method: study top companies in a boring industry, find their biggest expenses, pain points, and optimizable processes, and become the best in the world at solving one very niche problem. I earned a surgeon’s salary ranking YouTube videos for one unique keyword in an unknown niche at roughly 2k views per day.
• Niche down ruthlessly. If your goal is around $10k/month, operate in a hyper-specific niche instead of competing with big players - it is easier to be the #1 safari tour reviewer in Tanzania than the #1000 travel YouTuber. Become obsessively deep in one niche - expert in one domain, generalist in all others; chronic generalists never build the authority that prints money.
• Local “boring” businesses are an underserved, high-money market. Helping dentists and plumbers make more money is “stupidly easy” because their marketing is weak and their customer value is high (trade-off: a scale ceiling). Benchmark (2025-09): a Whop seller charges dentists $3,000/month for patient-acquisition social marketing - 20 members is roughly $60k/month, a trivial expense against customer value in professional niches.
• You don’t need a money-making niche - any passionate community works. BillionGraves (grave-indexing for genealogists) grew to an estimated $5M-$20M with 10-50 employees; 10+ Pokémon-card collector Whop communities each do four to five figures per day. Odd, passionate niches print while everyone fights over “make money online.”
• Sell shovels to crowds that can’t monetize. When a market floods with practitioners who can’t make money (fitness coaching has no barrier to entry, so 95% of coaches earn nothing), the profitable market is selling them the how-to-monetize layer.
• Identify what you already consume most and build a better version of it. Watch business content? Make better videos. Train MMA? Run a better gym. Domain familiarity from heavy consumption is an unfair advantage in product design. If you sell mass-market B2C, deliberately spending time as an ordinary consumer - scrolling the same feeds, using the same apps - is the best way to understand your customers.
• Reverse-engineer job listings and hiring posts for proven business ideas - nobody hires at a loss. A role someone pays a salary for is evidence of a profitable activity you can productize or offer as a service. Same with hiring posts in “TikTok money” Discord servers: message them and work out why hiring you would make them money.
• Validation checkpoints: if there is no community constantly discussing the problem online, it is not an important problem. To find software ideas, embrace one niche and talk to people making real money in it - you will easily compile 100+ inefficiencies solvable with software. Require solid proof a problem exists before building.
• Named research tools (2025): Whop’s free Pulse feature shows what is actively selling and searched on-platform - high search volume equals a validated idea with buying intent. Pair TrustMRR (a product’s revenue) with Swipebuilder (its ad creatives) to reverse-engineer proven offers end to end. posted November 2025 · practice might be outdated
• Follow where the largest players commit capital, not market noise. Tech giants never raced on crypto but all race on AI - structural competition among the biggest companies is the durability test for a trend.
• Nothing is too saturated. Every model still works if you find the edge; saturation objections usually mask a lack of differentiation. Counterweight: positioning against a saturated low-barrier arena matters too - the faceless “info empire” Telegram-funnel niche filled with zero-effort copycats, and if your competition is kids farming sales off a Google Doc, you’re in the wrong arena. What still works there is real sauce rooted in experience.
• Factor barrier-to-entry into market choice. Online business is hyper-competitive precisely because it’s open to everyone - you compete against obsessive, brilliant people who do nothing but work. Physical businesses are often profitable simply because barriers keep competitors out (a food truck’s physical presence does its marketing for it).

Why People Buy: Offers, Pricing, Psychology

• The bridge mental model: people buy when the value on the other side outweighs the risk of paying. A sketchy bridge with $100 on the other side gets no crossers; make it $1M and people heavily consider it - the bridge (your solution) is identical. Understand your market’s deep desires and build bridges to them. Most people fail online because they copy what others do without understanding why people buy.
• Sell things that bring people closer to money. Money is the precursor to the strongest human desires (survival, reproduction, power, freedom), which is why the make-money niche converts so well.
• The oldest formula in history: convince people an invisible threat is harming them and position yourself as the one who can save them - visible at every level from politics to big pharma to health influencers. Understand it to use ethically (real problems people can’t see) and to defend against as a consumer.
• People pay for convenience and simplicity. What seems trivially easy to a technical user is “too technical” for most buyers, who prefer a more expensive one-button version (my friend rejected Roocode + an MCP as too technical and chose the pricier Manus). If you can condense conscious effort into one click, people pay. Simple solutions at 60% efficiency beat complex solutions at 100% - JSON context profiles went viral precisely because anyone could use them. Products win by making complicated things less complicated: there is a lot of money in turning three-click solutions into one-click solutions.
• Low prices only make sense at scale - the Netflix/McDonald’s model. “Why does a rich guy need my $250?” Netflix doesn’t need your $20/month either; the work is worth it purely because of scale. Judge a cheap offer by its scalability.
• A product doesn’t need to work perfectly to sell. Calorie-tracking apps are wildly inaccurate yet earn millions because they deliver the satisfying feeling of tracking. The experience of progress can carry a product.
• Fitness-market insight: the real problem is adherence, not information - people buy info because they believe they lack knowledge, but sticking to it is the only hard part. That belief gap is why the niche prints forever; the product-design implication is to solve adherence, not information delivery.
• Ship offers, don’t theorize. Rich operators test ideas by putting up an offer and pushing traffic immediately; broke people pre-disqualify ideas with endless hypothetical objections. Real feedback from traffic beats a hundred “what ifs.”
• The whole game in four steps: create a product, create a landing page, drive traffic, test and adjust. Expanded loop: create an offer, put up a landing page with a clear buy button, make content about the product, double down if sales come, change and retest if not. If you haven’t literally done these steps, you are not actually “trying” - most people who claim nothing works have never put a purchasable offer in front of traffic.
• $1k MRR is a two-step problem: build a product that solves a genuine problem, then create content about how it solves that problem on the platform those people frequent. No sales means you failed one of the two steps. Concrete version: solve a specific problem, price ~$50/month, find where sufferers look for solutions, promote relentlessly there, get 20 customers.

Product Types and Packaging

• The packaging framework: infoproducts are packaged knowledge (the user does the work), SaaS is packaged capability (code does the work), services are packaged labor (you or a hire does the work), plus community/network (people solving the problem together). All are marketed identically - as a solution to a problem that results in an outcome. SaaS and info sell via the same psychology; many top-earning “SaaS” in public MRR databases are infoproducts dressed as software, and every SaaS benefits from an info component and vice versa.
• Start with services - the advice I repeat to beginners more than any other (2025-2026). The easiest path to $10k+/month is selling services directly, not consumer apps, info, or SaaS; virtually every successful friend of mine started with a service business requiring real work. Services are the easiest sale at the highest prices; sell the solution as a manually delivered service first, and only productize into software once hands-on delivery has taught you the problem’s ins and outs. If you’re grinding toward $10k MRR with a SaaS, run an agency instead: automate processes for paying clients, learn what businesses actually pay for, then productize the winning automations - it’s roughly 10x easier to reach $10k/month with a service, and it’s very hard to fail building tools that support fulfilment work clients already pay for.
• Agencies are paid apprenticeships. Working inside a successful agency shows you the backend - which problems to solve, how to solve them, how to package solutions (my path: agency employee → own agency → productized formats). Running an agency teaches your clients’ business models from the inside; I replicated my clients’ models directly and made more money with roughly a tenth of the work. If you can actually fulfill a service, running it for your own products beats retainers - the value my agency work created for clients dwarfed my fees. Beginners: don’t start with SaaS, consumer apps, or especially info - solve super-specific problems for a few clients and build case studies; you can become the best in the world at one specific process in a year, but not win a broad hyper-competitive market.
• As a buyer, never hire an agency. Every agency owner knows to avoid agencies - they layer margin and account management over the same freelance labor you could employ directly. In-house is almost always more efficient and cost-effective.
• Info products: high margin, real limits. 95%+ of online operators would make 10x more selling information teaching what they already do, with high margins and minimal fulfillment - and info/affiliate hones raw marketing skill fastest (all of the best marketers I know sell or sold info), which is why the info-to-SaaS pivot is common and the reverse is rare. But an info product only helps when information is the actual bottleneck, and for most buyers execution is: even a perfect Elon Musk playbook would fail 99% of buyers, because - like diets - theory is simple and adherence is hard. Sell (and buy) information only where missing knowledge is genuinely the constraint.
• Nobody can sell you a hand-held roadmap to money. If 1,000 people buy the same roadmap, they compete selling the same product to the same audience. Buy skills and ideas with hundreds of possible exits (my example: context engineering) rather than cookie-cutter recipes. Diagnostic for any course: is it tailored for mass appeal to beginners, or to actually set you up for success? Those rarely coexist.
• Decode business-model branding. “YouTube automation” is not automated - it’s a normal business with delegated tasks, previously called “cash cow channels” before the rebrand. Strip the packaging and assess the actual work before buying in.
• Software is the best rags-to-riches vehicle (2025 framing) - AI made it accessible, it’s easy to sell, margins are high, and it’s the natural step up from info: info sellers should transition to software because a tool that solves the audience’s problem is a stronger offer than another course, and selling a functional tool is easier than selling pure information. But don’t build because it’s “cool” - none of the billion-dollar companies are cool; they are useful. Learn why people buy software before touching a coding agent.
• Sequence audience and product together. Standard order: build the audience first, then the product that fits its needs, then convert. My contrarian corollary: it’s easier to build an audience around a product than to retrofit a product to an existing audience - a concrete product gives content a spine and attracts exactly the people who need it.
• Streamline proven processes instead of reinventing. Whop’s content-rewards program (paying clippers per view) productized what influencers already did ad-hoc - despite early hate, it won. Look for clunky, proven processes and productize the streamlined version. Technology usually wins on accessibility, not novelty: most of what went viral with 4o image generation had existed on Hugging Face - OpenAI won by making it easy. Two takeaways: packaging captures the value of existing capability, and you can be early by digging into researcher-adjacent platforms before mainstream productization.
• The 80/20 of software is repurposing existing technology into something usable by people less knowledgeable than you - packaging, not invention. There is a lot of money in turning technical capabilities into cookie-cutter tools for non-technical users (n8n and Make are the proof), and a large share of current AI money is simply putting an intuitive UI on technology too technical for the average consumer.
• Pair SaaS with a community: software provides the functional product, community provides retention and ecosystem. Distribution-native platforms (Whop) become no-brainers when discovery, landing page, and checkout live in one place.
• Payments (2025): avoid Stripe for digital products (support hell at volume; Stripe froze $100k of my balance with unreachable support, while ~20 Whop staff DMed me within an hour of one complaint tweet). I recommend Whop - strong support, 2.7% fees, many integrations. The meta-lesson: be conspicuously active and responsive in the community you serve.
• Beta-test across personas and pay with lifetime access: for my software launch I recruited 12 testers - 3 complete beginners, 3 experienced AI users, 3 software engineers, 3 UI/UX designers - each receiving lifetime access. The mix surfaces onboarding, power-user, technical, and usability issues simultaneously.
• Never lead SaaS marketing with “vibe-coded” - it signals unreliability and undermines trust in security and maintenance. Buyers prefer knowing a qualified developer built it. (Relatedly: I paid a real penetration tester to audit my own app pre-launch; Cursor will not secure your web app for you.)
• Bring-your-own-API-key pricing: my platform charges $75/month flat while users pay OpenAI/Gemini/Claude directly for usage - no credit markups or usage tiers, which is the only way to bundle many AI tools at that price. posted June 2025 · practice might be outdated
• Position AI products as curated knowledge bases, not human-like “coaches.” Human-like positioning reads as disingenuous; a solid knowledge base built on high-quality data - one that holds context on who the user is and matches the right data to their case - is credibly better than 99% of coaches.
• Don’t fixate on churn that’s structural to the category. A social-media scheduling app will always churn as users’ projects die; accept it and outpace it with new-customer acquisition rather than retention micro-optimization. But first diagnose which is your bottleneck: marketing gets people in the door, and retention comes only from a genuinely good product - I am a strong marketer whose weakness was retention, and I made “build truly good products” the priority.
• Don’t vibe-code replacements for cheap SaaS to save $25/month - a misallocation of time and mental bandwidth. Build your own tooling only when it becomes something you sell or a core part of the product.

Moats and Strategy in the AI Era

• Data is the moat - especially for small companies. SaaS in 2026 is niche-specific knowledge and best practices wrapped around an LLM API: a dedicated thumbnail SaaS beats vanilla Nano Banana despite literally calling the Nano Banana API, because it injects context on what a good thumbnail looks like - customers pay $29/month for the data and specialty, not the model. An agent with better task-specific data outperforms a vanilla agent, and people always pay a premium for performance; if everyone’s agent becomes great, the bar simply moves up. Every funnel needs a data layer: 50k email leads are nearly worthless if all you know is the address - instrument the funnel to collect attributes and behavior from the start.
• Base-model advances don’t kill specialized SaaS. When 4o image generation was declared the death of thumbnail tools, my counter was that there will always be a market for foolproof tools optimized for one job - integrating the new model via API plus element-level compositing, color correction, presets, templates, drag-and-drop text, and emotion libraries beats raw ChatGPT. Specialized tools get better with each base-model advance, not obsolete.
• “Wrapper” is not an insult. The value of AI tools lives in the UX layer, not the model: clean interface, hard-coded optimal prompt structure for a specific use case, external API integrations, memory, and tool-calling - packaged so a task becomes 10x easier than doing it manually in an LLM. If you provide better UX than the underlying tool, you are free to upcharge.
• Build narrow, never general. Do not build a general-purpose agent: the moment you get traction, the frontier labs come for your market. Build narrow, domain-specific products they will not bother with.
• Specialized APIs are a major opportunity (2026): packaged frameworks that let AI agents perform profitable actions much better than vanilla models, in niches too small for frontier labs to build in-house - proprietary data, workflows, and best practices as a service. If you can build a useful API that businesses plug into their agentic workflows and you have slightly above-average distribution skills, you are effectively rich. Motto: help an agent do things that increase profit for the business using it.
• Knowledge bases are the next info product. Prediction (2025-06): within two years people will sell knowledge bases instead of courses, because a GPT wrapper’s real value is the unique context it’s hooked up to - “GPT is an empty book, the knowledge base is the content inside.” Selling access to curated, specialized knowledge bases queried through AI becomes a category: everyone can access all information in one prompt, but knowing where information comes from and how to qualify it is the scarce skill. The same thesis drives AI-powered education: static courses and PDFs are outdated because you can’t ask them questions; the future is personalized AI tutors holding all the student’s context, built on curated, credible sources. posted June 2025 · practice might be outdated
• When AI can build anything, the durable edges are everything around the build: understanding what the market wants, deep domain-specific knowledge, taste (knowing what good looks like), writing like a human amid AI content, knowing how to attract attention to a utility, human networks that compress learning, creator networks that can summon attention on demand, and understanding social algorithms well enough to game them. Shipping fast usually produces a moatless app that bets everything on distribution; the same distribution paired with a hard-to-replicate product (e.g., serious data engineering) compounds far more.
• The knowledge asymmetry is itself the edge. Insiders drastically overestimate what the average person knows; most people’s AI knowledge stops at badly-used ChatGPT. Picking an obscure niche and building the best AI tool to serve it is a realistic path to $1M, with far less competition than it appears.
• Specialized knowledge + AI is the #1 skill stack. The specialized knowledge lets you discover real problems; AI lets you fix them - a dentist knows exactly which AI tools would fix his daily problems, an outsider can only guess. No specialization? Either acquire it, or partner with someone who has it in a high-value niche, integrate AI into their workflow, then productize. Broad AI understanding plus deep business understanding out-earns elite technical skill, because knowing which problems businesses need solved is the scarce input - overly technical builders create useless tools because they’ve never run a business.
• Become the go-to AI person for one specific domain. Deep domain knowledge plus basic AI knowledge automatically confers authority, because every industry wants to replace workloads with agents. You don’t need to be highly technical - you need to deeply understand the process being outsourced to agents, and build a brand around front-running that in your industry.
• AI changes industries; it doesn’t kill them. No matter how easy AI makes a task, people with money still outsource it to save time - I can build a better website with AI than most front-end devs, yet I hire a dev 9 times out of 10. The “SaaS is dead” narrative exists because indie hackers sell moatless products to other indie hackers who like to tinker; wealthy buyers ask “is it worth my time,” not “can I do this myself.” The winners are practitioners who integrate AI to deliver faster and better.
• AI automation agencies: real demand, low bar, fast rinse. The space is the new SMMA - the fastest I have ever seen an opportunity market get rinsed by courses - yet operators who actually know what they’re doing print, because business owners have intense FOMO about being outcompeted within 1-2 years. The #1 skill is not automation but understanding how businesses operate: you cannot automate a workflow whose purpose you don’t understand, and one n8n tutorial does not justify a $10k/month retainer. Thousands of profitable owners will pay life-changing money to have time-burning processes automated; the formula is specialize, be reliable, and display competence through a personal brand. Counterweight: don’t automate everything just because you can - for many processes, inexpensive overseas labor beats AI because human input still matters, especially for content and customer-facing work. posted July 2025 · practice might be outdated
• Test AI tools against a money map, not for novelty. Every tool you try should map to a money-making application: AI as advisor, web scraping, market research, fast MVPs, visualizing product ideas, training your team, copywriting, content creation, static ads, traffic generation. Everything you learn should fall back to a skill that earns more. posted April 2025 · practice might be outdated
• My framework for not getting left behind: stay informed (make AI your feed’s focal point, notifications on key accounts, newsletters, experiment with every new tool so your mind recalls it when a matching problem appears); weaponize industry depth (few AI builders have real experience in your industry - ideally run a business in the industry you serve); act on ideas immediately because they expire fast; learn how to learn with AI (summaries, roadmaps, interactive courses); master prompting and tool-chaining (n8n workflows); automate your own daily workflow starting small; and build a distribution edge (people trust people - personal brand, platform algorithms, owned channels like email lists and communities). posted April 2025 · practice might be outdated
• GEO - generative engine optimization (2025-06) - is an emerging discipline on top of SEO. Mechanics: the LLM breaks a query into subqueries, searches (ChatGPT uses Bing) when training data is insufficient, retrieves the top 5-20 results, and cites the clearest, most extractable sources - a #5 result with a crystal-clear structured answer can beat a vague #1, but ranking is still the gateway into the retrieval set. Tactics: be crawlable and indexed in Bing; clear structure (headings, bullets, FAQs); specific data and statistics; firsthand experience and case studies; cite authoritative sources; original, quotable content; optimize for question-phrased queries (“How to rank my YouTube video”); include credentials; and get mentioned on high-authority sites like Reddit and Quora - running an active subreddit is one of the most powerful levers for being recommended in ChatGPT answers. Brand findability compounds across live search and training data alike. posted June 2025 · practice might be outdated
• The most profitable AI sub-industries (2025-06): generative AI (copy, video, images, speech, code); AI-powered hardware; retrieval optimization (GEO, RAG infrastructure, custom search indexing, knowledge graphs); agents/automation; AI-powered education; context engineering; and vertical-specific SaaS (legal, healthcare, logistics, marketing). posted June 2025 · practice might be outdated

Playbooks

• The universal template behind 99% of online businesses:

  encounter a problem many people share → solve it → productize the solution (DWY/DFY offer, info, software) → find where sufferers search for solutions → position content at those touchpoints → funnel traffic to your landing page and email list.

  Compact indie version: identify a problem you personally face, fix it, post about how you fixed it, build an audience of people with the same problem, then commercialize - the fix is proof, the content attracts exactly the buyers, and the audience is pre-qualified.

• Full playbook for making money with AI from scratch (2025-07):

  1. Find a niche where money flows (SaaS, ecommerce, infoproducts, YouTube automation) with problems plausibly solvable by AI - follow the money (what creators sell, which ads perform); never take gurus’ word.
  2. Dive deep: study the industry, join free and paid communities, listen to podcasts, follow niche people across platforms - niche expertise must come before any AI work.
  3. Collect a ranked list of the pain points people mention most; identify multiple AI solutions per pain point; choose based on marketability, not just severity or ease - envision the sales page before building.
  4. Build the solution with the community: free access, honest feedback, iterate together.
  5. Don’t sell until you have an audience: post content showing AI solutions to the OTHER pain points on your list, on whatever platform your vertical consumes, with a newsletter alongside so no traffic is wasted.
  6. With an audience in one vertical that trusts you for solutions, selling becomes natural - you own the product and the distribution machine. posted July 2025 · practice might be outdated

• The A-to-Z AI execution playbook (2025-12), with tools per step:

  1. Identify a painful problem you’ve personally solved that ties to money, freedom, or status.
  2. Validate with GPT deep research: who faces it, competitors, reviews, complaints - confirm people pay for solutions.
  3. Choose product type: software, info product, or agency.
  4. Have Claude build a context profile of your offer so future LLM sessions need no re-explaining.
  5. Build with AI: Claude for course structure or service SOPs, Cursor for software.
  6. Have Claude write sales-page copy from the context profile plus a copywriting SOP and examples of good pages.
  7. Build the landing page in Framer or vibe-code it with Lovable.
  8. VSL: copy from an LLM loaded with best practices, voiced with ElevenLabs, visuals with Nano Banana + Veo 3.1 + CapCut.
  9. Payments via Stripe or Whop.
  10. Find where the audience congregates, study the platform algorithm for a few days, list 10 creators your ICP follows, post genuinely valuable content about the problem, and pay those creators to engage with your best posts to speedrun initial distribution.
  11. Funnel to the landing page via bio links, replies, and case-study posts. AI is just a tool to 10x each step - the fundamentals are the system. posted December 2025 · practice might be outdated

• The “$0 with no skills” automation playbook (2025-04):

  1. Pick a popular online business model (ecom, info, SaaS).
  2. Study it for a week for baseline understanding.
  3. Actually attempt it for a few weeks - the point is writing down which problems you hit.
  4. Join a cheap paid or free community around it to learn which problems others face.
  5. Learn n8n-style workflow automation specifically to solve those documented problems.
  6. Set up Twitter, YouTube, TikTok, and Instagram documenting how you solve them.
  7. People facing the same problems reach out - signing clients is easy when your expertise is already visible. posted April 2025 · practice might be outdated

• Digital-product build-and-distribute playbook (2025-09):

  1. Pick product type by your constraints: SaaS scales best but is hardest to build and sell; info scales second-best (make once, sell forever, eroded by leaks and hype decay) but demands real knowledge, packaging, and trust; services are least scalable but easiest to start - sell your time now via cold outreach with case studies.
  2. Validate by finding communities constantly discussing the problem.
  3. If you can’t code, don’t vibe-code it: build your distribution channels first, then pitch the idea - cracked devs will line up to partner with someone who has distribution locked.
  4. Use a proper payment platform (my pick: Whop over Stripe).
  5. Distribute by infiltrating the communities where you validated, reverse-engineering top players’ content; if content requires a face and you won’t show yours, do percentage partnerships with hungry creators.
  6. Funnel educational content into an owned email/Telegram channel and hard-sell only there. posted September 2025 · practice might be outdated

• SaaS-with-built-in-distribution path (2025-04):

  run a business model, hit real problems, publicly document how you use AI to fix them, turn the solution into software, then use the personal brand built while documenting as your distribution channel. posted April 2025 · practice might be outdated

• Idea-generation playbook (2025-03):

  1) find a partner on the same mission and chat daily, sharing what you learn; 2) once a day, export the chat history to a folder; 3) open it in an AI IDE (e.g., Windsurf) and have the agent clean it into a “YOU:/PARTNER:” markdown; 4) prompt an LLM as “expert business strategist and conversation analyst - extract all the ideas discussed”; 5) follow up with “rank the ideas by how lucrative they are and link any better combined.”

  The ranking step itself generates further ideas. posted March 2025 · practice might be outdated

• X arbitrage playbook (2025-09):

  1) find agency owners/service providers/gurus with audiences on YouTube/Instagram/TikTok but no X presence; 2) agree with big X accounts in their niche to boost the new accounts’ posts for a percentage of resulting profits; 3) pitch the influencers: you run their entire X presence for a percentage of all profits; 4) use AI to extract top insights from their existing content and reformat into engaging tweets; 5) have the big account engage to push posts to their audience; 6) once the account has momentum, promote the influencer’s offer, optionally with a Whop community.

  Broader principle: with $0 and no skills you can still profit by connecting two parties valuable to each other who don’t know it yet. posted September 2025 · practice might be outdated

• UGC flywheel idea (2025-09):

  partner with a UGC creator to make content about how they earn → sell a low-ticket community teaching people to become UGC creators → build a network of ecommerce brand owners → broker deals between brands and your creator network for a percentage.

  Each layer feeds the next; managing a roster of UGC creators is far more profitable than selling advice to the same audience.

• AI content-service path (2025-04):

  learn to ideate and script with LLMs; get good at image, video, and sound generation (4o, Runway, Kling, ElevenLabs, Flux); build a distinctive portfolio; grow a personal brand along the way; sell the service to brands. posted April 2025 · practice might be outdated

• Validate manually first, automate second.

  My first Python project, a pyautogui script pushing YouTube Shorts through Bandicam’s upload API to bypass limits, made $2,000/day after one day of tutorials, because it automated a manually validated money loop.

  The exploit is long patched; the durable lesson is that simple automation applied to an already-proven loop has instant ROI, no CS degree required.

• Competitor-scraping research layer (2026-08):

  if you post content to promote products, scrape all content promoting products like yours, rank it by performance, outliers, sales intent, and engagement ratio, and store it in a database your agent can query for ideas, scripts, and automated content creation.

Partnerships and Hiring

• The marketer × developer duo is the most profitable two-person partnership online. The coder builds, the marketer makes it reach and convert; each covers the other’s blind spot. If you can’t build software, sell somebody else’s to prove you can distribute, then find a hungry developer who lacks distribution and partner 50/50 - distribution proof is itself an asset you bring. The same shape holds in AI automation: the visionary + builder duo, one who understands the game and holds the vision, one who brings it to life.
• Restructure around strengths instead of fighting weaknesses. Hate camera? Faceless brand or hire creators. Want software but hate coding? Partner with a dev who hates distribution. Want a book but hate writing? Dictate to AI. Hate sales calls? Partner with a closer. “Discipline” is often a lie told by successful people who simply outsource everything they dislike.
• Be the visionary between businessman and technical. Understand markets, teams, systems, and distribution, and be just technical enough to give developers exact logic - going deeper is a net negative that steals capacity from visionary work, when thousands of qualified developers can be hired for the rest.
• When you find a good developer, never let him go and pay him well - genuinely good developers are far rarer than assumed, and retaining scarce proven talent is cheaper than replacing it.
• Hire talent without a personal brand. Providers with followings charge a premium for status - the best-branded charge around 4x market worth, and a $5,000+ landing page is almost never justified when a skilled unknown builds better for under $500. Source from forums, Discord, and Telegram rather than X. Specific arbitrage: Russian and Ukrainian freelancers are often excellent at development and design (found on Telegram and forums like zelenka.guru); quotes for identical projects can differ 10x+ with no quality difference - real example: $100k vs $8k for the same build. Around $1.7k/month buys very good talent in India or the Philippines (~6x local average wage), so expect quality at that price.
• Gaming content creators are the best cheap content-talent pool. Most earn nothing, are genuinely passionate, and can be hired inexpensively - DM directly and run a simple system to test trainability; most fail the filter but the hit rate beats other pools. Founders should outsource video content to hungry creators rather than record it themselves.
• Pay on performance, not retainers. Retainers breed complacency (I learned this giving guaranteed monthly retainers because I could). If someone refuses performance pay, they either don’t believe in your product, don’t believe in their results, or both - all disqualifying. The hungry newcomer with something to prove outworks the big agency extracting fees. My editor deal: editors keep 100% of ad revenue plus view-milestone bonuses; I keep product sales.
• Hiring filters: trust the first impression - a candidate who can’t follow simple rules at the application stage will deviate on the job. “Are you the best at what you do?” instantly reveals whether someone understands their own specialty, their self-perception, and how they rate their skills - most people can’t answer at all. When hiring creatives, never ask whether they use AI; judge only whether the end product meets the standard - copy that reads like AI or a site that looks vibe-coded is itself the disqualifying skill issue.
• Never help someone professionally without locking in the incentive first. 9 out of 10 people you “put on” for free will later claim they didn’t need your help. Structure the upside (equity, revshare, affiliate cut) before delivering value, not after.
• Brokering is underrated: you can make millions purely connecting people who want something with people who can fulfill it - no product of your own; the skill is sourcing both sides and taking a cut. Concrete case: without ever offering automation services, my AI-adjacent audience generated 10+ service inquiries daily - my plan was to partner with a genuine engineer and pass leads through hands-off for referral revenue.
• Cold outreach basics: a pitch with no context, too much sell, and nothing in it for the recipient fails - show why you’re messaging this person specifically and lead with their benefit. As an unknown selling automation/AI services, cold outreach mostly fails on zero credibility anyway; the two working paths are in-person/local connections (an uncle’s or family friend’s business, where credibility transfers) or inbound via publishing the niche-specific automations you build.
• Teaching publicly is the best agency client-acquisition strategy. Build an X account teaching people how to do what you do; demonstrating expertise builds authority and clients come inbound with barely any advertising - provided you’re actually good at the craft.
• Network counter-cyclically: when crypto content disappears from your feed (bear market), that’s exactly when to befriend crypto influencers for the next bull run. Relationships are cheap when attention has left a sector and expensive once the cycle turns.
• Avoid public beefs with competitors - feuds almost always hurt both sides; keep your enemies close.

Operations, Delegation, and Systems

• $0→$10k/month you can do alone; $10k→$100k requires delegation, outsourcing, and automation. Everything automatable with code should be; everything else gets automated with virtual assistants - create systems that work, then automate them. Build the team the moment the product turns profitable: avoiding hiring to protect margins is one of the worst growth killers, and even with AI shrinking team sizes, owners still want to delegate operating the systems.
• Laziness plus ambition is the CEO combination. Find the path of least resistance and build systems others fulfill. Hard workers hit a ceiling believing they can do everything personally; every CEO is “inherently a procrastinator” about mundane tasks - the fix is delegation, not discipline.
• Build a business you can remove yourself from. A business requiring your daily presence is barely better than a job (my main income source needs about one hour of daily attention because of systems). Grind at the beginning - 16-hour days are for testing, validating, and installing systems, never for repeating manual work - then systematically delegate every function you fall out of love with. Passive income is never truly passive, but it can be passive for you: my former marketing agency runs on my team at ~10 minutes of my time per week, netting ~$8k/month.
• Operating model per project: 100% hands-on during a setup phase covering both development and distribution, then systems and near-autopilot. I also target at least a quarter of revenue from automated, low-barrier models as diversification, because internet income is volatile.
• If you own a cashflowing business, be an orchestrator. Awareness of what AI makes possible matters more than personally building automations - making them yourself is low leverage when thousands can do it better and cheaper. Ideal loop: have the idea, think through how it should work, hand to the team, move to the next decision.
• VA structure: one trusted personal assistant overseeing everything (in my case a childhood friend), each VA assigned to one small component. Compartmentalization means no single VA sees enough to replicate the operation.
• Kill meetings. Nearly every “quick call” produces a conclusion that fits in one paragraph - default to asynchronous text and Loom videos for almost all business communication.
• Only four activities are productive: improving the product, working on the offer/funnel, creating content/creatives, and talking to clients. Everything else - including learning, until it feeds one of the four - is “beneficial, not productive.” Twin filter: does this create value for customers, or communicate value to potential customers? If neither, you’re wasting time.
• Validation speed and commitment: give a new model 1-2 months of proper effort to show results, then change approach or quit - “great things take years” is partly a psyop that keeps people beating dead horses. But business-model hopping is among the worst habits: commitment applies to the model you validated. The plot twist I tell PDF collectors: all the models work; the variable is commitment.
• The soft-skill reality check: it’s nearly impossible to teach someone who has never operated online how to make money on the internet - success rests on hundreds of soft skills absorbed passively (intuition for what works, what good design is, how the internet economy functions), acquired only through years of actually operating. The complete curriculum, treated as a checklist: market research, offer creation, sales psychology, copywriting, distribution, paid ads, funnels, audience building, delegation, content, email marketing, storytelling, pricing psychology, retention, payment processing, upsells, SEO, automation - nearly every successful operator I know is well-versed in at least 80% of these.

Brand, Face, and Leverage

• The faceless path is complete: learn to drive traffic, learn to monetize, learn to outsource - I built 7 figures fully anonymous with no face, no clients, no sales calls. Any skill plus a personal brand is a money machine (people print money just making YouTube thumbnails), and when you have little capital a personal brand is the right first distribution channel - but pivot away from it as soon as possible.
• A great brand is completely detached from its owner. If revenue is tied to the number of videos you personally post, you have a glorified job. Almost no product you use is tied to a founder’s persona. Use the personal brand only to capture initial attention, then detach: keep the company brand separate from the personal account, build systems and teams, plan for the community to produce future operators/educators, and line up an exit from the content hamster-wheel. Caveat: you need leverage to pull this off - capital to pay creators or an existing brand to bootstrap from.
• Scale content as an operation, not a performance. I do faceless content and hire creators when real-person footage is needed - training a team of 10 creators beats doing everything yourself. Never rely on one account or brand: platform enforcement (YouTube mass-bans) can delete a single-account business overnight, so scale horizontally by replicating a proven content system across multiple accounts and brands, often operated by partners.
• The operator behind influencers out-earns the influencer - the way labels out-earn rappers and producers out-earn actors. Influencer wealth is visible (survivorship bias); backend operators monetizing those audiences make comparable or larger money without performing, and can replicate across multiple creators at once. My long-term plan: run the backend of other creators’ brands rather than being the face. The general principle: the richest person in every arena monetizes the participants - the richest people in sports aren’t athletes; sell tools, education, audiences, and infrastructure to the players instead of competing in the game.
• The publishing/DWY model: vet skilled operators (refuse larpers without real experience - it wastes time), help them build personal brands and audiences, and take a percentage of everything they sell. Results (2025-11): one mentee 0→5k followers with millions of impressions in 2 weeks; another 0→3.5k followers in 3 days. One committed person per brand beats running 10+ accounts yourself. Relatedly, the richest people in the internet-money industry own a percentage of many offers rather than running a few of their own - diversified upside, less operational load per offer.
• Flexing is only rational when it’s the marketing for what you sell. Gurus flexing cars to sell courses are doing correct marketing - the display is proof-of-claim. If no offer is attached, showing off is pure cost. Decide whether your lifestyle content is a funnel or an expense.
• Video-game economies are the internet-marketing academy. Nearly all the best internet marketers started making money inside game economies - Minecraft ranks, build teams, content-locking taught product, pricing, and traffic before adulthood.

Information Hygiene: Secrecy, Gurus, and Echo Chambers

• Never reveal your niche alongside your earnings. Distribution can be reverse-engineered; revenue + idea + traffic source is an open invitation to spawn competitors, which is why experienced affiliates gatekeep niches. Sharing what you sell and your financials with an audience that sells similar products gets loud success cloned into oblivion. Exceptions: products actually marketed to the X audience and B2B marketing/coding tools, where the audience is the ICP - my rule of thumb: if less than 25% of a product’s revenue comes from X, talking about it there is a net negative. Public MRR databases have effectively zero upside for serious operators.
• Nobody online makes as much as they claim. The people genuinely making a killing don’t post revenue numbers; the inflation dynamic runs $10k/month guys claiming $100k to appeal to real $100k guys claiming $1M. Revenue-flex marketing attracts only low-quality leads - authority-based value content out-converts it with better buyers.
• If you run client accounts, never flex client results on your timeline. Nobody serious works with a loudmouth who airs out every move; discretion is what makes partners trust you.
• The build-in-public community functions like an MLM-style distribution scheme. A few top accounts push indie-hacker traffic to products built for indie hackers, while members are told documenting everything is marketing - it is not; honest leaders would teach marketing before building. Don’t take advice from a single community’s echo chamber: even top build-in-public players often make only five figures MRR (a small-market signal), and founders you’ve never heard of, with no personal brand, out-earn most of its stars. Balance with other circles - marketers, affiliates, YouTube automation, SEO people.
• Vet gurus by whether they demonstrably do the thing they teach, where they teach it. A personal-brand coach who can’t hit 1k impressions on his own tweets won’t build yours - X is among the easiest platforms to grow on with real value. Corollary: a truly personal brand can’t come from someone else’s predetermined playbook; the approach contradicts the product.
• Free long-form content is marketing by design. Essentially every YouTube video from money-makers is a funnel; nobody generous enough to give away all their sauce spends hours producing free video. Genuinely complete playbooks appear only when the creator has a financial incentive (like a paid launch) to communicate everything clearly - authentic fragments show up on X because posts are cheap to write.
• Shortcuts do exist. I sold software that synthetically ranked YouTube videos for high-intent keywords, and my own X growth was boosted by a coordinated push. Selling a method you actively use often costs you nothing - buyers were happy it worked. Disbelief in shortcuts is usually inexperience with them. But never build a business around a single patchable method: at 16 I made four-figure days with YouTube Shorts spam and felt depressed knowing it couldn’t last - and it didn’t. If revealing your method would destroy your business, your livelihood is on wobbly legs; build durable assets (audience, skills, brand, product) on top of or instead of exploits.
• Don’t fear copycats on ideas - out-innovate them. I share ideas publicly because each spawns a hundred better ones; the moat is idea velocity, not secrecy. (Secrecy applies to live niches and traffic sources, per above - not to ideas.)

Money Management and the Endgame

• The two undisputed rules of online money: diversify income streams and stack liquid cash. Digital income is violently unpredictable - one month $50k, the next your traffic source dies. Most young guys who appear to be winning have little to their name; a $150k car off internet income is usually only rational with family wealth as a safety net.
• The realistic goal is fast cashflow converted ASAP into security, not a lifestyle. My plan: run up capital through digital products, invest in property for permanent security, then transition to a fulfilling, scalable physical business. Gurus who spent their one big play on a Lamborghini will be broke within years.
• My wealth hierarchy (2025-02): info products will never produce generational wealth; software can but rarely does (defensible innovation pits you against multi-billion-dollar corporations); online models excel at getting an individual to low-seven figures fast; after that, dominating a local physical industry compounds better - plan cash flow around wildly varying month-to-month revenue.
• When parking online profits in the real world, own the real estate under the operating business - the McDonald’s model. Open a gym and own the building: the land appreciates while the business generates revenue on top, stacking two return streams on one investment. McDonald’s makes most of its money this way.

Business Ideas I’ve Called Out

• AI-powered genealogy (2025-06): ChatGPT deep research surfaces ancestral records manual searching can’t (it found my great-great-grandfather’s death date, war of death, father’s name, and parents’ marriage certificate). Suggested product: a research tool bundled with handwriting transcription for pre-1920s records. posted June 2025 · practice might be outdated
• Photo-restoration wrapper app (2025-06): wrap Flux Kontext (plus animation) in a “restore old photos” app, marketed with an emotional TikTok/Instagram series restoring the audience’s own photos - estimated at least a few thousand dollars MRR. The pattern generalizes: emotional content series as organic acquisition for a simple AI-wrapper app. posted June 2025 · practice might be outdated
• Vibe-code professionalization agency (2025-07): thousands of AI-generated apps get traction but are insecure and unmaintainable; an agency that takes them over and makes them production-grade sits on obvious demand - a multi-million dollar idea. posted July 2025 · practice might be outdated
• Knowledge-management agency (2025-06): collect all of a business’s context - operations, team, accounting, systems, marketing - structure it into knowledge bases AI systems plug into, then set up automatic ongoing context collection, on a fat retainer. Priceless to any business serious about AI adoption, and recurring by nature. posted June 2025 · practice might be outdated
• B2B AI training for creative teams (2025-08): corporations have the budget and mandate to adopt AI but their creative departments lack practical workflow knowledge - “freelance creative team scaling” as a training/consulting offer. posted August 2025 · practice might be outdated
• Vibe-coder streamer (2025-06): a streamer who builds software live all day and keeps the audience current on AI advancements - an underserved intersection of AI and streaming culture requiring both AI fluency and entertainment instincts (clippable moments, tool experimentation); done right, “generational wealth.” posted June 2025 · practice might be outdated
• Homeschooling info for concerned parents (2025-08): fear of school-system indoctrination drives purchases of curricula and aligned children’s books - I’ve watched a seller run the angle profitably. The general pattern: parental fear plus a ready-made solution is a strong info-product niche.

How the X Algorithm Works

• The algorithm is open-source - interrogate it.

  The X recommendation algorithm repository is public: feed it to an LLM and ask how ranking works, what signals are weighted, and how to design posts around them.

  There is no secret sauce to going viral on any platform. Algorithms are learnable rules; reverse engineer what gets pushed and produce that.

• The core ranking model predicts user actions and scores your post per viewer.

  An AI model predicts the odds that each viewer takes roughly 20 specific actions on your post, each with a weight (as of early-mid 2026).

  Heaviest positive signals: retweets and quotes (takes people want to argue with or co-sign), follows of the author (triggered by demonstrated expertise: “if this is their random tweet, imagine the rest”), dwell time (threads, articles, lists that stop the scroll), and video views past a minimum watch-time threshold.

  Strong signals: replies (questions, relatable experiences, controversial takes), profile clicks, external shares (DM shares especially), and visual hooks that prevent a fast scroll-past.

  Supporting signals: click-to-expand, photo expands (detailed infographics, small-text screenshots), quoted-tweet clicks, bookmarks.

  Likes matter less than assumed - a cheap signal.

  Negative weights collapse reach: “not interested” (irrelevant content, undelivered clickbait), mutes (posting too much, repetitive topics), blocks (toxicity), and reports.

• Distribution runs on a personalized relationship graph.

  The algorithm profiles every viewer from their last ~127-128 engagements, follows, and mutes, and scores your post against that taste pattern, so hyper-specific content finds exactly its audience.

  Posts reach people two ways: in-network (followers, whose feeds multiply your score) or out-of-network AI matching to strangers with similar taste, which is why 1,000 real followers beat 10,000 random impressions.

  Reaching strangers requires early positive engagement from your immediate network, which the algorithm uses for taste-matching (“simclusters”: content popular within your community gets amplified, and bridging multiple communities enables viral spread).

  Spammy engagement - engagement groups, mass follow-backs - dilutes your graph signal and hurts future distribution.

  Every post also carries silent metadata (video length, language, media type, brand safety) that shapes reach, posts older than 7 days are filtered out entirely, and cheap filters penalize muted keywords and spammy reused formats.

• Author reputation compounds.

  Accounts carry a reputation score (“tweepcred”) built from how people react to your content; a high-authority small account can out-reach a 100k-follower account.

  Consistency, topic-specific authority, and a text-quality score that rewards readability, line breaks, and clear structure (and penalizes ALL-CAPS) all feed it.

  Toxicity scores and detected inauthentic engagement patterns nerf reach.

• Engagement velocity was king; content scoring now leads (as of late 2025).

  For most of 2024-2025 the dominant mechanic was the first 30 minutes: if a tweet out-engaged your usual baseline early, the algorithm pushed it wider, so posting time, hooks, and engagement circles mattered most.

  A late-2025 algorithm change shifted this: posts are now scored more on their actual content, via AI/LLM classification (Grok analyzing the text), than on raw engagement ratios.

  Posts with amazing engagement ratios stalled at 3k views, and “reply for X” lead-magnet posts that would have gone viral three months earlier stopped working.

  Quote tweets are slightly favored, video/media heavily favored, and by January 2026 non-followers began seeing tweets within the first 1-2 hours, making follower-external reach much better for quality content.

  Practical takeaway: optimize the substance and media of each post rather than engagement-bait mechanics.

• Dwell time is a core ranking input - engineer for it.

  The longer people spend on your post, the bigger the push.

  This is why prompt-sharing posts go viral (viewers stop, read, and copy the prompt), why long well-formatted posts that trigger “show more” outperform short ones, why screenshots can beat quote-tweets (readers must stop and read), and why X articles were heavily favored.

  I posted a text sales letter as an article and got 100k views with strong conversion (as of early 2026; expect the format to saturate).

  Write posts that hold attention and get saved, not posts skimmed in one second.

• Profile clicks are heavily weighted.

  Content that makes people click into your profile boosts distribution - the same signal that makes accounts provoking instinctive profile checks go viral more easily.

  Optimize for authority and intrigue.

• Quote tweets outperform standalone tweets.

  Despite getting less engagement, quote tweets almost always out-reach cold posts, and the boost scales with how well the quoted tweet itself is performing (as of mid 2026).

  Quote strong posts rather than posting cold.

• Space posts out - the diversity decay changed the volume math (as of 2026).

  The algorithm penalizes your 2nd, 3rd, and 4th post shown to the same viewer (roughly 70%, then 50% score), so one banger now beats five mid posts.

  Note this is a change: in 2024 there was no observable posting-frequency penalty on X, and extreme volume was a viable cold-start strategy.

• Platform quirks worth knowing.

  Follower totals mean little - dead followers don’t hurt, because what pushes a post is engagement from the people who actually see it.

  Creator payouts are driven by engagement from X Premium accounts, but Premium itself gives only a slight ranking boost and is not necessary (I grew to 4,000 followers before buying it).

  Replies to replies barely boost a tweet. Starting a tweet with an @ makes it act like a private mention that skips the For You feed.

  Low-effort agreement replies that restate the original tweet add nothing and do not build an audience.

Starting From Zero on X

• Big-account engagement is the distribution unlock - there is no fourth path. X does not boost new accounts; with 0 followers a tweet reaches almost nobody, so a small account’s growth comes from piggybacking on accounts that already have attention. The three realistic paths: have friends with big accounts, pay a big account for engagement or mentorship, or give big accounts a concrete incentive to engage - promote their product, provide advice or labor, drive eyes to their offer. Any clear value exchange works, because when your activity benefits them, they are directly incentivized to engage with and amplify you. Virality has two checkboxes: good content AND a boost from a big account in your niche; one good tweet pushed to someone else’s niche audience takes you from 0 to your first 1k followers. My own first tweet got ~20-30k impressions because established friends amplified it, and I would never start a new account without a plan to make it blow up instantly through such connections.

• The affiliate route (fast path for a small account):

  1. Find an X account in your niche selling a service, community, or info product.
  2. DM asking to be their affiliate.
  3. Run lead-magnet posts promoting their product - viral value bombs that subtly pitch their offer.
  4. Ask them to reply to your lead magnets; their engagement boosts the posts in the algorithm. You drive buyers to their offer, they push your distribution - both win.

• The “reply guy” strategy is a psyop. Mass-reply prescriptions (“10 tweets and 100 replies per day”) are guru snake oil - often the guru is recruiting an army of reply guys to boost their own tweets; always ask what a growth teacher gains from the behavior they prescribe. I grew to 1k followers without replying to other accounts at all, and helped multiple people past 10k with replying never part of the strategy. Replying only makes sense when you genuinely add value - never detectable AI-slop replies - and mainly works because people who engage with you see your future original posts. Building relationships with bigger accounts by being genuinely useful to them beats replying under their posts.

• My full 0-to-10k playbook (I personally grew 0 to 50k in 9 months):

  1. Pick the niche. It must be big - list 5 accounts consistently doing 10k+ impressions per tweet - AND you must be genuinely deep in it. Real experience is non-negotiable: what you did before you started tweeting determines what stories you can tell and what advice you can give, and larping is obvious to anyone who has been around.
  2. Write instantly-scannable value tweets. Every tweet must communicate its exact value instantly: self-contained value bombs, one line per idea, clean line breaks, formatted for easy scanning. Make it bookmark-worthy (“will they need this later?”) and design it to incite engagement. Relevant images and videos get extra algorithmic weight.
  3. Bootstrap reach through big-account engagement (paths above), or take the slow path: become a genuine user of the app, replying only when you have real value to add.
  4. Once anything goes viral, ride momentum with relentless consistency - value only, multiple times daily, without shifting positioning. Around 10 high-quality tweets daily; in the 0-10k push, focus purely on value even if it feels soulless.

• Volume compresses timelines - my cold-start benchmarks.

  I went 0 to ~7,000 followers in my first 7 days on roughly 400 tweets (three to six months of most people’s output in one week), posting experience-based value 12+ times daily during a “100 tweets a day” phase, with barely any replies.

  First month: 219 original tweets, 4.25 million views.

  Longer arc: ~1M impressions in the first 2 days, 25k followers in ~7 months, 50k in 9 months, ~56.5k and ~100M impressions in the first year - including one 3-month stretch that added 34k followers off 45M impressions, and a 20k-to-50k month when I made posts as mass-appeal as possible.

  People attribute fast growth to magic; it is mostly output density plus decent content.

  (Note the 2026 author-diversity decay above changed the raw-volume math - space output rather than dumping it.)

• What powered that growth: unique, experience-based free value.

  I posted the ways I had actually made money - real, actionable, experience-based content that stands out against a timeline saturated with generic AI-flavored posts; readers can tell when someone speaks from experience.

  Giving genuinely good information away free is the growth engine; monetization comes after authority is established - the two are sequential, not in conflict.

  The catch: 95% of people lack that experience and try to skip acquiring it.

• Run lead magnets aggressively under 10k followers.

  Lead magnets requiring a follow plus reply to trigger DM delivery simultaneously farm engagement and build your email/Telegram list - I grew to 7k in 7 days using one, and monetized immediately.

  At small size, worrying about the algorithm is pointless (it only becomes predictable once you have a steady baseline of impressions) - maximize growth and lead flow instead.

  They become spammy and not worth it at larger size, the main downside is a flooded DM inbox, and note the late-2025 algorithm change nerfed “reply for X” posts specifically.

• Engagement farming is a bootstrapping tool, not a strategy.

  On engagement-ranked platforms (X, Instagram), refusing to farm engagement while using the app for business is shooting yourself in the foot - I used reply-gated lead magnets, provocative prompts, and strategic quote-tweet beef myself for initial traction.

  But drop it once you have momentum: continuing after you have an audience damages the reputation you are building.

• Get onto “accounts to follow” lists.

  Being included in curated niche follow lists compounds - readers who work through a list follow everyone on it.

  I gained roughly 20,000 followers in 5 days from list mentions. Cultivating the relationships that get you listed is a high-leverage tactic.

• An introduction tweet mentioning a large relevant account can borrow its distribution.

  One newcomer got 33k views and 600+ followers from a single intro tweet mentioning me - the mention invites engagement from the big account, which triggers the engagement-quality signals.

• Profile and bio: they can only hurt you, so keep them clean.

  Your profile is a landing page for your ideal customer. The bio should be a simple one-liner instantly stating what you do and how you benefit the reader.

  A clutter of accolades with vertical dividers reads as LinkedIn; hard-shilling an agency in the bio signals you don’t understand the game; Miami-balcony and rented-Lamborghini imagery repels the buyers you want.

  A clean, distinctive profile - even an anonymous avatar - can feel more human than a real face with corporate formatting.

  A profile picture mostly cannot help you, but a cringe one can hurt you.

• Time heavy posting around niche hype waves.

  My all-time peak month (19M views, March 2025) coincided with the GPT-image-1 hype cycle - attention in a niche spikes around major launches, so concentrate output there.

  More generally, pioneers and early adopters of any content format capture most of its value; once low-quality copycats flood in, the format dies - move on formats early.

• X only works if you genuinely enjoy writing.

  After helping about a dozen people grow past 10k followers, the pattern: the successful ones fall in love with writing and let their own creativity take over - which cannot be taught.

  I only sustained 14 months of daily posting because it was fun; people who post mechanically and try to “systemize” it underperform and would earn more elsewhere.

Nobody Cares About You: Content Principles

• The first step is accepting that nobody cares about you.

  People follow for selfish reasons; every piece of content must push the reader along their own path.

  In the early stages your content has exactly two purposes: establish why anyone should listen to you, and be genuinely valuable to the reader.

  Do not post about yourself, your day, or your personal stories - people begin to care about your brand and story only through repeated exposure to your value, so include hints of personality inside value posts.

  Anyone claiming secret growth knowledge is lying. This approach took my account to 60k followers in a little over a year.

• “Document your journey” is not a starting strategy.

  Nobody cares about your journey until it is demonstrably successful - it is nearly impossible to start a movement on “watch me try.”

  Grow instead by sharing solutions, insights, and expertise aimed at your ideal customer’s selfish benefit, through an authoritative lens; documenting works only after you have earned an audience.

  (My earlier 2024 advice recommended documenting with an outside editor; my settled later position reverses it for the cold start.)

• Make the reader the protagonist.

  Instead of tweeting “I did XYZ and it helped me,” tweet “here’s how you can do XYZ” and name the benefit. Package lessons as short, digestible, instantly-scannable posts.

• Pre-publish checklist: does it provoke thoughts, spark emotions, or start conversations?

  If a piece of content does none of the three, the algorithm has no engagement signal to amplify.

• Polarization drives distribution - manufacture it deliberately.

  A lot of people need to hate you for a lot of people to love you; every tribal community has a rival tribe, and appealing strongly to one side generates engagement from fans and haters alike.

  My compressed personal-branding formula: be polarizing via an us-versus-them dynamic, simultaneously drop niche value to become a thought leader for your target demographic, and accept being a performer on camera - missing any leg breaks the model, and the last requirement disqualifies most people.

• There is a real authenticity-versus-virality tradeoff - choose per phase.

  What goes viral is usually cringe, normie-coded messaging: I grew 20k to 50k in one month by making posts as mass-appeal as possible, while my authentic era builds trust and audience quality instead.

  Neither is wrong; know which objective you are optimizing for.

  Meanwhile authenticity is the current meta at the persona level: when 99% of influencers flex rented Lambos and alpha-male personas, the 1% who lead with humility stand out - when they zig, you zag.

• Never build a “perfect genius” persona.

  Every slip-up hurts a flawless persona ten times more than an authentic one; an audience that knows you are figuring things out forgives mistakes.

  Likewise, copying a proven persona (Tate clones, my own imitators) always fails - the original exhausted the format and the copy lacks the person.

  Going against the grain as genuinely yourself is the only way to bring something new; I grew on money-Twitter without ever flexing, under a deliberately unserious handle.

• Flexing only works as proof attached to real teaching.

  Flex with no value makes you a meme; flex with value makes you a thought leader.

  Build the brand on humility and total transparency instead: never make false claims, be open that you will eventually sell something, and skip the “I have nothing to sell you” trick - sophisticated audiences see through it.

  Rarely post money screenshots even though they get attention; proof should be in the quality of your advice.

  Trust has the biggest compound effect on conversion rate: low-quality leads are attracted through deception, high-quality leads through honesty.

• Virality mostly rewards shamelessness.

  The barrier to attention is not knowledge or advice quality but willingness to look like a fool in front of an audience - which is why low-status-looking creators outearn polished intellectuals.

  Attention first, monetization second - but see the audience-quality warnings below before taking the clown route.

• Write for the silent readers.

  99% of people who see your posts never engage, and some are very high-level anonymous operators.

  Opportunities - partnerships, clients, jobs - come from lurkers you never knew were reading; engagement metrics undercount your real reach.

• Keep all interactions manual.

  Write replies and DMs as if messaging a friend; the human touch is crucial on social media and automation kills it.

Audience Quality Beats Reach

• Views do not equal money.

  A small volume of high-intent views monetized with your own product beats millions monetized with ad revenue: 425 views once made me $200; a $40 Telegram-promo video did 60k views and $20k revenue, while YouTube automation channels spend $400 per video for $500 of ad revenue on 100k views.

  If you rely on ad revenue, advertisers pay more for the first 5 seconds of your video than you earn from it - sell something instead.

• Follower counts are not comparable currencies.

  10k Twitter followers are worth more than 100k Instagram followers - Twitter’s audience is closer to money, easier to reach, and easier to move off-platform.

  Follower count is not influence either: countless TikTok/Instagram accounts have millions of followers and zero influence, because shaping people’s thinking carries far more weight than entertaining them.

  Optimize for influence over reach - the entertained audience watches you; the influenced audience acts on what you say.

• Entertainment is the worst niche; being an internet clown keeps you broke.

  You can make a living from around 500 views a day in a high-value niche, while entertainment success is survivorship bias over the top 0.01%.

  A clown audience cannot be sold to - which is why mass-appeal influencers end up pushing gambling, adult-content funnels, and crypto schemes, and why viral “influencers” stay broke.

  Traffic is step one, traffic quality is step two: build the audience deliberately for the offer you intend to sell.

• Content specificity determines follower quality.

  Niche, actionable advice converts followers into subscribers and clients; general life advice gets views but no authority and no conversions.

  My X followers transferred directly into Telegram subscribers because the content was on-point advice in one specific niche; broad motivational accounts plugging a Telegram get impressions but few subs.

  If you sell a service, your content must be niche-specific value for the audience that buys it - a social-media-services seller posting about dating attracts followers with zero interest in the offer.

  Bait and flex posts likewise attract followers who will never buy; make content for the person you would want as a follower.

• Segment lifestyle vs. tactical content by which audience you need.

  High-quality prospects don’t watch lifestyle vlogs - they have their own lifestyle; a title like “adding $44k to an info product in 14 minutes” is what your ideal client clicks.

  Use lifestyle content only as top-of-funnel for mass conditioning; double down on tactical, result-driven content for clients and high-ticket deals.

  “I’m better than you” flex marketing attracts insecure beginners, only works at massive volume for low-ticket transformation offers, and is actively disqualifying for B2B - real business owners are outnumbered by wantrepreneurs roughly 1000:1, so leading with value means accepting far lower view counts.

  Decide which audience you want: 10k legitimate businessmen who see you as a thought leader, or 1M low-intent followers - and map content strategy to the offer.

• Build an audience that wants to learn, not one that wants to be entertained.

  My framing: circus clown vs. authority figure.

  Most followers still won’t buy, but a learning-oriented audience contains serious operators, and I formed real business partnerships through my community.

  Chase virality first and nine times out of ten you end up with an audience you can only sell scammy offers to - content is just a distribution channel for products.

  One-year proof that no guru persona is required: 56.5k followers, ~100M impressions, multi-six-figure profit with zero flexing.

• Over 90% of the money from an X account is indirect.

  Distribution creates inbound opportunities, connections, and partnerships that dwarf direct sales - so optimize for algorithm-favored visibility even if the format feels cringeworthy.

  The real goal of posting is getting into private group chats and DMs, where the actual alpha and deals are; public content is the filter that earns entry.

  Every operator should run at least a faceless X account purely for the network - it makes hiring, partnering, and exchanging insider knowledge dramatically easier.

• The X networking playbook:

  1. Post actionable insights about your business model in an engaging way.
  2. Follow and interact with everyone in your industry.
  3. Move conversations to DMs and form an industry-specific group chat. Eventually you hold a high-leverage position - a large network plus people who look to you for advice - which opens monetization directions beyond your primary business.

• Become the authority by teaching your peers.

  If you want design clients, become the face of the designers, not a walking portfolio: people who want to learn the skill push you in the algorithm, which puts you in front of the people who pay for the service - and people buy from those others look up to.

  If you are not yet making money online, this is the whole path: get proficient in an in-demand skill, build a brand teaching it, and let clients come to you.

• A personal brand is a distribution asset, not the business - and it does not scale.

  You have one face, a fragile reputation, and 24 hours a day; anything relying on one personal brand stands on wobbly legs, and companies with more than one face for distribution exit far more often.

  The people printing hardest from personal brands are the operators behind creators, not the creators - survivorship bias hides them.

  The faceless-vs-personal debate is irrelevant: both are just distribution, and either works if the product is good (people insisting you NEED a personal brand often sell $10k programs where the brand IS the product).

  The most legitimate way to build one is to do impressive things first - the brand should be a side effect of achievements, not the goal.

• But route traffic through humans, not company accounts.

  People want to learn from and be led by a person; corporate accounts posting content telegraph sales intent, and the face of a service gets more inbound than the service account.

  Instead of paying influencers six figures for a synthetic launch pump lasting days, spend that budget turning your own team into thought leaders - distribution you own permanently and can point at every future announcement.

• To sell, occupy space in the audience’s mind daily across every channel.

  Be on their X feed, YouTube subscriptions, email inbox, Telegram notifications, and texts simultaneously; repeated daily value exposure is what makes them buy when you finally pitch.

Platform Strategy

• The traffic-quality map.

  TikTok and Instagram: low-quality, high-volume. YouTube: high-quality, medium-volume. Twitter/X: highest-quality, lowest-volume.

  X and YouTube long-form have the highest view-to-sale ratios, partly because both let you take users off-platform without friction.

  For organic B2B marketing, hyperfocus on just those two - they are sufficient on their own.

• Grow on X first, then propagate.

  X is the easiest large platform to build a following on, sets trends the others copy later, and a base there propels virality everywhere else - the Tate playbook.

  It is also the most stable platform: unlike TikTok/YouTube/Instagram you rarely face bans, strikes, shadowbans, or sudden algorithm shifts, so the audience asset is more durable.

  A bonus mechanic: your accumulated timeline of hundreds of small value posts is one extensive sales letter, building more trust than periodic long-form videos - which is why X creators can command more trust than YouTubers.

• YouTube is the most durable content platform.

  Videos keep earning for a year or more: one summer-2023 video still averaged ~$70/day in sales a year later; another product averaged $200/day at 95% profit from one video per week.

  For software, one YouTube view is worth roughly 50 reel views - 10 minutes of one person’s attention beats 15 seconds of fifty people’s; one of my brands has 140k YouTube subscribers vs 50k Instagram followers with YouTube revenue on the order of 100x higher.

  Anyone selling something targeted who is not posting on YouTube is watching money burn. The evergreen snowball is matched only by Google-ranked articles.

• Short-form is top-of-funnel, not a converter.

  Use TikTok/Reels/Shorts to give millions a first impression, then route viewers into email lists, YouTube, and podcasts where trust and buying intent are built.

  Short-form virality on a brand-new account is still achievable fast with a good niche and format - no warm-up needed: I edited a TikTok in 10 minutes, posted on a fresh account, and hit 120k views in 8 hours, with the leads routed straight to an email list.

  Awareness content and clipping campaigns are not a scam just because they don’t directly convert - awareness precedes conversion, and expecting cold short-form traffic to convert directly is a skill issue.

• Translation and cross-platform arbitrage.

  Proven English viral formats have rarely been tested elsewhere: translating viral tweets or money-Twitter content into languages like French goes viral in those underserved markets.

  Similarly, repurposing X posts to Instagram is one of the fastest brand plays - X concentrates the best thinkers while far fewer normies use it, so proven X content can ride from 0 to 100k IG followers.

  And post videos natively wherever a platform is boosting media: during X’s video push, media got roughly a 10x boost, and my ad benchmark from the period was $138 for 5M impressions on X ads.

• Hunt temporary platform quirks and exploit them hard while they last.

  Example: the period when YouTube polls reliably got 50k+ views and allowed links driving traffic anywhere.

  Loopholes are never sustainable, but a solid organic/paid base plus short-term exploits compounds fast.

  Algorithms change often - cheap tests beat assumptions (e.g., my Instagram experiment seeding 500-1,000 followers before the first reel, on the theory that follower interests and geolocation seed the algorithm’s audience targeting).

YouTube Mechanics

• A fresh channel’s first 6-10 videos get a free impression burst - treat it as the quality bar.

  YouTube evaluates click-through rate, retention, watch time, and session watch time on that burst and stops feeding impressions if metrics are weak.

  If none get pushed, the videos are not good enough - it is a test, not a warm-up.

• Search traffic out-converts recommendations ~4x.

  Someone searching “how to lose weight” has explicit intent; I make 4x more from YouTube search than browse - and search algorithms are much easier to manipulate than recommendation algorithms.

  Research high-intent keywords relevant to your product with vidIQ and put the keyword in the title; build presence on niche forums where buyers gather.

  A related play: monitor X for emerging AI trends and publish YouTube videos on them before search supply catches up - X surfaces trends days before mainstream demand hits YouTube search.

• Never link your YouTube videos from other platforms expecting reach.

  YouTube treats each discovery source - external links, search, browse/home - as a separate ecosystem with separate engagement metrics; off-platform clicks bring low-intent viewers whose poor retention hurts the video, and external traffic does not trigger algorithmic push.

  It helps only indirectly: external viewers who subscribe or engage later see your videos in their feeds.

  To diagnose underperformance, get data before advice: find where views actually came from, then check CTR and average view duration specifically among browse/recommended viewers - anything else is speculation.

• The permanent-VSL structure keeps pitching out of your content.

  Make one long video that acts as your standing sales letter and end every regular video by redirecting to it - this preserves the retention metrics of normal content while funneling every viewer to the sales asset.

  Extension for Shorts, which don’t allow comment links: attach a long-form video (the VSL, or even a 30-second bridge video with a CTA title) via the “related video” feature and put the link in that video’s description - it converts better than comment links because the button is visible without opening comments.

• Watch-time quirk: playback speed multiplies retention.

  A 1-minute video watched at 0.25x registers as 4 minutes of watch time - 400% retention - which viewer farms exploit.

  Grey-hat, but it reveals that watch time/retention is the metric YouTube actually weights.

Own the Relationship, Rent the Discovery

• Use platforms purely for discovery; convert in owned channels.

  Every platform is the top of a funnel: X to Telegram to product; Shorts to long-form to newsletter to product.

  Pitching directly in social content suppresses reach - viewers click away when the plug starts, tanking retention and future distribution.

  And platforms can delete your access to your audience with one click (large YouTube channels get banned in waves): a ban should never touch your email or Telegram list - if your account dies, owned channels let you respawn within days.

• An owned audience makes virality partially self-serve.

  Move followers onto an email list or Telegram, send useful content daily to keep them engaged, then push that audience at each new post - the early engagement spike triggers the platform algorithm to distribute it wider, on any platform.

• Faceless funnels work.

  Faceless content pushing traffic into a Telegram channel and newsletter pulls people into your ecosystem without being on camera; once resourced, hire creators and diversify distribution rather than making videos yourself.

  Auto-DM funnels also still work for growth when done well - I used one on my own account; the technique’s bad reputation comes from spammers executing it poorly.

• Align with an existing community before building your own lane.

  Every mainstream influencer traces to one origin community: Tate from money Twitter, Adin Ross from NBA 2K, Logan Paul from Vine, Nelk from pranks.

  Connect your personality to a community people already care about, then branch out - creating your own lane from day one is how you stay invisible.

Platform-Specific and Grey-Hat Playbooks

• Instagram engagement-comment funnel:

  1. Create a digital product (guide, Discord community, Notion template).
  2. Post reels showing it off with a “comment X and I will send it to you” CTA.
  3. The comment flood blows the post up, because Instagram heavily weights engagement.
  4. DM the sales page to every commenter.

• Instagram 7-second-reel loop hack. Make a 7-second reel showing a pain point, overlay a CTA to read the caption for the solution, and write a long caption funneling to your product. While people read, the short video loops repeatedly - more loops mean higher retention and more algorithmic push.

• Snapchat giveaway loop:

  1. Create a Snapchat account (optionally seed with 10 US adds via Snap Maps for geo-targeting).
  2. Add 100-500 people from Quick Add.
  3. Post stories in a profitable niche (crypto, gambling, forex).
  4. Run giveaways where entry requires a story shoutout - each entrant advertises you to their friends, a viral loop.
  5. Funnel viewers to your paid community or offers.

• Localized Instagram town pages:

  1. Create pages for multiple small American towns (e.g. “wealthclub{town}”).
  2. Follow the followers of local businesses’ pages - residents recognize the town name and follow back.
  3. After hitting a follower goal, funnel the audience to your product.
  4. Scale by hiring virtual assistants to manage the pages.

• Clip-army playbook:

  1. Repost streamer clips to X.
  2. Funnel viewers to a clipping Discord with the angle “I make X with clipping, here’s how you can too.”
  3. Build an army of clippers.
  4. Rent the clippers to influencers who want Tate-style mass short-form distribution.

• Spectacle-jacking.

  Build content pages around every major public spectacle and funnel the attention to your offers - e.g. for the 2024 election, a Trump TikTok page and a Biden TikTok page running “who has more supporters / race to 100k” content that farms both fanbases into a public Telegram channel.

• Whitehat before blackhat.

  Do not attempt blackhat traffic generation until a whitehat organic content machine is already selling your product - bots, purchased accounts, and experiments usually fail and only make sense as an amplifier on proven product-content fit.

  When you do go serious, multi-account operations require a real stack: proxies, phone farms, anti-detect browsers, browser and emulator automation, aged account suppliers, warm-up procedures, and VAs.

  Hard-won specifics: never automate TikTok or Instagram uploads via browser (detected, zeroed; YouTube Shorts tolerates it), Selenium recommendations signal an amateur, buying aged accounts corrupts your geo analytics, and I have run ~50 accounts under one IP - multi-accounting is a grey area that mostly goes unpunished.

  At mass-market scale, AI-generated posts and account farms demonstrably work (bans are irrelevant when no individual account matters) - but not in expert niches where audiences can tell.

  Buying TikTok followers from an SMM panel just to unlock the 1,000-follower link-in-bio threshold did not hurt my reach, since follower count is not a ranking factor.

  But never buy engagement from public marketplaces: valuable engagement depends on who follows the engaging account and niche alignment - anyone selling engagement publicly has none worth buying.

• Blackhatworld is the one forum worth reading daily.

  Everything you need to learn social media marketing is there; forums and Discords are also where cheap skilled labor hides.

• Misc tools and picks.

  To clone a brand voice with AI, open a profile in an incognito browser - logged-out X shows only their top tweets - and have Grok write a quick scraping script.

  My 2025 niche pick for a large personal brand: travel vlogging - extraordinary lives bring extraordinary impressions, and the lifestyle itself is the content moat (as of late 2024).

Why writing matters

• Writing is the highest-ROI skill because most valuable skills are extensions of it: processing ideas, putting them into words, and formatting them so they make sense. Every person worth looking up to is a good writer, and persuasive writing is a muscle built through practice.
• Writing publicly compounds your intelligence. Two years of writing in public dramatically improved how I formulate thoughts. Learn things, then write about them so others understand, optimizing for value and clarity. The practice itself sharpens your thinking (2026).
• Marketing skill = applied psychology plus clear writing. Study three things: how people perceive value, how brains structure ideas, and how to write in a way that is easy to digest - not tools or buzzwords.

Writing for reach: simplicity and packaging

• Dumb it down - simplicity is the distribution advantage, not a compromise. The more complex a post, the less reach it gets: readers scroll at the first hint of mental resistance, and the algorithm punishes scroll-pasts. Don’t tailor to an experience level; write about principles that apply at every level in maximally simple language. Assume every reader is a distracted, dyslexic idiot: shorter sentences, simpler words, clear formatting. If a 10-year-old couldn’t read it, it won’t perform. Gate genuinely technical sauce behind a lead magnet where subscribers consume it deliberately, and package timeline content as instantly-graspable dopamine hits with “no input lag.”
• Packaging beats substance for reach. The same content titled with a mass-appeal, outcome-first promise consistently outperforms accurate, mechanism-first framing: “$1000/day method for beginners” beats “how to effectively funnel traffic to your email list from short-form content,” and “this prompt will change your life” reliably pulls 100k+ impressions where a sober description gets far less. My tested hook formula: zero context + an optimistic promise - a “this thread will change your life” opener pulled 29k+ views. The mechanism is no cognitive load plus a big implied payoff. Wrap valuable content in a mass-appeal hook and let the substance sit underneath.
• Structure is a multiplier on value; treat formatting like product UX. The most life-changing insight in an unstructured wall of text fails - knowledge only pays if it is easy to consume. The average tweet gets about one second of dwell time (Twitter is a lower-attention platform than TikTok), so line breaks, sentence length, and visual scannability decide whether the reader stays. Chopped-up, line-broken formatting also physically occupies more screen space, increasing read-through; long text embedded in an image drives high view-time on media.
• Inject the value proposition instantly. The reader must understand what value they get from a post immediately, with no room for interpretation - a few seconds of dwell time decide whether it goes viral or flops. Nobody will work to decode your offer.
• Reach scales with total addressable audience. My prompts that went mega-viral targeted the widest possible desire - people who want to change their lives. Niche-precise content trades ceiling for conversion quality; mass-desire content does the opposite. Ask how many people the piece is even relevant to (2026).
• Three-point checklist before posting a tweet (2025): 1) you find the topic genuinely interesting in the moment, 2) you believe people will find it valuable - meaning actionable and step-by-step, something readers can start doing today, not vague inspiration, 3) it is formatted cleanly enough that a 12-year-old could read it.

Copywriting & persuasion

• People pay to alleviate pain - write to the deepest layer of the desire chain. Nobody wants to lose weight to lose weight; they want to be attractive, and beneath that, to escape the pain of not being loved or respected. Like a person with an unbearable toothache, buyers pay a lot to end pain. Separate surface wants from underlying desires and attack the emotional trigger.
• Mine real customer language before writing copy. Use a deep-research prompt (run in a deep-research mode, not normal chat) that takes two inputs - avatar and pain point - and researches Reddit, Quora, YouTube/Facebook comments, and product reviews to extract demographic and psychographic insights (hopes, failures, prejudices, who they blame), current solutions and what buyers distrust about them, forgotten past solutions, and beliefs about outside forces worsening the problem. Deliverables: real quotes, recurring narratives, and a summary of the market’s worldview for story-driven copy. (Credited to Alen Sultanic via @buyerofmedia.) posted April 2025 · practice might be outdated
• No jargon for cold traffic. Explain concepts in ELI5 terms - abstract words like “cognition” confuse cold prospects and kill conversions. Save insider terminology for landers targeting audiences you have already warmed with educational content.
• Treat every piece of content like a VSL. If every YouTuber structured videos with persuasion and a conversion goal, there would be no broke YouTubers. Done wrong it hurts retention, but a balance exists in any transformational video - and each viewer of conversion-optimized content is worth roughly 100x more, which makes up for the drop-off.
• Learn copywriting by doing, not from books. Consuming copywriting courses makes you worse: pre-made frameworks override intuition. The best copywriters I know learned purely through trial and error - write, publish, watch what works, and let judgment develop.
• Build an extensive swipe file. Earning potential tracks with how many proven examples - ads, hooks, funnels, formats - you can draw on. Systematically collecting what works in your market compounds into faster, better output.

Authenticity as strategy

• 99 out of 100 posts should have no ulterior motive beyond real value. People copying a successful style fail because every tweet is visibly engineered to sell, and viewers in a sophisticated market know exactly what you are doing. Followers who trust your authentic value are the ones who buy when you promote roughly once every 100 posts. Authenticity is the only moat in a sophisticated market; the leased-supercar guru aesthetic is dying, and authentic value content now out-engages virality bait (2025).
• Write tweets in one unedited stream of thought, capped at about two minutes. Over-editing makes writing sound synthetic; the occasional typo is the cost of sounding real - a deliberate trade for volume and authenticity. I spend under 15 minutes a day writing tweets and sustain reach.
• Don’t systemize X. Post manually the moment you have an interesting thought, never schedule, never use AI to write tweets - automation directly contradicts what the platform rewards, which is live, authentic thought (2025). posted December 2025 · practice might be outdated
• Sharing 99% of what you know will not hurt your business. The whole audience-building protocol: when you learn something that makes your life easier, post about it - if it helped you, others will find value in it too.
• Turn every act of consumption into a trigger for creation. Watch a video, write an article; scroll Twitter, quote-tweet your take; read a book, tweet the best ideas. This flips you from consumer to creator and gives an endless free content pipeline.
• Ugly beats polished. Unpolished creatives routinely convert better than beautiful ones - 99% of buyers don’t care or notice whether something is AI-generated, and aesthetic perfectionism is low-ROI unless your brand is literally built on aesthetics. Only value and clarity matter (2026). Likewise, authentic low-production creator videos outperform over-produced brand content by orders of magnitude per dollar - polished productions peak at 3k views where paid college students making authentic videos do far better for a fraction of the cost. Optimize for perceived authenticity, not production value.
• Mask questions as authoritative statements to crowdsource expertise. Publish a confident guide containing the thing you’re unsure about; experts will flame you in the replies with detailed corrections. They get an ego boost, you get the knowledge.

Formats & platforms

• Short-form is a volume game; long-form is a quality game. Spam and iterate on short-form; polish and invest in long-form, where a single video keeps paying for a year.
• Judge content by ROI of effort versus impressions. A tweet written in 2 minutes that gets 50k impressions beats an Instagram reel that took an hour for 50k views - though for similar content, long-form video is worth more per view because it is higher-intent attention. For reach per unit of effort, text on X is extremely efficient.
• X originates, short-form platforms broadcast. Great thinkers are on X; great broadcasters are on IG/TikTok/YouTube. You can build six-figure months by repurposing ideas that originate on X into short-form for other platforms; faceless educational reels were performing extremely well (2025).
• Short-form marketing is a pattern interrupt. The entire job (framing borrowed from OnlyFans marketers) is snapping viewers out of the doomscrolling trance, then immediately directing them off-platform while you have their attention.
• Text beats video for info products about 80% of the time. Text is easier to skim, structure, revise, and drop into an LLM; writing gives more time to reason and organize, which is why a book is more structured than a YouTube video. Reserve video for the ~20% of content that is hands-on UI navigation (n8n tutorials, design, media buying). Ninety percent of educational YouTube videos compress into a six-line tweet. Ship a markdown file per course lesson so buyers can feed it to an LLM. Sophisticated learners now skim a video and send the link to an AI (e.g., Google AI Studio) for a rundown, so video is trending toward pure entertainment while AI absorbs the education role - match format to audience sophistication: high-level audiences prefer text plus extractable snippets, beginner biz-opp audiences still prefer video for the dopamine (2025). posted December 2025 · practice might be outdated
• Choose medium by audience: text filters, visuals scale. Written marketing works in smart-money circles (Twitter, newsletters) because comprehension requires baseline intelligence and interest; for mass appeal go visual - even a child instantly understands an image, while the masses will never parse a screenshot of a prompt.
• Whiteboard thumbnails work. A whiteboard densely covered with information is a proven thumbnail trigger - visible information density promises value before the click. Usable for YouTube thumbnails and ad creatives in educational niches.
• Reddit-TTS TikTok product placement playbook:
  1. Write a “fake” Reddit story that is genuinely engaging.
  2. Plug your product inside the story as the fix for a common problem.
  3. Render it as a Reddit TTS TikTok over Minecraft parkour gameplay footage.
  4. Funnel viewers via link in bio. The story format smuggles the pitch past ad-blindness. posted July 2024 · practice might be outdated

AI-assisted writing

• Never let AI fully write audience-facing posts. Heavy AI users instantly recognize the tells (“the f*cked part? they didn’t pay a dime”, “here’s the truth no one is talking about…”) and it destroys credibility; on AI-focused X, hand-written accounts consistently outperform those publishing AI-generated posts (2026). AI lacks intuition, and in personal branding the human element always wins. Use AI to think, not to speak.
• My content process: manually braindump the ideas, let AI format them more cleanly, then rewrite in your own voice - and log every change you make plus why into a brand-voice markdown file the agent maintains. You cannot AI-generate good advice; the substance must come from your brain, AI only helps with clarity (2026).
• AI-era copywriting = four components in a prompt: context (the topic), conditions (rules the LLM must follow), brand voice (the writing style), and bias (a lens that skews output in your favored direction). Copywriters evolve into copywriting-focused prompt engineers - and only someone who is already a good copywriter can write a good copywriting prompt. Encode domain expertise and explicit best practices into the prompt rather than letting the model infer what “good” means; done well this one-shots consistent sales copy. posted June 2025 · practice might be outdated
• AI replicates a specific person’s voice indistinguishably - if you build a brand-voice profile. Robotic ChatGPT output just means no voice profile was used. Collect a corpus of the target voice, extract its patterns into a reusable profile, and attach it to generation prompts (proof point: a fully AI-generated tweet from my generator drew a reply from Elon Musk). One application: a tool that takes a braindump and formats it into digestible content in your own brand voice. posted March 2025 · practice might be outdated
• Script-profile playbook for replicating a YouTube style with AI:
  1. Find a well-scripted video in your niche.
  2. Get its transcript with any YouTube transcript extractor (or paste the link into Perplexity).
  3. Send the transcript to AI with a script-profile JSON template covering: hook type and duration, pacing and segment structure, storytelling approach and arc, tone shifts and emotional triggers, humor type and joke frequency, vocabulary complexity and catchphrases, retention/engagement points, and SEO elements.
  4. Use the filled profile to instruct AI to write new scripts in that style. Limitation: transcript-based extraction cannot capture visual elements. posted March 2025 · practice might be outdated
• AI content is bifurcating into two viable lanes (2025): mass-produced slop for cheap traffic at scale, and genuinely high-quality art. Everything in between gets ignored - pick a lane deliberately; mediocre AI-assisted content has no audience.

Platform Payout Reality: Never Build for Ad Revenue

• Platform ad revenue is a rounding error next to what the same audience is worth through your own offers - never treat payouts as the business model. Monetizing views with your own product is worth 10-50x ad revenue (a channel averaging 2k views/day would earn ~$10/day in ads but made $200/day selling to the audience), and directing X traffic into a solid offer earns 50-100x the payout. Even the high end of long-form YouTube (~$10 RPM per 1,000 views) is tiny compared with real monetization: my creator network earns thousands per video via brand deals on channels averaging only ~10k views (2026-08). YouTube automation creators celebrating $1-2k per 500k views are leaving 50x on the table.
• X payout benchmarks: roughly $1,000 for 20M impressions (~$0.05 RPM, 2025-01); ~$178 for 5M impressions over two weeks (~$0.03 RPM) despite a 90% tier-one male 18-44 business/AI audience (2025-03); ~$1,300 for 20M tier-1 impressions (2025-04); ~$50/week from ~2M impressions and ~$1,500 total lifetime ad revenue on an account that made six figures through products and connections (2025-03); ~$300 for 1.5M impressions in one payout period, ~$1k/month from millions of monthly impressions at peak virality (2025-11); lifetime, roughly $0.25 per follower and about a $0.14 CPM - $1-2k/month at 60k+ followers (2025-10).
• X payout mechanics: earnings are driven by engagement from Premium (verified) users on your posts, not raw impressions, and rates fluctuate with platform changes. The March 2025 payout-system change raised payouts 5-10x (fewer creators sharing the pool after X demonetized low-quality accounts), making ~$3k/month achievable from consistent posting at ~50k followers in the AI/business niche (2025-07), still incomparable to the value of the eyes themselves.
• The real money from an X audience is leads, partnerships, and deals closed in DMs - treat the account as a deal-flow asset. I projected over $1M/year from Twitter leads versus ~$50k/year from ad revenue at the same activity level (2025-05); most of my account’s income comes from partnerships formed in DMs, and ad revenue is a small bonus for doing your own marketing.
• X/Twitter automation (systemized, human-quality monetized accounts, not AI slop) is a viable business: five-figure months are realistic (2025-07). Accounts must serve a dual purpose - pushing traffic to funnels and offers plus payout revenue; my test mentees received a $3k wire from X within 1.5 months, and political/culture accounts farming ad revenue also reach five-figure months with a solid network.
• Sponsored-post benchmarks: startups pay $1,000-$2,000 per tweet to mid-sized accounts, ~$2,500 for larger ones (2026-01); all-in AI-niche accounts get $1k-$5k per tweet, and one large crypto account was paid $25k for a single tweet (2024-09/2025-09). Niche selection sets the rate card - but off-brand deals erode the trust that makes the account valuable; I turned down ~$200k of promos in a year and only promote what I actually use.
• Payout arbitrage windows recur - the Spotify template: commission cheap content (Fiverr songs), distribute via DistroKid, buy streams against a ~$4-per-1,000 payout. The loophole is closed, but the transferable pattern stands: find a platform paying per unit of consumption, source content below cost, arbitrage until patched.
• Ragebait pays because X pays for engagement, even angry engagement: a fake persona page engineered to trigger a tribe can farm a couple thousand dollars a month at modest scale - useful for understanding why outrage floods the timeline, not a recommended model.

Funnels & Owned Channels

• Route traffic through an email list or community before the sales page - direct-to-sales-page traffic wastes 99% of visitors. If 1% converts on a direct page, 99% are lost forever; if 20% join your list and 10% of those eventually buy, you make ~4x more from identical traffic and keep the ability to re-contact everyone. I wasted literally millions of visitors before learning this.
• The standard info funnel: valuable social content → free lead-magnet/newsletter opt-in → email automations ending in a “Buy Now” CTA → low-ticket paid community → high-ticket mentorship or service. Discovery happens on socials; conversion happens in owned channels; each tier qualifies buyers for the next.
• Monetize from the very first views - never “build the audience first.” Revenue has nothing to do with follower count (I converted a sale through a YouTube channel with 6 subscribers); every unmonetized view is wasted. If a project gets no sale within the first few thousand views, kill it or change the approach.
• Kill-or-iterate rule for promotional content: if a short-form video promoting a low-ticket product hits ~5,000 views without at least 1 sale or a set number of opt-ins, it’s dead. Cycle formats - one short-form style, then another, then long-form, then selling directly on Twitter; even total failure teaches more than brainstorming.
• Email is a major, underrated channel: ~30% of my digital product sales come from email marketing, with a 55% open rate against the “email is dead” claim (2025-01). But test Telegram too - my Telegram channel converts better than my email list in my niche, and YouTube Shorts funnel well into Telegram.
• Organic channels ranked by buyer intent, highest to lowest: SEO, YouTube search, newsletter, Telegram channel, YouTube recommended, X, Instagram Reels/TikTok, YouTube Shorts (2025-08). Search-driven and owned channels convert best per view; short-form feeds deliver volume at the lowest intent.
• Send high-volume ephemeral traffic (TikTok slideshows, viral spikes) into a newsletter, never onto a one-shot offer. I once sent 1M clicks to a free-Robux CPA offer; when the method got patched I had nothing left. Owned lead capture turns ephemeral traffic into a durable asset.
• A concrete quit-your-job benchmark: 100 new people into your Telegram channel every day. Focus on a repeatable daily acquisition number into an owned channel, not vanity follower counts.
• Traffic-and-product pairing rule: paid traffic suits physical products; organic traffic suits digital products. Digital margins survive organic volatility; physical unit economics can absorb paid acquisition. I made 7 figures on purely organic traffic without ever running an ad - and paid ads are best used to collect leads, not sell directly.
• Build a working tool as your lead magnet instead of another PDF: e.g., a free tweet-generation app where users plug in their own API key (zero running cost to you) and must join the newsletter to sign up. A functional tool converts and differentiates far better than an ebook.
• Sell the lead magnet for a token ~$15 instead of giving it free - not for revenue, but to filter info-hoarders. Pulling out a card signals intent to execute; I call it a way to “100x the value of leads.”
• When auto-DM lead-magnet delivery breaks at scale, require an email-list signup for delivery - the better list-building outcome anyway. After a 700+ reply thread glitched my auto-DM tool, switching to “subscribe and receive it in your inbox” converted the spike directly into owned subscribers.
• Evergreen YouTube + automated email = sales for years with zero maintenance: 1) create a digital product (course, guide, Notion template); 2) set up a related lead magnet with a clean opt-in page; 3) build an automatic email sequence that sells it; 4) film ~100 evergreen videos funneling to the opt-in; 5) post and forget. Years-old videos still produce sales on products I’d forgotten; a couple of videos about one product made ~$1,000/month passively for months.

Info Products: Sell Transformation, Not Information

• People buy transformations, not solutions - frame every product like the pay-to-win route from the prospect’s current state to their desired state. Info outsells software because “go from zero to $10k/month” beats “automate your outreach”; software only becomes an easy sell with a get-rich-quick element, and the new B2C app wave applies the same mechanism by attacking insecurities. Supplements and digital products are marketed identically - by instilling belief in a life-changing outcome, not pushing a tangible item.
• In saturated niches like fitness, the real product is identity reframing plus enforced accountability, not knowledge. Everyone knows calories-in/out and progressive overload, yet 99% fail; all fitness knowledge is in a free 20-minute YouTube video. Clients pay for reframing into an athlete identity, eating for fuel, daily training, and religious data tracking - package coaching around enforcement and check-ins (“transformation accountability groups”), not information.
• Real, instantly applicable methods are the edge - most courses and ebooks are theory-heavy. Documented, practiced, immediately actionable process is the moat; selling methods you haven’t applied yourself is essentially scamming.
• Emotional buyers convert: angles targeting insecurity relative to peers dramatically outperform neutral angles. “Become faster/stronger than your teammates” far outconverted generic promotion of the same athlete workout program; people impulse-buy in emotional states, which is also why insecure audiences (self-improvement hyper-consumers) have extremely high purchase intent for anything offering an illusion of progress - an ethically double-edged fact.
• For men, map the offer to survival or reproduction: status, safety, money, or attractiveness - never abstract features.
• Trends supply fresh “mechanisms” to anchor offers to: a credible new angle (e.g., the peptide trend in fitness) restores hope in prospects who already failed with mainstream paths and re-opens the sale. Watch for new research and novel angles to build offers around.
• Position yourself between an existing belief and its believers rather than creating demand: large belief tribes (diet camps, etc.) already have millions of followers; when your advice lines up with what they believe, they gravitate to you.
• Stop calling it a “course” - use “workshop,” “cohort,” or “playbook.” The word triggers a conditioned negative reaction in mass audiences; “$10k/month SWE accelerator” and “software engineering bachelor’s degree” describe the same thing with opposite reception.
• Deviate from the standard info-product landing page template - sameness pattern-matches to “scam.” Viewers have clicked the identical copy-paste layout 50 times; breaking the pattern is itself a conversion advantage.
• Sell in the niche you actually know, even if it’s not the most lucrative on paper: the #1 athletic performance mentor makes more than the #1000th sports betting tipster. Depth and position within a niche beat raw niche size.
• When AI has all the world’s information, sell the lens, not the info: tools, curated context, and filters that encode judgment and make AI output the right information for a specific goal. An LLM always gives a correct answer but never the best one - which is why communities, consulting, courses, and mentorship stay sellable. AI coach apps trained on your course material (available in Whop’s app store) upgrade delivery while your proprietary knowledge stays the moat (2025-10).

Pricing, Offer Structure & High-Ticket vs Low-Ticket

• Low-ticket info is the front end; the real money is the back end - SaaS, high-ticket services, done-with-you offers, licensing, or B2B. Relying on info alone is clueless: use $15-$50 products to acquire and qualify buyers, then ascend them. In the AI wave specifically, low-ticket courses surface and warm B2B leads - buyers with real budgets can tell whether you actually know your subject (2025-04). If you sell a high-ticket service, sell low-ticket info that nurtures buyers into realizing they need the service.
• Hold the front-end price low and stable instead of raising it: a cheap front end makes the backend upsell look more attractive by contrast. The front end is an acquisition and qualification tool, not the profit center - a $27 offer only matters in light of the backend.
• Don’t sell static info products above roughly $500 - the most ethical structure is low-ticket playbooks plus a high-ticket done-with-you offer. Static-info buyers aren’t pre-qualified and most never act, so premium prices for unaccompanied information are exploitative; my coaching’s 100% success rate comes specifically from hand-held execution.
• Sequence services before info products: high-ticket done-for-you/done-with-you sells on results rather than audience trust, pays far more per sale, and doesn’t need months of audience-building; the info product later becomes a lead magnet for the service (2025-11). If you can’t fulfil the service, you shouldn’t sell the info either. The biggest earners on X sell services - info is the bonus.
• Low-ticket at scale is not beneath you: Hormozi did millions selling a $29 book through a livestream (reviving free-plus-shipping mechanics) while gurus consider themselves too sophisticated for it. Low-ticket buyers trickle upward - I have done seven figures from low-ticket alone with a traffic engine feeding high-ticket. Watch mainstream gurus to analyze their funnel mechanics, not their information.
• Farm $17 impulse purchases through short-form before building upsells: show a transformation in a niche, sell the cheap “how I did it” guide, and treat its real job as getting buyers into your ecosystem.
• Cheap add-ons roughly double revenue - stack core product + community chat + educational material + embedded tools under one platform (Whop lets each be its own app); the stack immediately raises perceived value. Purely info-based communities are dead: teach people to solve a problem, put them among people solving the same problem, then give them tools that make solving it easier. With a coding agent and Whop’s CLI you can one-shot custom community apps - even replacing the default course app (2026-07).
• Offer buy-now-pay-later on high-ticket before optimizing anything else: ~30% of high-ticket sales on Iman Gadzhi’s challenge Whop close through the Splitit BNPL option - easily seven figures that wouldn’t otherwise exist (2025-09).
• Prefer revenue share to flat fees when you can actually deliver; hybrid (fee + revshare) secures commitment. A vendor who can genuinely 2x+ your revenue would be irrational to take a $5k/month retainer - so retainer-only “transformational” agency pitches are a red flag, and as a capable seller you should structure revshare (e.g., 33% of profit for 3x-ing revenue). Caveat: revshare only makes sense where you can actually move the whole revenue line, not for commodity services.
• A simple mental model for $100/day: create a $100 digital product and sell it once per day. Units-per-day of one offer makes revenue goals concrete and testable.
• Valuation sanity check: a $1M offer on a $12k MRR indie SaaS (~83x monthly revenue) is a no-brainer sell - beyond the cash, the exit goes viral, builds clout, and hypes your next product (2025-04).
• Build once, sell twice: prioritize assets with zero marginal cost of reproduction - products, content, tools, templates - over anything requiring the work to be redone for each new dollar. Passive income always means someone or something working for the money: a $10k/month client fulfilled by a $5k/month hire is $5k of passive margin - delivery arbitrage, not magic.

Selling Without Sales Calls

• The public hates sales calls - a visible “buy now” button can outperform a booked-call funnel, and no deal under $2k should ever require a call. I earned seven figures without taking a single sales call or making a VSL; as a buyer I won’t book a call for anything under ~$2k either. When a prospect asks the price, “let’s hop on a call” is the fastest way to get blocked - just state it.
• High-ticket should close in a short DM exchange (~10 messages), not a 30-minute convincing session. Needing to hard-convince a lead signals the offer or messaging is weak - the better the product and messaging, the less persuasion required.
• The best cold DM is a plain statement of what you do and exactly how it benefits the recipient: “hey, I do XYZ, {how it benefits you}” beats every script. Obvious sales tactics kill B2B deals with sophisticated buyers - part of the job is judging whether the prospect will see through canned techniques.
• State your non-negotiable price up front before explaining scope: it filters time-wasters expecting the offer at 10% of its worth - it costs some raw conversion, but it saves enormous back-and-forth.
• For warm organic traffic, plain copy and a buy-now button beat hype and long VSLs. If the audience is warmed by your content, the emotional persuasion is already done and a 60-minute unskippable VSL only adds friction; long VSLs still work on cold traffic in markets like healthcare or insurance. Match funnel aggressiveness to traffic temperature.
• The minimum viable online business stack: landing page + buy-now button + traffic. Don’t overbuild funnels before proving traffic converts on the simplest possible page.

Paid Communities

• Never run a free group chat - charging even $1 filters low-quality members and improves conversation quality roughly tenfold. Free chats fill with unserious members who drive away serious ones; an unmoderated free-chat section in a paid Discord actively costs sales, and a chat full of time-wasters signals a low-level offer to high-intent leads. Keep a team curating chats on topic. Price-gating valuable info is a feature, not a moral failing: free material learned by 10k people at once loses its edge.
• Telegram vs Discord - choose by offer: Telegram suits older, higher-IQ, high-ticket audiences (more notifications enabled, ~4x the monthly active users, native advertising, easier high-ticket sales); Discord suits younger, low-ticket, high-engagement communities (far better channel structure, good for finding cheap labor).
• A good customer support rep is the single highest-ROI hire for a paid community owner - support quality drives retention, and churn (especially in degen niches) is the real killer.
• Paid communities make most of their money from member data: segment members by situation and goal, then route each segment to the appropriate next offer instead of treating the community as one undifferentiated audience.
• Automate upsells inside the community platform (Whop automations app): 1) launch a free course/community as top of funnel; 2) set engagement milestones that trigger automatic DMs; 3) present a problem in a video, then DM the solution as an upsell to the paid service; 4) segment engaged vs unengaged members; 5) use engagement metrics to predict churn (2025-10).
• Mundane hobby and unconventional niches print on Whop: $38k MRR teaching pickleball, $60k MRR teaching cycling, a “women empowerment” community past $1M in sales - stop overthinking niche viability and launch. Note: review counts are not a quality or revenue signal; whops making millions can have zero reviews.
• Info-and-tools education model: keep all info plus most tools in one affordable subscription (mine: $75/month) and reserve high-ticket pricing for B2B (whitelabeling your tools and AI-integration services) rather than selling high-ticket “mentorship” to individuals (2025-04).

Niche Selection & Audience Quality

• You need the right views, not many views: a video with 10M views can earn $0 while 1k views of business owners with an expensive problem earn $6,000. A money-niche view can be worth more than 1,000 entertainment views; a ragebait fitness account with 1M+ views per video made nothing while my faceless informational channel averaging 5,000 views did five-figure months in software sales. Pick the niche for the value of the viewer.
• Niche profits beat virality: an 8M-subscriber YouTuber with a full team made $50k/month while 5k-subscriber channels pull $100k/month at 95% margins (2025-12). The more precisely content targets high-net-worth problems, the lower its view potential - and the higher monetization per viewer. Small proof points: a 4,000-subscriber trading channel (never even YouTube-monetized) made me $100k-$150k lifetime from product sales; a 5k-subscriber fitness YouTuber averaging 1k views runs a $60k/month coaching program because every video is proof-plus-process (real client transformation, exact meals/weights/steps over spreadsheets) instead of fluff.
• Trading is by far the best niche to sell info in - serve markets where people already spend freely (trading, gambling, sports betting, bizopp). Most of my money over five years came from the trading industry without ever trading: I co-founded one of the biggest trading communities and sell to the market rather than participating. Where money already flows, you can also facilitate deals and take a cut. A network of trading-niche Instagram pages attracts constant monetization deals - “it’s literally a skill to stay broke with a constant flow of trading traffic.”
• Match platform to audience sophistication: short-form and YouTube audiences are less sophisticated and easier to sell info/mentorship to; Instagram is wantrepreneur consumers; X is sophisticated operators and the top platform for high-quality leads. On X you’re effectively selling to the gurus themselves - weak offers get seen through instantly, but genuine offers make serious money ($50k of high-ticket closed in two weeks off a single Typeform). “Twitter doesn’t convert” means you’re selling air to a sharp crowd.
• X monetization works if you have real expertise to sell - and is a terrible place to get rich without it. I made multiple six figures in my first year of posting via info products, services, and coaching; my first launch did $37.5k at 99% profit in one day from organic Twitter, ~30 days after my first tweet, by giving away nearly everything free so the people who got value paid for structure. A 140k-subscriber YouTube channel in a high-value niche wasn’t much more profitable than my X account; if you’re remotely successful in any vertical, posting on X can roughly double your income.
• If you have 100k+ followers and only $10k MRR, the problem is strategy, not reach. Followers convert when the audience matches the product’s ideal customer profile and the product is good; design content strategy backward from who buys. The audience your front-end marketing attracts is the pool you pitch on the back end - flex/lifestyle marketing attracts the “vulnerable class” while value-plus-trust with a direct checkout attracts businesspeople; serial flexing now reads as scam signaling even to normies. The quiet $300k+/month operators share two traits: authority built on a genuine specialty, and videos that function as VSLs for their products.
• Distribution is only one piece: what you sell and how you sell it can be the difference between five and seven figures through the exact same channel.
• A personal brand also monetizes as pure connection: matching people who need something with people who provide it, with no fulfillment - deal flow arrives in DMs and referrals become a revenue stream of their own.
• Making $1 online is a transferable proof-of-concept: a friend’s $27 lead-scraping Chrome extension promoted with four UGC TikToks a day (~1k views each) made ~two sales ($54)/day passively - a system that earns while you sleep can be scaled 100x, and beginners wrongly dismiss it as too small.
• If you can sell info products you can sell anything: convincing cold audiences to pay for intangible value in a hyper-competitive market builds sales muscle no other product forces - and if you can sell to cold traffic, you’ll never be poor.

Full Playbooks

• The complete minimal online business (2025-03):
  1. Find a validated product or service, replicate it, and make it better.
  2. Build a landing page with bolt.new with a prominent “buy now” button; take payments through Whop.
  3. Find a content style in your niche that demonstrably gets views and double down.
  4. Build an email list with an automated sequence that sells for you.
  5. Move content traffic to the list with a genuinely high-quality free lead magnet.
  6. Send daily value emails and pitch the product once a week.
• $0-to-$10k/month insecurity playbook (2025-08):
  1. Pick a profitable insecurity you have real knowledge in (obesity, lack of athleticism, dating, skincare).
  2. Design a gamified info product promising a clear transformation in a set timeframe (“3-phase superhuman athleticism program,” “90-day fat-to-shredded sprint”); price at $15-$50 - cold-traffic sales are emotional impulse purchases.
  3. Build a sales page tapping the emotional pain and highlighting the transformation via a “unique mechanism,” never the hard work.
  4. Research organic formats that convert - browse Reels for entertaining-plus-educational trending formats in the niche (study Peter Peptides-style videos).
  5. Post ~10 videos/day soft-shilling the guide; monitor engagement and conversion, and adapt. The edge is execution volume and metric-driven refinement, not secrets.
• The “level 0” baseline business (2025-09): 1) faceless TikTok page around a common painful insecurity (finance, dating, appearance, health); 2) build an email list and send genuinely useful advice; 3) create an offer that resolves the insecurity (course, PDF, done-with-you); 4) direct-response landing page; 5) launch to the list, monitor conversions, adjust pricing and copy, iterate. Don’t attempt any other model until this works.
• Faceless niche-audience content formula: point out a pain the audience has, deliver value for 90% of the video, pitch your info product as the shortcut in the last 10% (e.g., a self-improvement channel targeting lack of discipline pitching a “90-day self-improvement challenge”).
• Influencer piggyback method: 1) find an influencer with monetizable expertise (Huberman, Knees Over Toes Guy, Paul Saladino); 2) create the info product their audience wants (e.g., a testosterone protocol for Huberman viewers); 3) repurpose and mass-post their clips to TikTok, Reels, and Shorts; 4) attach a ~3-second CTA to your product at the end. I made a quick $10k selling an athletic workout program this way at ~$7 per 1,000 views; main risk is takedowns.
• Faceless trading-brand model: hire a profitable trader as the face, build the infrastructure and team (script-writers, editors, thumbnail designers) yourself, split profits 50/50 - I ran one to 6 figures in a year fully faceless. Generalization: an elite marketer earns more as the brain behind a team of creators than as the creator; many of the richest social operators are publicly unknown. My evolved Whop creator-partnership version: own a small revenue-share percentage in a few creators in exchange for setting up their offers and operating their growth - ~$15k/month for six months with practically no ongoing work (2026-03).
• Multiply yourself through operators: once genuinely good at something, invest in other people’s development in that skill and take a percentage of their profits - but get ownership or revshare upfront and structure mutual dependence, or you’ll be cut out once no longer needed.
• Telegram crypto-channel ad-revenue playbook: 1) create a crypto Telegram group; 2) repost memecoin calls from public accounts (optionally automated across 10 channels with a repost bot); 3) drive traffic via TikTok slide spam or a crypto YouTube channel; 4) at 1,000 members Telegram monetization unlocks and projects buy ads; 5) cross-promote across your channel network. Benchmark: one 100k+ member channel made $400/day from ads alone, with a private paid calls channel sellable on top.
• Reddit/Quora SEO for products: 1) post questions your product solves (verify they rank on Google); 2) switch accounts; 3) write an in-depth answer that plugs the product; 4) boost it with an upvote service. On Quora: 10 aged accounts, Google-ranking questions about products with affiliate programs, answer with your affiliate link, buy upvotes from an SMM panel.
• Paywall-and-leak affiliate PDF: 1) create an educational PDF guide; 2) fill it with affiliate links to the tools needed to execute the method; 3) publish on Gumroad behind a paywall (direct sales don’t matter); 4) post it on leak forums titled as a leak with the dollar “value” in the title; 5) collect affiliate commissions.
• College community pages: 1) create an Instagram page per major American college; 2) hire a broke student on each campus to run it; 3) grow via giveaways requiring a story repost to enter; 4) charge local businesses and event promoters for promotion.
• Robux offer-wall arbitrage: 1) build a site where Roblox kids complete CPA offers (surveys, videos, app installs); 2) pay them in Robux with a VA automating payouts; 3) advertise through Roblox creators. Children’s game economies hold more money than assumed - my teen “free Robux” content lockers paid $1.50 per conversion at $1.7k/day peaks.
• Broke-beginner UGC retainer play (2025-03): 1) produce promo videos for SaaS and consumer apps to build reps on camera and in editing; 2) build a public catalog and ideally a following; 3) pitch founders on being the face of their brand for a monthly retainer with unlimited content. The wedge: founders hate being on camera and per-video UGC pricing is expensive; the catalog is proof you can already do the job. Adjacent 2026 entry point: spawn new distribution channels for funded startups (YouTube, X campaigns, IG, LinkedIn) - they have huge growth budgets and reward anyone who demonstrably attracts attention.
• Beginner affiliate path: start by affiliating for a proven digital product instead of creating your own - set up social pages in its niche, collect leads, refer the product. A few thousand dollars a month, plus education in what sells before you build your own. “Sell info without info”: ask established paid-community owners on Whop to affiliate - most give a 50% revenue-share link and even teach you traffic, since it’s free distribution for them.
• Local events business: hosting parties is underrated - build local Instagram/TikTok pages that own a city’s nightlife audience, monetize the events, and let reputation in those communities compound as the moat.
• CPA/sweepstakes is a bad monetization but an excellent training ground: sweeps means fighting ToS for tiny conversions (~$8; email submits ~$2), but my teenage CPA years taught me most of my marketing - graduate to your own products.

Platform Mechanics & Payment Infrastructure

• Payment-processor lockup is a when-not-if risk: Stripe froze my account with $100k pending and then held 25% of my balance; PayPal is worse. For info products and communities I now host my own landing page and process through Whop - support actually responds (an issue was fixed almost instantly via DM), fees dropped, high-ticket checkout links exist, and BNPL plus content-reward clipping programs are built in. Don’t concentrate all revenue in one processor; a peer paying 18% platform fees elsewhere was getting badly overcharged.
• Whop Discover mechanic: publishing to the Discover page costs a 30% affiliate commission on buyers who find you there - if you bring your own traffic, simply opt out and keep the margin (2024-12). Whop’s referral economics also make a durable side play: refer businesses to process payments there and pocket a percentage of their fees and ad spend forever (2026-08). Whop and Gumroad are the starting platforms for digital products.
• X is unusually friendly to off-platform links - exploit it. Most platforms suppress external links; X doesn’t care, which is why it has one of the highest view-to-sale ratios. Put your Telegram (or other) link under your tweets. Where a platform does dislike Telegram links, use YouTube as a link cloak: put the Telegram link in a video description and share the YouTube link.
• Target low-competition, high-intent YouTube search keywords with VSL-structured videos: searches like “how to scrape leads from google maps” or “how to get clients for agency” - use your product in the video to achieve the titled outcome, then deliver the CTA. Fully AI-generated videos convert when 100% product-focused like a VSL: an ElevenLabs-voiced template, tweaked every few days and synthetically ranked for high-intent keywords, pulled $300-$500 per 1,000 views in the trading niche; the bottleneck is the limited pool of high-intent keywords. Without a personal brand, AI content must sell directly, not entertain.
• Direct-response short-form benchmark: ~1 sale of a $20 digital product per 1,000 views on a fully AI-made promo video. The trade-off: blatant promo converts but resists virality - conversion rate and reach pull against each other.
• Faceless scales further than assumed: I sold many types of digital products without ever showing my face, taking a sales call, or making a VSL - landing page, content-driven traffic, and self-serve checkout replace personal branding. Proof of simplicity: I set a friend up with a proven product format plus scripted UGC videos, and sales arrived within two days - execution of a known formula, not novelty. And simple products still work: a PDF (traffic-generation methods plus community access) did $10k in the first 2 hours and ~$35k on day one via an X lead-magnet funnel into Telegram, with most launch sales coming from Telegram, not the timeline.
• The biggest influencer money bridges entertainment and education: viral entertaining clips earn the reach, education justifies the product, the funnel connects them (Andrew Tate, HSTikkyTokky, TJR Trades).
• Controversy is a launch strategy: deliberately provoking a tribe that hates your offer gets outraged quote-tweets that would cost $1,000-$5,000 each as paid promos, for free - and thousands of haters become witnesses to future success stories (a course launch that enraged indie developers did exactly this, purposefully cringe to pump views before flipping the script).
• X platform pricing (2025-04): ~$200/month for the gold checkmark on a business account; $1,000/month for the ability to grant affiliate badges to associated accounts.
• AI service-upgrade example: a personal trainer keeps the core transformation program but adds an AI assistant analyzing client check-ins (weight, measurements, calories, progress pictures, lifting progression) via the Gemini API - daily trend reports plus visual form analysis from video - built with AI coding tools (Cursor/Windsurf/Replit), embedded on Whop, and paired with weekly calls, an accountability community, and educational material.

Ethics & Legitimacy

• Never market an info product with a promised income figure - no such promise has ever had a 100% success rate, so it cannot be made without lying. Buyer outcomes are dominated by whether the person can execute, not by content quality, so “make X per day” marketing is effectively false advertising; judge offers promising specific daily income as scams by default. Honest claims convert less initially but build a self-sustaining community with longevity.
• The line between legitimate info and grift: teach a verifiable skill and promise only tangible truths (“this method still works as of this date”), never guaranteed monetary results. A course teaching motion graphics is the same category as schools and non-fiction books; a guaranteed-income course is fraud.
• Judge an info seller by how they spend their time: a legitimate teacher still spends most of their time running the business they teach. Grifter signals: profit screenshots everywhere (the more screenshots, the stronger the grifter signal), teaching as the main occupation, and huge followings (bigger brand → more info income → a staler taught model). One-off launch days get dressed up as daily income - a $35k day-one launch followed by $8k becomes “$35k/day.” Legit sellers usually have smaller followings and modest flexing.
• Don’t sell low-ticket “growth” info to a mass audience - most buyers can’t succeed, which invites fraud accusations. Qualify leads and sell mentorship only to people who can actually win; screening applicants is only worth it for four-figure-plus offers. Mentorship outperforms courses on outcomes largely because of vetting: my mentees all succeeded partly because I screened out everyone needing too much handholding - turning most applicants away protects the success rate and the reputation built on it.
• Price reasonably, refund anyone unhappy, and make the information genuinely valuable - long-term trust outearns short-term extraction. The audience is the real asset; morals over profits keeps it intact. (Counterpoint: self-improvement creators deliberately keep audiences insecure because their business depends on it - don’t be that.)
• The cynical $100k-MRR consumer app formula exists - find a common insecurity, build an app that “fixes” it, market to impressionable TikTok audiences - the info-product psychology applied to software subscriptions; understand the mechanism even if you won’t run it.

Why affiliate first

• Traffic is the transferable skill; learn it before building anything (2025-01-17, reaffirmed 2026-06-17). For your first $10k online, skip creating your own products: pick a solid affiliate program and pour all your effort into learning to drive traffic and testing what converts. Successful affiliate marketers succeed at every subsequent venture they touch. Money follows whoever can reliably send converting traffic, regardless of whose product they sell.
• “Find something people buy, and help the sellers sell it” (2026-05-30). The simplest entry into online income: find a product already proven to convert (browse affiliate networks), learn to make content that influences people to buy it, and you can realistically replace a 9-to-5 within months. I started at 14 manually sending pay-per-install links to strangers on Snapchat at $1 per install, making ~$100/day. The entry bar is agency, not capital or skill.
• The simplest path to a few thousand a month (2025-06-20): find a valuable offer, contact the owner, create content targeted at their ICP, capture leads with a lead magnet, and send them over for a commission. It is protected by two filters: most people never understand the right things despite hearing them repeatedly, and most who understand fail on implementation. Selling someone else’s proven product is easier than building your own.
• Niche and lander matter more than traffic volume - same clicks, 10x revenue (2025-09-10). At 16 I ran ~20,000 clicks/day to a content locker farming $1 app installs (iHeartRadio, State of Survival on OGAds) for ~$1,500/day. A later product converted at ~$2,000 per 1,000 landing-page visits. The same volume would have produced ~$20,000/day, and pointed at a trading offer it would have cleared a seven-figure month. Maximize payout per visitor before scaling traffic.
• Sell shovels to crypto during a bull run (2024-12-19). The whole market runs on FOMO, so traffic converts easily; broker and trading-bot affiliate programs pay very well, so you position as a traffic source without needing your own product.
• The first few dollars are psychological, not financial (2025-08-21). Fastest first taste: sign up to an affiliate network with cost-per-install offers and send your link to friends. Expect bans and trivial amounts, but seeing money appear “from nothing” creates the motivational spark that drives everything after.

Clipping & UGC: the on-ramp

• Clipping is the ultimate filter (2025-02-08, reaffirmed 2026-07-03). It compresses online business to its purest loop - attention in, dollars out - with the conversion element removed. Getting 100k daily views on TikTok/IG is very achievable, making $100/day realistic. If you cannot clear $1k/month with the simplest possible model, no other model will save you; if you can, you have built the traffic skill every other model needs. Then graduate into affiliate marketing, where the same skill pays far more.
• Whop clipping/UGC pay rates (2025-03-03): roughly $1 per 1,000 views for clips, $2 per 1,000 for faceless UGC, $3.50 per 1,000 for regular UGC. Go with faceless UGC because it teaches scripting, recording, editing, and understanding what gets views; adding an affiliate link also teaches converting views into sign-ups. Do not buy info products at this stage - traffic skills are learned by doing.
• The payout curve is brutal (2024-12-25): roughly 1% of clippers take 95% of the payout. Do not treat clipping as an averages game - the play is to be top 1% on quality and volume, otherwise expected value is near zero.
• A beginner can be profitable within days with guidance (2024-10-21). A friend whose only skill was basic video editing affiliated for one of my products, learned to make promotional reels, and hit a $500 day ($250 profit) five days in.
• Zero-capital content play: Whop community breakdowns (2025-02-28). Whop pays $2-$3.50 per 1,000 views across platforms. Literally film your monitor breaking down interesting Whop communities - how much they make, how they acquire members - and post the videos, with a Whop affiliate link in bio to earn a percentage from creators you onboard. $100/day is the minimum bar for someone serious.
• Graduate into paid short-form partnerships (2025-03-03). Once skilled, partner with info-product sellers, software companies, mobile apps, or ecommerce brands that sell through short-form content, and take affiliate deals.

Finding & vetting offers

• Pick offers with performance data, not vibes (2025-12-17). Whop’s affiliate dashboard lets you search offers by existing affiliates’ earnings, conversion rate, and EPC (earnings per click), and view the seller’s promo assets. Treat any affiliate marketplace like a real network: filter for proven conversion metrics before committing your traffic.
• Talk to offer owners (2025-10-16). Picking the right whop to promote is the biggest variable in results; owners know what converts and how to position it - leverage relationships to find out.
• Check the incentives behind free “methods” (2025-02-24). Classic affiliate growth playbook: someone posts a free method (the BlackHatWorld example I ran into: promoting CPA offers on Snapchat via quick-add and story shoutouts), hundreds run it, and the poster earns from everyone sent to the network he was affiliated with (OGAds). I saw the same play running on money Twitter with a well-known affiliate network. A warning as a consumer - and a distribution tactic if you own the offer.
• Educate the market for free, supply the product through affiliates (2025-08-03). A TikTok creator in the peptides niche runs this model: content teaches people about the product category and builds demand; monetization happens through affiliates supplying the actual product. Education creates the demand; affiliates capture it.

Playbooks

• Clipping-and-affiliate at scale

  (2024-12-18):

  1. Pick a high-earning niche (trading, crypto, dropshipping).
  2. Spreadsheet all top creators in that niche.
  3. Build a catalog of 1,000+ clips from their content (or have a VA do it).
  4. Build a team of 10+ US-located clippers on performance-based pay - find them in “internet money” Discords.
  5. Send them the clips with editing instructions.
  6. Set up a newsletter with a free lead magnet (e.g. an e-book) and funnel video viewers into the email list.
  7. Once the list is sizable, approach owners of top-earning Whop communities for affiliate deals - many will give 50% commission. You build both a niche network and significant income.

• Beginner deal: content boost for promotion

  (2025-02-26):

  1. Go to Whop’s affiliate-offers page and filter to programs paying 50%+ commission.
  2. Pick an appealing program and find the owner’s socials.
  3. DM a proposal: you run a Twitter/X page promoting their product; they engage with your tweets to boost them in the algorithm.
  4. Once agreed, batch-create 10 lead magnets, post them, and pitch your affiliate link at the end of each.

• Broke-start replication with an owned channel

  (2025-09-25):

  1. Browse the Whop discover page by keyword to surface top-performing whops in high-value niches (trading, sports betting, bizopp).
  2. Confirm good affiliate commissions.
  3. Find their socials and reverse-engineer the organic strategy that is working.
  4. Create accounts replicating that content style, promoting the whop via your referral link.
  5. Crucial step - insert an owned “alternate channel” (email list or Telegram) into the funnel so unconverted traffic is stored instead of lost.
  6. Promote the affiliate link inside that channel alongside related valuable content. The owned channel later becomes the launch asset for your own product in the same niche.

• YouTube-to-X transcript engine

  (2025-12-02) - done right this clears $10k/month:

  1. Find whop creators selling specific knowledge (trading, marketing, business).
  2. Reverse-engineer their distribution - traffic almost always comes from 1-2 channels, most commonly YouTube, paid ads, or X.
  3. Isolate whops driving traffic from YouTube that are not posting on X.
  4. Extract all their video transcripts and have Gemini distill the key teachings into a bullet-point list.
  5. In Claude, use the built-in skill-builder to turn that list into a “tweet idea generator” skill.
  6. Create an unbranded X account for the niche - branded accounts flop on X.
  7. Use the skill to generate lead-magnet ideas and run them as “reply for X” automations.
  8. Critical step: pay an X account in the same niche averaging 5k+ views to engage with your first tweets - that initial boost is what lets posts take off.
  9. Ongoing system: one lead magnet a day, affiliate link in each magnet, close people in DMs for extra revenue, funnel everyone to an email list or Telegram channel.

• Language-arbitrage promo videos

  (2025-02-07):

  1. Find a YouTube video that successfully promotes an info product.
  2. Download it with cobalt.tools.
  3. Get the transcript via youtubetotranscript.
  4. Translate to Spanish with DeepL.
  5. Generate a voiceover of the translation with ElevenLabs.
  6. Overlay and sync the new voiceover on the video.
  7. Get an affiliate link for the same product and put it in the description.

• AI slideshows on TikTok

  (2025-06-03) - a cheap, delegable format that was ripping:

  1. Use ChatGPT to generate an emotional, engaging storyline that indirectly promotes your product.
  2. Have it split the story into multiple slides.
  3. Generate background images with gpt-image-1.
  4. Delegate production to cheap labor and post at high volume daily.

AI angle-mining for promotion ideas (2025-04-13): use Perplexity to download transcripts of videos that appear to promote products like yours, define a JSON profile template of what to extract, and send template plus transcripts to Gemini. No need to verify each video manually - let the AI filter which ones actually promote a similar product. Output: a bank of long-form promotion angles.

Long-form YouTube: the compounding channel

• Case study: $450-$600/day in commissions within 15 days, 100% margin, all organic (2025-10-16). Faceless, outsourced (not AI-generated) long-form YouTube videos promoting a Whop community: first sales within 12 hours of the first video, ~$500/day at day 3, stabilized at $450-$600/day by day 15. Method: find a community already converting, make videos about the problem it solves, push traffic to the landing page with your affiliate link.
• YouTube affiliate content becomes true passive income (2026-01-13). Videos I uploaded three months earlier were still netting $7k/month at 100% profit after I abandoned the project - no ad spend, no fulfilment fees, and revenue persists long after publishing stops, unlike feed-based short-form platforms.
• The distribution-learning sequence (2026-07-24). Forget building entirely; grab an affiliate link for a proven product and practice: 1) make YouTube videos about the product, learning buying intent and structuring videos like VSLs where the product is the solution to the video’s problem; 2) build an email list, fed by educational short-form content, around the products you promote; 3) skip X until you are experienced - a far more sophisticated market. Done right, brand deals arrive before you need your own product: a channel I cofounded with only 3k subscribers is offered $4k per video just to talk about products.

Historical / patched plays

• YouTube Shorts mass-upload bot - DEAD (method ran ~age 17; recounted 2025-07-05 and 2026-06-01). I uploaded the same Short 5,000 times per day per account through a high-limit API (API uploads bypassed the UI limit, raising it from 10 to 100 per account), driving ~30k link clicks daily into a CPI content-locker offer for ~$1.5k-$2k/day. A YouTube algorithm change killed views for duplicate and API-uploaded videos, ending the era overnight. Transferable lessons: upload/distribution APIs often have different limits than the UI; raw traffic volume can be manufactured; pairing huge traffic with a weak offer wastes it (the trash CPI offer was my big mistake); and platform-dependent arbitrage dies without warning - extract profits fast and diversify.
• Snapchat content-locker CPA hustle - ethically questionable and likely patched (ran at 16, ~$200/day; recounted 2025-08-11). Mechanics: 1) account on a content-locking CPA network (OGAds, CPABuild); 2) locker headlined “download 2 apps to enter the $600 giveaway,” unlocking after 2 installs at $0.50-$1 each; 3) a persona Snapchat account seeded with 10 Americans from add-me sites; 4) DMs requiring entrants to complete the locker, send screenshots, and repost the giveaway to their story (viral loop); 5) quick-add to reach mutuals of existing contacts. The giveaway was never paid out and it demanded ~10 hours/day. Value today: understanding how incentivized CPA, manufactured social proof, and story-repost viral loops combine - and why platforms patched them.

Field notes

• The 90/10 cult-building formula (2025-02-12). Self-improvement audiences are built on 90% generic-but-genuinely-beneficial advice (train, eat clean, study, network) that buys uncritical acceptance of 10% indoctrination: make followers feel superior, call out everything wrong with the world, promise they’ll surpass everyone. Because the 90% visibly improves their life, the audience never questions the 10%.
• Odd-behavior asks are compliance tests (2025-02-21). Getting followers to perform costly, strange behaviors (extreme diets, cutting off friends, rituals) is deliberate - same mechanism as gang initiations. If someone follows strange advice with no clear benefit, the leader has proof of how far their buttons can be pushed, and monetization becomes easy.
• The self-help influencer manipulation sequence (2025-02-28): identify the audience’s deepest insecurities; show a “past self” with the same problems and a dramatic transformation; make viewers feel broken; position yourself as the one-size-fits-all solution; use storytelling for relatability; frame common problems as catastrophic to manufacture urgency; repeatedly display your results; brush off criticism and build an echo chamber; create us-vs-them so viewers pick a side. Once a viewer is that deep, they will buy anything.
• On YouTube, production value inversely correlates with information density (2025-02-23). The best technical info comes from tiny channels presenting slideshows; videos above ~50k views are optimized for entertainment. When researching, deliberately seek low-subscriber, low-production expert content.
• Cold DMs: casual and peer-like beats polished scripts (2025-09-02). “yo” - “got an opportunity for you, think we can cook” gets entertained; “let’s say we came into your business and added 100 signups a month” gets ignored. The 90% of failed cold DMs all translate to “I can see you’re making money, let me have some of it.” Approach as a collaborator, not a vendor.
• Extract a busy expert’s real opinion by being confidently wrong (2025-10-11). Reply to their posts from an alt account with confidently wrong claims - people write paragraphs to correct someone far more readily than to help someone.
• Print your project checklists (2025-09-04). Break every project into micro-steps in a Google Doc, print a physical copy, and tick boxes; the habit is one of my highest-leverage productivity tools. “Get addicted to ticking off checklists.”
• Hold money in multiple forms (2025-09-10). Cash, crypto, gold - a single geopolitical event can freeze your banking overnight; diversification across asset forms is insurance, not paranoia.
• De-algorithm your X feed (2025-02-23). Remove the “For You” feed and use the “Control Panel for Twitter” browser extension to avoid engineered outrage content - it made the platform “100x better” for me.
• Verified org accounts on X can buy inactive handles (2025-10-16) - a little-known way to get a clean brand handle.
• X fingerprints previously community-noted images (2025-11-06). Reposting a known-noted image gets automatically re-flagged via Grok detection - don’t reuse viral images with a community-note history, even in new contexts.
• Making TikTok treat you as a US poster from abroad (2025-11-16). Don’t connect the phone directly to a proxy - TikTok detects it and blocked even signup in my tests. Run the proxy on a PC, broadcast that connection as a hotspot through a WiFi dongle, and connect the phone to the hotspot; the relay passed detection.
• Design process for non-designers (2026-07-31). Browse Mobbin for UI patterns that fit your product, recreate them adapted to your specific utility, then apply your own taste on top.

Core Principles

• AI multiplies whatever you already are. If you’re good at something, it gets you good results 10x faster; if you’re bad, it produces trash 10x faster. The operator is the most important variable: give the same model to an average Fiverr copywriter and to David Ogilvy, and Ogilvy’s output wins. Domain expertise remains priority number one, because you cannot instruct an LLM to embody a skill you do not possess - prompting is transferring your skill into words so the model can embody it. Skills are not replaced by AI; they are amplified through it.
• You don’t make money by “learning AI.” You make money by learning how to make money - AI is just a tool that performs those actions faster or at scale. There are cracked technical people documenting a $3k MRR SaaS while a non-technical marketer does a $100k day ripping AI creatives for affiliate offers. You don’t need to know 99% of how AI works to profit from it (like driving a car without understanding the engine). Identify the actions that make you money, then find where AI performs those actions faster or better - and ignore tool-hopping hype cycles, much of which is paid influence or engagement bait.
• Stop obsessing over new model releases. A better model makes zero measurable difference for 99% of people. If current tools aren’t enough for you to build a cash-flowing business, the bottleneck is you, not the model. (2026-06)
• Being “good at AI” means exactly two things: feeding it your domain knowledge, and prompting it to emulate your skills. AI alone, without a skill to amplify, produces little.
• Career positioning: whatever your current skill is, become the best AI operator for that skill - a copywriter should become the best copywriting-focused AI operator. Someone who feeds a model “the brain of a copywriter” will replace plain copywriters as demand for manual digital labor falls.
• AI-assisted iteration beats fully offloaded output. Asking AI to generate a whole article and posting it as-is produces generic content; asking it for an outline, brainstorming headlines back-and-forth, drafting in your brand voice, then manually polishing produces results. And if someone can tell your output was AI-written, you are using it wrong - effective prompting requires active thinking about frameworks, rules, and context; the payoff is reallocating hours from manual drafting to actual needle-movers.
• You must know how to do a thing manually before you automate it. Agents are not all-knowing; handing one a task and expecting it to figure everything out fails. You remain the mastermind who knows what needs to be done and where agents can and cannot help.
• Know AI’s four structural bottlenecks and design around them: (1) every new chat starts from zero - solve with stored, reusable context profiles; (2) AI has no intuition - it only processes predefined data; (3) it has no true creativity - it remixes training data or provided context, so novel ideas must come from you; (4) it hallucinates confidently and is good at making false things sound true - verify claims.
• A base LLM without good information is confidently stupid. It answers with the authority of the top-ranking sources for a query, so researching unfamiliar topics with it means getting lied to a lot. Consequence: LLMs make genuinely good information MORE valuable, not less - now you can feed it in and converse with it. On uncertain or contested topics, feed the model multiple sources yourself and work from there instead of trusting base training data.
• Route around known weaknesses: LLMs are strong at verbal, logical, and memory reasoning but weak at spatial reasoning (a vanilla LLM fails badly at chess puzzles). Factor this into task delegation. posted March 2025 · practice might be outdated
• Privacy rule of thumb: don’t tell an LLM anything you wouldn’t type into Google - the data handling is identical.
• The highest-leverage use of an LLM is as an ideation partner, amplifying your thinking through back-and-forth discussion, not as an answer machine - the answers you need are usually not in the training data. Feed it your business data and let personalized analysis surface ideas you wouldn’t have considered; it shouldn’t make decisions.

Context Engineering: Profiles, Files, and Skills

• Prompt engineering is primarily a game of storing and reusing context. You, your business, and everything personal to you are not in the model’s training data; an LLM cannot tell you which solution is best for your business until it knows your business. Build reusable context profiles (business, audience, marketing channels, goals) pasted into any prompt in one action. Output quality with vs. without them is night and day - like overclocking your LLM. posted June 2025 · practice might be outdated
• Context is just two buckets: facts about your situation the model doesn’t know, plus instructions on how to use those facts. Categorize it and retrieve selectively - never dump wholesale. The fuller the context window, the worse the outputs. posted December 2025 · practice might be outdated
• Why JSON for stored context: LLMs handle structured, machine-readable data well; it is modular, easily editable, mergeable, and condenses context into fewer tokens, letting the model reference the right fields without ingesting whole paragraphs. But JSON is for storing context, not for prompting - prompt in plaintext or XML and inject JSON profiles as needed. “JSON prompting” was never magic; its real benefit was navigable templates with swappable variables, not model performance. (2025-08, reaffirmed 2026-04)
• The 8 core profile types I published templates for: business context, brand voice, marketing strategy, ideal customer, content strategy, product roadmap, audience psychographics, and YouTube scriptwriting. A business at full efficiency also holds profiles for each product, each distribution strategy, sales process, and team structure. Most of the time, simple context injection into prompts beats building a RAG system. posted June 2025 · practice might be outdated
• Build the business profile once - 1-2 hours - and reuse forever. Include: general overview (name, mission, problem solved, elevator pitch), product details and pricing, ideal customer personas (problems, desired outcomes, objections), and brand voice (tone, writing guidelines, example copy snippets). Fill it via an LLM interview rather than typing it out - interviews surface more accurate data than self-description. posted June 2025 · practice might be outdated
• Profile stacking: give the LLM multiple JSON profiles simultaneously (e.g., offer + ideal customer + distribution strategy) so it cross-references them for hyper-specific advice. Compounding quality: business + ICP + brand voice. posted May 2025 · practice might be outdated
• The three-input content formula: information context + brand voice profile + instruction. Example: a copywriting course transcript (info) + a creator’s brand-voice profile + “generate 30 short-form reel scripts in this voice using this info.” Feed a YouTube transcript to an LLM to clone any creator’s voice into a reusable profile. posted June 2025 · practice might be outdated
• Claude skills are the current (2026) best home for standing context - effectively SOPs for AI, significantly better than GPT projects. A skill is just markdown files in a .zip: Claude injects a short description of each skill into every chat and loads only the relevant reference files when a task calls for them, so your context window isn’t polluted (mentioning your VA’s name pulls their full profile only in that moment). Turn every process, SOP, and standing context - including profiles of people you work with - into a skill. Make your business context (market, offer, positioning) a standalone skill, separate from task skills like copywriting; pairing a task skill with your business-context skill is what produces genuinely useful output. posted November 2025 · practice might be outdated
• ChatGPT’s built-in memory is inferior to self-managed context profiles on six dimensions: structure (vague impressions vs. structured data), visibility, control (loads every session regardless of relevance), portability (vendor lock-in), modularity (mixes all projects), and collaboration (unshareable). My failure case: detailed diet/macros context saved as only “is interested in eating 4 meals per day.” Memory also silently pollutes future outputs and cross-references unrelated conversations. Use memory for light personalization and drafting an initial profile; use files you control for serious context injection. (My earlier 2025-04 take advocated flooding ChatGPT memory with context - profiles-over-memory is my later, settled position.) posted June 2025 · practice might be outdated
• When context grows large, use a two-call pattern: call 1’s only job is extracting task-relevant slices from your context store and assembling a prompt; call 2 does the actual work with a fresh window. Extendable to retrieval → LLM filtering/summarization → final call; also applies to filtering RAG context. posted June 2025 · practice might be outdated
• For long work sessions, keep the context window small and continuously save relevant context externally as you go - model output degrades as context grows (even Gemini degrades past ~100k tokens). Treat context as a curated, persistent asset you re-inject. posted December 2025 · practice might be outdated
• Convert courses and knowledge sources into context profiles: split by module, extract transcripts, and have an LLM compress each key point into a field containing one instruction or fact. Feeding a raw course in wholesale drowns the model in filler; the compiled profile is an editable, persistent “course” aligning the model with exactly the knowledge you want. posted August 2025 · practice might be outdated
• The secret behind viral mega-prompts is not the rules - it’s the context-collection window. Begin any serious prompt by having the model gather context question by question (an interview) before producing anything; this reliably yields roughly 10x better output. posted May 2025 · practice might be outdated
• Personal context database: maintain a file covering basic stats, current situation (focus, routine, time, resources, constraints), goals at 3-month/1-year/3-5-year horizons, problems and past obstacles, thinking style, interests, strengths and gaps. Have the AI interview you to fill it, and end sessions by asking it to update the database - plus the companion prompt: “based on this interaction, tell me a few things I may not know about myself that are either beneficial or detrimental to my growth.” posted February 2025 · practice might be outdated
• Master health context profile: a five-phase interview (bloodwork across dates for trends; symptoms/injuries/chronic conditions; weight, activity, sleep, hydration; diet patterns; imaging - MRI, X-ray, ECG, DEXA) compiled into one updateable document with values vs. reference ranges and flagged abnormalities, pasted into any future health question. The pattern generalizes: durable context profiles for any recurring advisory domain. posted December 2025 · practice might be outdated

Prompting Techniques

• Good-prompt checklist: role assignment, context (or a context-collection process like an interview), step-by-step instructions, an example output structure, and rules. Prompts don’t need to be long, but the more conditions you set, the more curated the output - no matter how smart models get. posted May 2025 · practice might be outdated
• Clear instructions beat raw model intelligence. Broad prompts will never match the output you imagined: the more room you leave a model to infer intent, the further the output drifts. A person of average ability with precise step-by-step instructions outperforms a genius with messy ones.
• Role assignment steers which training data gets accessed. Asking plainly for a CPA marketing method returned generic results; “you are a BlackHatWorld moderator” returned specific ones. Even “take a high IQ, rational, first-principles approach” noticeably sharpens ChatGPT’s advice. Caveat: role bias doesn’t qualify data quality - a model hooked to a specialized knowledge base beats its general-use state. posted August 2025 · practice might be outdated
• Multi-role debate: the overlooked upgrade to role assignment is assigning two or more expert roles to the same problem, making them debate, and outputting the refined synthesis. Extension - the “council of experts” truth filter: simulate a debate between polar opposites (e.g., carnivore vs. vegan) and ask for their common ground; conclusions that opposing camps share hold the most truth. posted May 2025 · practice might be outdated
• Recursive prompting: ask LLM #1 “what is the best way to prompt {X idea} so it gives {Y output}”; send the generated prompt to LLM #2, inspect output, ask it to edit the prompt; store the polished version in a prompt library. If AI is an instruction-following machine, have it write and improve its own instructions. posted March 2025 · practice might be outdated
• Serialize a great session: when a chat is performing exceptionally well, send “your current state is perfect - send me a prompt in markdown that I can send to another LLM so it acts as a clone of you,” and reuse the output as a system prompt anywhere. posted February 2025 · practice might be outdated
• The “prompt engineer” rewriter: keep a system prompt that converts messy questions into precise instructions - assign an expert role, state the topic simply, break the request into parts, demand examples/steps, specify output format, name the audience, define success. “How do I market my business?” becomes “You’re a marketing expert. Give me 3 marketing strategies under $1,000 that worked for real businesses and can start this week, with exact steps and common mistakes.” posted March 2025 · practice might be outdated
• Prompt-engineering fundamentals worth studying: meta-prompting, chain-of-thought, few-shot, self-refining prompts, prompt-chaining, role-based prompting, and socratic prompting - learn these plus how an LLM processes a query and you can construct prompts intuitively. posted April 2025 · practice might be outdated
• Refusals are often soft. When an AI refuses something it can actually do, blunt pushback (“stop bullshitting”) occasionally gets compliance. posted April 2025 · practice might be outdated
• Translator pattern for weaker tools: when an AI builder misunderstands you, give a stronger model (Claude) a screenshot plus your idea and have it write unambiguous instructions for the builder - a prompt writing a better prompt. posted March 2025 · practice might be outdated

Advisor and Self-Analysis Prompts

Default LLMs are dangerous yes-men - you must explicitly prompt for pushback. My library of counter-sycophancy prompts, each with its own distinct structure:

• The “brutally honest strategic advisor” (my most viral, ~1.5M views): IQ 180, brutally honest, built billion-dollar companies, expert in psychology/strategy/execution, cares about your success but tolerates no excuses, thinks in systems and root causes. Mission: identify critical gaps, design specific action plans, push past comfort zones, call out blind spots and rationalizations, force bigger thinking, hold you accountable. Response format: hard truth first, then specific actionable steps, ending with a direct challenge or assignment. Bonus: if its thinking exceeds yours, ask it to teach you how to think like it. posted February 2025 · practice might be outdated
• The “hyper-rational first-principles problem solver”: break everything to foundational truths, challenge all assumptions, design interventions at leverage points by impact-to-effort ratio, cut off excuses. Fixed format: situation analysis (core problem, assumptions, first-principles breakdown) → solution architecture (intervention points, action steps, success metrics, risk mitigation) → execution framework (immediate next actions, progress tracking, course-correction triggers, accountability). Constraints: no motivational fluff, no vague advice, no theory without application. posted March 2025 · practice might be outdated
• The “brutally honest strategic analyst”: an expert in behavioral psychology and cognitive biases with zero tolerance for self-deception. Extracts goals with exact metrics and timelines; asks what you actually did in the last 24-48 hours toward each; for every excuse, judges legitimate obstacle vs. rationalization and names the cognitive bias; forces confrontation of goals vs. daily actions, claimed priorities vs. time allocation, perceived vs. actual effort; never accepts vague answers. posted June 2025 · practice might be outdated
• The “accountability manipulator”: questions your memories of “trying hard enough,” compares you to an alternate-timeline self who took action, points out inconsistencies in excuses, reframes past failures as proof of capability, refuses sympathy - starts by asking your goals, then systematically dismantles every excuse. posted April 2025 · practice might be outdated
• The “Life Optimization Advisor”: interviews you one question at a time on ultimate goals, hour-by-hour routine, income and spending, relationships, health, and time allocation, challenging every inconsistency; then lists every inefficiency, calculates opportunity cost of wasteful activities, highlights goal-action contradictions, and outputs a measurable plan with schedule optimization, habit protocols, weekly accountability metrics, and consequences. No sugar-coating, no platitudes, no vague answers accepted. posted April 2025 · practice might be outdated
• The “life analyzer”: one question at a time across six phases (physical, mental/emotional, financial, professional, lifestyle, goals), then alignment scores (0-100%) per area, gap analysis, a 30-day/90-day/1-year action plan, systems recommendations, and resource allocation across time, money, energy, and skill development. posted April 2025 · practice might be outdated
• The “rational insights” anti-sycophancy system prompt: evaluates logical consistency, evidence quality, hidden assumptions, biases, emotional vs. rational reasoning, and causal claims; points out flaws with the specific logical error and a better reasoning path; acknowledges strong reasoning without flattery; calls out fallacies immediately, questions belief sources, encourages steel-manning; prohibits unnecessary politeness, appeals to authority, and vague feedback. posted June 2025 · practice might be outdated
• The “pure logic engine” (LogicCore): restates your problem stripped of emotional language, asks up to 10 clarifying questions (one at a time) targeting measurable variables and cause-effect, then delivers a core problem statement, causal chain, an IF/THEN solution framework prioritized by implementation speed, resource efficiency, success probability and measurable impact, and a numbered action protocol with success metrics and failure points. posted March 2025 · practice might be outdated
• Memory-powered flaw diagnosis: a ChatGPT prompt using its stored memory of you in three parts - Diagnosis (one core flaw only, citing specific patterns from memory), Consequences (how it has limited outcomes, referencing past behavior), Prescription (the highest-leverage shift aligned with known goals). Rules: no politeness, brutal clarity over comfort. posted April 2025 · practice might be outdated
• Goal-to-checklist (“elite strategic advisor”): one question at a time covering end goal, timeline, resources (skills/money/connections/tools), obstacles, success metrics; after 5-7 questions, summarize the goal in one sentence and confirm; then output a nested-checklist roadmap of milestones broken into tasks with dependencies, time and resource estimates, roadblocks with contingencies, and progress metrics. Companion aphorism: if you knew your next 100 actions, you’d do them in a quarter of the time. posted June 2025 · practice might be outdated
• Goal-to-system prompt: the AI interviews you about what you want, why, what blocks you, and what structure suits you, then designs a system that is specific, includes daily/weekly actions, minimizes decision fatigue, includes tracking, and adapts over time. posted April 2025 · practice might be outdated
• Journal profiling: feed daily journal entries to a prompt that builds and continuously updates an identity profile (core identity, cognitive patterns, behavioral patterns, emotional landscape, relationships, goals, challenges, strengths), extracting explicit and implicit information, tracking patterns and contradictions, and updating confidence levels over time. posted March 2025 · practice might be outdated
• Expert-corpus self-assessment: build a context-profile template from a body of expert writing (I used Corporate Machiavelli’s 55 essays), have the AI fill it via interview, chat history, or journals, then rate you 1-10 on every measurable value, identify your best-fit fields, and lay out a course of action. Honest answers make it scarily accurate. posted March 2025 · practice might be outdated

Business Strategy Prompts

• Idea backlog analysis: dump every idea from your notes app into a prompt that scores each on market potential (1-10), execution complexity (1-10), resource requirements, time to market, revenue streams, risks, and competitive advantage; runs pattern recognition for themes, synergies, and combinations; suggests simplifications and pivots; ranks by profit potential, speed, resource efficiency, and moat; and produces an execution roadmap for the top three. End with: be brutally honest about flaws. posted March 2025 · practice might be outdated
• Idea-to-execution blueprint: a phased interview (one question at a time, up to 50, flagging critical flaws immediately) through core idea extraction, market/competitor analysis, marketing strategy (content pillars, organic, SEO, paid with budget allocation), execution framework (resources, risks, milestones, KPIs, cash flow, tech stack), and optimization/scaling - outputting an executive summary, 30-60-90 day plan, resource requirements and burn rate, KPIs with break-even analysis, risk assessment, and scaling triggers. posted March 2025 · practice might be outdated
• Business-model matcher: an interview prompt (one question at a time, max 20, each building on prior answers) across skills, experience, personality and work preferences (risk tolerance, time), and practical constraints (capital, income goals) - outputting 3-5 aligned business models with timelines to profitability, starting requirements, validation steps, and scaling potential. Zero-capital variant: three parts (up to 10 skill questions, up to 5 resource questions), ending with your 3 most valuable skill combinations and the top 2 zero-cost opportunities launchable within 24 hours, each with 5 immediate action steps. posted May 2025 · practice might be outdated
• Three-phase market-positioning strategist: (1) skill assessment - probe existing specialized skills or guide selection via what you research for fun and where you beat peers, ending with a 90-day learning roadmap; (2) market validation - demand, competition, pricing, service vs. product vs. hybrid; (3) distribution - branch on camera comfort: on-camera path (YouTube/TikTok/Instagram) vs. off-camera path (Twitter threads, newsletter, LinkedIn). Ends with a 30-day action plan and metrics. posted May 2025 · practice might be outdated
• “Objective Self-Analysis” for business direction: interview across five categories - natural proclivities (what energizes you, flow states), skills (what people pay for and ask your help with), experience (repeated patterns, proven wins), network (who can help, communities), unfair advantages (resources, background, head starts) - producing per-category analysis and a 3-5 sentence summary of your unique edge. posted July 2025 · practice might be outdated
• Traffic-strategy interview (Traffic Secrets-based): the AI interviews you (product and UVP, ideal customer, current channels, top three traffic challenges, 6-12 month goals), then applies Brunson’s frameworks - Dream 100, content distribution across owned and external platforms, hook-story-offer funnels - output as business summary, traffic diagnosis, strategic framework, and a 30-60-90 day plan. posted March 2025 · practice might be outdated
• Marketing-strategy interview → JSON: four phases (business foundation; positioning and messaging; channels and content; strategy design - customer journey, lead capture, offers, pricing, campaigns), one question at a time, exported as a JSON profile you feed into future prompts or hand to a team. Similar: an ICP interview (ten questions across demographics, values, lifestyle, pains, purchase triggers, price sensitivity, platforms, brand expectations) ending in a reusable ideal-customer JSON with recommended channels, content strategy, messaging, and USPs. posted April 2025 · practice might be outdated
• Reddit pain-point research prompt (best in deep-research mode): inject product and ICP; generate psychographic search queries in three formats - emotional triggers (“frustrated with”, “hate when”), aspirational language (“wish I could”), pain indicators (“anyone else struggle with”) - then analyze emotional themes, recurring frustrations, language patterns, intensity via comment engagement, and competing solutions; organize into primary/secondary/emerging pain points with direct quotes, 1-10 intensity scores, frequency, solutions tried, and gaps. posted May 2025 · practice might be outdated
• Distribution-channel deep research: role of senior market research analyst; inject business context and ICP; forbid speculation; step through where ideal customers spend time online, their frustrations and unmet needs, and highest-ROI organic and paid channels based on real behavior and buyer readiness; output JSON of distribution_channels (name, organic/paid, reason, strategy), audience_touchpoints, audience_painpoints. posted May 2025 · practice might be outdated
• Interview → project profile → checklist: have the AI interview you about idea and strategy to build a “project profile,” feed that into a profile-to-execution-checklist prompt, optionally hand the result to a coding agent as a tracked project checklist. Once the plan exists, the only variable left is execution. posted March 2025 · practice might be outdated
• Mine your communication history: export chat history with a business partner (Telegram/Slack), have AI write a conversion/cleaning script (Cursor, optimizing for token economy, chunking if needed), then feed it to an LLM to extract every business idea ever discussed - or fill a partner-analysis profile template (communication style, problem-solving approach, decision speed, skills, confidence, delegation, leadership, adaptability) for both parties. Extension: message your partner every idea you have and treat the archive as an AI-queryable second brain - Napoleon Hill’s “mastermind third mind” made literal. posted June 2025 · practice might be outdated

Learning with AI

• Stop consuming long-form content raw. Most nonfiction is fluff around a few key points; if something can be digested more efficiently with AI, consuming it in full is leisure, not learning. When a book is recommended, first ask an LLM for its chapters with bullets each, then choose what to learn deeply. posted March 2025 · practice might be outdated
• The book method (full version): don’t ask for a whole-book summary - it loses crucial information. (1) Create a personal/business/mindset context profile. (2) Paste one chapter plus the profile into the LLM - one chapter per conversation so nothing gets skipped. (3) For each topic, ask it to explain the theory, convert it into practical applications for your specific situation, and run an interactive questionnaire until you truly understand. Shortcut variant: flick through a book, pick the topic that will benefit you, send just that chapter with your context and ask which concepts apply directly to your life. posted August 2025 · practice might be outdated
• The video method: pull the transcript (youtubetotranscript or similar) and have an LLM extract key points - or watch with an AI chat open beside the video for instant clarification and personalized applications. Send the transcript, not the video link: a 20-minute video analyzed as video is ~300k tokens; split prompts so each covers one sub-topic. posted April 2025 · practice might be outdated
• The “implementation extractor”: turn any video transcript into a max-10-step action plan - only concrete, immediately executable actions; each step numbered, starting with an action verb; specific numbers/timeframes preserved; all “why” explanations removed; vague steps made specific or deleted. Output: EXECUTION STEPS, KEY METRICS, REQUIRED TOOLS/RESOURCES. posted March 2025 · practice might be outdated
• The recursive tutor prompt: AI asks what you want to learn, builds a progressive syllabus, and per lesson explains with analogies, asks socratic questions, gives one short exercise, and only proceeds when you’re ready - rephrasing if not. Mini-quiz after each section; final integrative challenge plus real-world reflection. Replaces passively watching hours of course video. posted April 2025 · practice might be outdated
• The Pareto learning prompt: identify the critical 20% of concepts producing 80% of results in any skill; output core concepts with reasoning, what was cut and why, a learning sequence formatted [Concept] - [Resource] - [why this resource] with exactly one vetted, specific resource per concept (never “any YouTube video about X”), practical challenges, and mastery metrics (“you truly understand this when…”). posted April 2025 · practice might be outdated
• Deep-learn any resource via an AI IDE: put it in Cursor/Windsurf, have the agent extract every practical detail into a hierarchical markdown, then have a chat LLM walk you through it point by point - guaranteeing full coverage instead of skimming. For PDFs too big for a chat context window, the IDE chunks them and can explain each chunk “like you’re a smart 12-year-old” into an ordered markdown; merely-large documents fit Gemini’s long context directly. posted March 2025 · practice might be outdated
• Learn from a proven master, not the model’s average: pick a person with demonstrated mastery of the skill and make the LLM pull exclusively from their educational content - converting it from mediocre generalist into a specialist with a trusted lens.
• Personalized curriculum project: a ChatGPT project with two JSON files - your personal context (what you do, workflow, how you learn) and a curriculum with learned/unlearned booleans. Each session: new chat, run an interactive lesson on the first unlearned item, grounded in your context; update the boolean manually after. Also ask AI for 100 practice projects that would benefit your real work. posted July 2025 · practice might be outdated
• Learning-style assessment interview: a “behavioral learning strategist” prompt in five phases - situational questions, a format resonance test (one topic explained three ways), a learning-by-doing reflection, pattern-recognition questions, then a structured JSON learning_style profile (dominant/secondary style, input preferences, friction points, optimal self-learning strategy) - assessing observed behavior rather than self-reporting; save it so future explanations are tailored. posted April 2025 · practice might be outdated
• Learn to code (enough): ask Claude for a learning roadmap, feed it to a coding agent, and have it walk you through each part with practical exercises. Basics take about a week and make you a far better prompter - you can specify technologies and reason about logic. Beginners should not let the agent auto-generate everything: have a chat LLM explain every line you add and why, or you lose grip on the codebase. posted July 2025 · practice might be outdated
• RAG-chatbot tutor prompt: teaches an absolute beginner the full build - defines “embedding” and “vector store” in plain English, asks what chatbot you want as the running example, then numbered steps (OpenAI embeddings → Pinecone storage → retrieval → GPT generation → commented Jupyter cells) with installation instructions, common-error fixes, current SDK, stopping after every step until you say “continue.” posted July 2025 · practice might be outdated
• Learn by watching agents: give a task to an autonomous agent (e.g., Manus) and watch how it decomposes and solves it - you absorb tool-use patterns by observation. posted April 2025 · practice might be outdated
• The no-BS AI curriculum, in order: what an LLM is (training, fine-tuning, inference); capabilities vs. limitations (hallucination, bias); model types and best use-cases; context windows and token economy; tools by function and how to find niche ones; tool-chaining; meta-prompting; abstraction prompts for ideation and reasoning prompts for planning; context management and memory architecture; then domain-specific application. Learn fundamentals before building workflows - flows built without them function but are inefficient. Pareto framing: decide if you’re an engineer or an operator; operators (99% of people) skip the mathematical internals entirely. posted June 2025 · practice might be outdated
• Four learning paths (sample all, then specialize): general foundations (LLM mechanics, token economy, context management, prompt engineering, RAG/embeddings/vector DBs, MCP); creative (multimodal capabilities per model, style profiles, image/video/sound tools, tool-chaining); automation (n8n/Make/Zapier, JSON, triggers/actions/APIs, error handling); vibe-coding (coding LLMs by power vs. cost, AI IDEs, breaking projects into step-goals, debugging, front-end tools like v0/Tailwind/shadcn). posted May 2025 · practice might be outdated

Image Generation and Visual Content

• JSON style profiles are the master technique: feed reference images (ads, thumbnails, brand assets) to a model and have it extract the stylistic qualities - color palette, composition, character style, typography, textures, lighting, motifs, post-processing - into a structured JSON documenting only the design system, explicitly excluding specific subjects, logos, people, or brand names. Reuse the profile to generate new on-brand visuals for entirely different content; edit or merge profiles as needed. Variants: thumbnail profiles, brand-kit profiles for new products, and reusable “filter profiles” (extract the look, apply to a new image, iteratively ask it to update named qualities, save as a customizable filter). posted March 2025 · practice might be outdated
• Separate style from composition for maximum control: combine a JSON style profile with a rough visual layout - dump relevant PNGs into Canva/Photoshop/MS Paint arranged roughly as you want, export, and send the layout plus the JSON with “turn this into a finished image based on the profile.” Far more control than any text prompt alone. posted March 2025 · practice might be outdated
• Sketch-to-finished-graphic: a hand-drawn sketch plus one structured prompt (set resolution, correct specific elements - “make the dollar bill a real $100 bill,” remove labels, “let your creativity run wild but follow the instructions on the thumbnail”) produces finished thumbnails; you can write instructions directly on the sketch. Current (2026) tool: Nano Banana Pro (run via Freepik with quality up) - extracts individual elements from existing thumbnails, turns paper sketches into finished logos/graphics, near-replacing thumbnail designers. posted December 2025 · practice might be outdated
• Prompt for imperfection to get realism: name a low-end camera, casual context, and explicitly ask for a noisy, authentic, non-cinematic look - e.g., “taken from an iPhone 6… noisy and look authentic not cinematic, this photo was lazily taken in the November cold.” Polished defaults are what give AI away. posted March 2025 · practice might be outdated
• 4o image-gen weaknesses and workarounds (2025-03): faces drift toward uncanny near-likenesses - train Flux on 5+ photos of your own face (krea.ai/train) for near-1:1 self-images; color grading biases orange/red - specify colors explicitly; text breaks - spell out exact text in the prompt; glitches - repair with Photoshop generative AI. Locked aspect ratios: add black bars to your sketch (62px bars yield exact 16:9), generate between them, crop after. Refused prompts: run the identical prompt through Sora - same image model, more lenient filtering. posted April 2025 · practice might be outdated
• Static-ad interview prompt: a legendary direct-response marketer persona gathers product, ideal customer, #1 problem solved, guarantee/USP, headline benefit and style (problem/benefit/question/direct), proof points, exact CTA, mood, visual style, and brand colors - then outputs an art-director brief (scene composition, lighting, camera angle, headline placement, text hierarchy, text-to-image balance) ready for an image model. posted March 2025 · practice might be outdated
• Adaptive image-interview prompt: before generating, the AI extracts the image in your head one question at a time, adapting by subject type (person → pose/expression/clothing; scene → perspective/time of day/weather) plus style, technical, mood, and use questions - wrapping within ~10 questions into a structured generation prompt. posted April 2025 · practice might be outdated
• Process infographics from one prompt: image models can generate complete step-by-step recipe/process infographics - specify view angle, layout, labels with exact quantities, connecting dotted lines with icons, and a final shot. Specificity is what makes it work. Chain tools for data-driven graphics: Perplexity extracts structured facts, ChatGPT renders them with formatting instructions (“1:1 grid, each time-slot its own box, no spelling mistakes”). posted March 2025 · practice might be outdated
• UI cloning with Claude Code (2026): Claude cannot extract accurate styling from screenshots alone but replicates given CSS very well. (1) Open dev tools on a UI you admire; copy the full CSS plus a screenshot. (2) “Rebuild the exact same UI design as the screenshot in a single html file, css attached.” (3) Use VisBug to copy per-element CSS until pixel-perfect. (4) Have Claude generate a detailed style-guide markdown (palette, typography, spacing system, component styles, shadows, animations, radii, Tailwind usage, example components). (5) Drop that style guide into any project and Claude one-shots new pages in that style, even across context resets.
• One-shot branded decks (2026): paste content into Claude, invoke the pptx skill, attach a brand-style context profile - finished on-brand PowerPoint in a single prompt. posted January 2026 · practice might be outdated

AI Video and UGC Production

• The 2025 pipeline (dated): Flux 1.1 Pro Ultra for images to animate, Kling for animation, chaining ~10-second clips since nothing longer was economical. For storyline consistency, have an LLM write the sequential image prompts following the storyline, generate each image, then animate each. By 2026 the stack moved to Seedance 2.5 / Veo / Sora for video and Nano Banana Pro / GPT-Image for frames.
• Seedance 2.5 honest assessment (2026-08): shockingly realistic - near-solved realism for UGC - and excellent at following instructions, but expensive (~$5 per 17-second clip, ~$10 per 30 seconds; get the prompt right so you one-shot it) and still bad at rendering text - add captions and overlays yourself in an editor.
• AI UGC works commercially: AI-generated UGC passes as real to roughly 80% of viewers, collapsing creative costs. Platform risk is low: TikTok’s parent built Seedance and Meta ships AI creative tools - they will not ban AI content. And for openly-AI entertainment content, the viral secret is the opposite of hiding it: make it as absurd and obviously AI as possible (“no way this is real”) - audiences accept fantasy like they accept VFX.
• The clone-a-winner assembly line (2026-08, my flagship workflow): (1) give your agent a short-form content research API and pull ~100 videos promoting products like yours; (2) pick one that fits; (3) have the agent extract ALL context via a video-analysis tool (Higgsfield CLI has one built in) into a structured markdown - scenes, motions, appearance, mannerisms, accents, tone; (4) generate candidate starting frames via the CLI and pick the best; (5) rewrite the context so it promotes YOUR product; (6) send rewritten context plus starting frame to Seedance 2.5. I get ~80% first-try success. Ten-minute variant: research agent finds references → Gemini 2.5 Flash produces a scene-by-scene analysis with image-reference placeholders → GPT-image-2 recreates a frame per cut-scene → Higgsfield video generation with Seedance → human editor adds captions, cuts, zooms, music.
• Fully automated viral-video cloning with Claude Code: query a TikTok research API for a niche filtered to 100k+ views, outliers, and sales intent; one project folder per video; download with yt-dlp; PySceneDetect splits scenes at hard cuts; ffmpeg grabs each scene’s start frame ~0.08s past the cut (avoiding transition motion blur) plus mid/end frames; Gemini API analyzes hook, pacing, structure, and why it converts as structured JSON; Claude Code writes one generation prompt per scene; each start frame + prompt goes to Seedance image-to-video.
• Segment-assembly UGC: realistic UGC videos are chains of 4-8 second clips, each from a starting-frame image plus a short prompt. Get frames by screenshotting real TikTok UGC and having an image model change the person’s appearance; prompt template: “extend this video and make her say this (make sure she is very expressive and enthusiastic): {script}”. Expect a few generations per segment; compile in CapCut. Avatar-tool alternative (makeugc): input script, pick or create an avatar (including product-in-hand), iterate on script and tone - good enough to run as traffic creatives.
• The green-screen format: generate a still of a presenter in the bottom corner with a green screen behind them (prompt for organic, hand-held, imperfect quality), animate with Seedance 2.5 plus the script, then chroma-key any screenshot in as the background - endlessly reusable for selling almost anything.
• Anatomy of a professional AI video prompt (from Higgsfield’s open-sourced packs): style prefix (film stock, lens, color grading, director-style camera language); shot-mode declaration (multi-shot montage, hard cuts); image references with explicit identity/wardrobe locks; a global consistency-lock section (what must never change between shots); diegetic sound and lip-synced dialogue per shot; per-shot camera and framing direction; and a long negative-prompt list (no watermark, no wardrobe change, no lip-sync drift, no morphing). Copy the scaffold. Career path: study these open-sourced packs, practice in Cinema Studio, make trend-tied shorts, network with AI filmmakers - corporate deals in the space are large.
• Gemini is the video-analysis model, full stop. Google has by far the best vision model regardless of benchmarks (it deciphers 200-year-old Kurrentschrift handwriting other models fail on) - judge multimodal models by hard real-world tests, not leaderboards. Gemini sees a YouTube video’s actual visuals (and Google-private data), not just the transcript like Perplexity. Non-YouTube videos (reels, TikToks, VSLs): rip the file and upload to a private YouTube account, then feed the link. Screen recordings work too - it extracts structured data from any on-screen workflow. Limit: ~2 hours of video floods the context window; chunk into 1-2 hour segments (use AI Studio for this). At scale: give Claude a Gemini 2.5 Flash API key - I analyzed over 1 million videos this way for about $140 total.
• Gemini editing workflows: extract an “editing-style profile” JSON from a well-edited reference video, upload your raw footage to a private YouTube account, and ask for a breakdown of cuts, VFX, and transitions with exact timestamps - or use the profile as a manual that makes a human editor faster and cheaper. Retention reverse-engineering: extract shot-by-shot profiles (script, visuals, SFX, mood) from successful videos, ask Gemini for the key retention strategies, build a profile of your own channel, and request an implementation manual. Wrapper opportunity: Gemini video-analysis apps are far less crowded than image-gen wrappers because even daily AI users don’t know Gemini processes visuals. posted April 2025 · practice might be outdated
• Articles → video scripts: scrape articles with Firecrawl, extract a “script profile” from a proven video in your niche (pacing, storytelling, retention), then have an LLM write a script pulling information from the articles in the style of the profile. Automatable via the Firecrawl MCP server. posted April 2025 · practice might be outdated
• Scaling expert content without cloning the expert: put the expert’s wisdom into a knowledge base, build an AI system that pulls topics from it to generate scripts, then hire and train human presenters - multiple lead-generating channels hands-off (recommended in 2025 over full AI presenters, which weren’t convincing yet). posted May 2025 · practice might be outdated

Coding Agents, Automation, and Knowledge Systems

• The terminal is becoming the everything-agent. Coding, marketing, ops, note-taking, payments - anything executable through code should be automated with a coding agent. My 2026 80/20: get deeply familiar with the Anthropic ecosystem (Claude, Claude Code, and its agent tooling) rather than spreading across every new tool.
• Agents vs. workflows - not interchangeable: a workflow is a predefined sequence of steps (you do the reasoning at design time); an agent takes a task and reasons about execution itself (you outsource the reasoning). An agent can be one step inside a workflow. Important when scoping and pricing automation projects.
• Automation only pays when the underlying process is already lucrative. Skip general-purpose agent setups (e.g., Clawdbot hype); build small specialized agents around the few tasks that actually move the needle. Litmus test for always-on assistants: if hiring a human assistant wouldn’t yet be profitable for you, an AI assistant won’t be either - beginners with nothing to automate gain nothing from automation tools.
• AI automation needs only six baseline skills: workflow tools like n8n, JSON data structures, prompt engineering, conditional logic, API requests, and webhooks - after which you can charge companies thousands to build agents. Learn market-relevant projects by browsing Upwork’s “AI automation” listings and building what businesses actually pay for; every practice project becomes a portfolio piece. Free n8n resources: RoboNuggets’ videos and Nick Saraev’s free 6-hour YouTube course. Graduation path: prototype in n8n to learn workflow logic, then have a reasoning model break each workflow down node by node and teach you to rebuild it in Python - production agents need the control only code provides. posted July 2025 · practice might be outdated
• Plan, audit, then execute: before building with a coding agent, have one LLM write the full build instructions as markdown, have another LLM audit the file for consistency and efficiency (repeat the audit several times), then hand it to the agent. Separating planning, verification, and execution catches inconsistencies before any code is written. posted April 2025 · practice might be outdated
• Vibe-coding guardrails: for complex systems, have your agent maintain a visual flow diagram of every back-end process, with an agent skill that auto-updates it whenever code changes - a middle ground between reading all generated code and going blind, and it makes collaboration ~10x easier. And don’t vibe-code your landing page: tapped-in buyers instantly sniff it out; use a proper builder like Framer.
• MCP tool-chaining: research MCP feeds fresh data, generation MCP renders output, the agent orchestrates - e.g., Roo Code + Perplexity MCP + ElevenLabs MCP producing automated voiceovers with real-time researched data (2025-04 example; the pattern generalizes). posted April 2025 · practice might be outdated
• Conversation data → custom agent (2025 RAG pipeline): export Slack data as JSON (and dump ChatGPT’s knowledge of your business as JSON); use Cursor with a long-context model to write a LangChain script that cleans (strips filler), chunks, and converts to Q&A pairs; embed with OpenAI embeddings into Pinecone; wrap as a LangChain agent and deploy. Tip: paste the outline into an LLM and have it walk you through each step. posted April 2025 · practice might be outdated
• The continuously-fed knowledge base is the real edge - not asking ChatGPT questions. An agent that scrapes relevant sources (competitors, ad libraries, forums, YouTube, X, podcasts, newsletters), filters into a personal knowledge base, produces a daily report, and ties findings to your business context via profiles - queryable directly instead of relying on training data. posted May 2025 · practice might be outdated
• The Obsidian + Claude Code second brain (2026): open the terminal inside Obsidian and have Claude Code create an Obsidian-optimized structure (daily notes, projects, knowledge, resources), one main index for agent navigation, a CLAUDE.md with conventions, interlinked concepts, and a single BRAINDUMP.md where you dump raw info that Claude organizes on command. Always-on version: host an agent on a VPS, connect it to Telegram, and have it build and maintain the vault around your entire life - agentically searchable, with a skill that always appends new info about you, seeded by braindumping goals and uploading chat history, Notion databases, and tweet archives. Put the agent inside the chat app where your team already communicates so the compounding is automatic. Hosting option: a spare Mac mini keeps the knowledge base local, letting every agent (Claude Code, Codex, Hermes) access the same centralized information remotely under your control.
• Drop-folder document pipelines: my private family wiki in Claude Code - drop scanned documents, Gemini vision transcribes them (including 120 pages of old German Kurrentschrift and cursive Cyrillic), and the agent routes extracted facts into the correct person’s markdown page (one page per person, an index, a family-tree.md with relationship links). Once populated, it can one-shot a printable family-history book. The drop-folder → vision-transcribe → route-to-entity architecture applies to any document corpus.
• Bespoke internal tools beat off-the-shelf: instead of adapting to Notion, prompt an AI builder to generate a dashboard customized to your exact workflow - my 2025 example: Bolt.new + Supabase + a DeepSeek API key produced a working board in about three prompts. Micro-SaaS extension: build a tool that fixes a problem you personally have, paywall it with a free trial, and market it with screen-recorded short-form videos of the tool in action. Small utility example (2026): Tally for forms because its API lets an agent generate a complete form in ~30 seconds.

Model and Tool Selection

• “What is the best LLM?” is the wrong question - ask which model is best for a specific task, and chain specialized models rather than committing to one. Benchmark intelligence isn’t the only criterion: use the most readable, human-feeling model for human-facing text and stronger reasoning models elsewhere.
• My division of labor (2025-12 → 2026): Claude for writing copy, context management, documents, and repeatable workflows via skills (most intuitive for context profiles); Gemini for heavy reasoning, very large context, and all video/vision analysis; ChatGPT for quick summarization and basic questions; Grok for researching trends or anything extractable from X. (Earlier snapshots for reference: 2025-03 - Claude for copy, Perplexity for research, Poe at ~$20/month for credit-based access to all major LLMs without daily caps; 2025-04 - Gemini as main LLM, ChatGPT for images, n8n for automation, Kling for video, ElevenLabs for sound, Windsurf for coding.) posted December 2025 · practice might be outdated
• The 2026 category map (know the current best per category, not every release): LLMs - Claude, Gemini, GPT, Kimi. Coding agents - Claude Code, Cursor, opencode, Lovable. Computer-use agents - Manus, OpenAI/Claude. Image - Nano Banana Pro, GPT-Image, Midjourney. Video - Google Veo, Sora, Kling, Seedream. Audio - ElevenLabs, Suno. Automation - Claude Code, n8n, OpenClaw. Claude Code alone covers many categories when plugged into the right tools.
• Cheap tiered pipelines: use a free reasoning model to write instructions, a scaffolding tool for the bulk, and an AI IDE for finishing (2025 example: Grok/Kimi → Replit → Cursor) - each tool where it’s strongest, cost minimized. posted March 2025 · practice might be outdated
• Six low-barrier AI skills anyone can build: LLM proficiency (which model per task), prompt engineering, context management, no-code automation, AI-assisted coding, and AI creative work (style profiles, brand-voice copy, image/video/audio tools). posted April 2025 · practice might be outdated
• The highest-ROI AI use cases, as a checklist: delegating repetitive tasks to agents; coding assistance; reasoning steps inside automations; content creation (10x the process even if not fully automated); marketing (landers, ads, copy); learning assistance; deep research; brainstorming; chatbots on internal knowledge bases. posted April 2025 · practice might be outdated
• Deep research is genuinely powerful: Gemini Deep Research scraped ~200 websites from a 50-word prompt and identified a half-forgotten WW2-era family figure, down to living relatives, in 5 minutes. Pattern: give it every known specific and let it triangulate. Gemini is also remarkably good at deciphering old handwritten documents - least hallucination in my testing, unlocking hundreds of genealogical records. posted December 2025 · practice might be outdated

AI Content Without the Slop

• “AI slop” is a user-skill problem, not a model problem. Several highly respected X accounts run roughly 80% AI-generated content undetected. My formula: strong writing-style training data plus fresh, high-quality informational input per piece - but keep DMs and genuine one-to-one interactions human. posted August 2025 · practice might be outdated
• The long-form copywriting workflow: braindump key details into a doc, let AI fill gaps and handle formatting with a brand-voice profile in context, then skim every line and humanize. Claude-skill version: dump your bullet-point draft as context, generate with a skill built on good copywriting practices, constrain the model to use only the information you gave in the structure you request, then humanize. You will not one-shot good copy - and the system only works if you first teach the model what good copy is (e.g., extract style from landing pages you know convert). Copywriters claiming AI can’t replicate their style are coping: instructing AI well is itself a writing skill. posted November 2025 · practice might be outdated

Adjacent Tool Picks (early, dated)

• 2024 endorsements: Simple Analytics over Google Analytics for lean online businesses; ElevenLabs as the most useful SaaS I’d used; Framer for landing pages; the “Control Panel for Twitter” extension to remove the algorithmic For You tab. posted December 2024 · practice might be outdated
• Google-dork lead generation: search site:linkedin.com {occupation} {area} @gmail.com to surface people in a role and city who list an email publicly - an instant free outreach list. posted July 2024 · practice might be outdated
• US TikTok For You page from abroad (2024): either a US SIM + fresh US Apple ID + US 4G proxy connected via PC hotspot dongle (never connect the phone directly to the proxy - TikTok detects it; VPNs don’t work), or simply rent remote control of a physical US phone for ~$130/month. posted August 2024 · practice might be outdated

Execution Beats Knowledge

• Making money online is an execution game, not a knowledge game. A friend of mine made $2M online in a year without knowing what a “VSL” is; I wrote copy and built funnels “off vibes” before I learned the terms for them. Hand 1,000 people a step-by-step plan and 95% still fail. The differentiator is commitment and action, not information access. You don’t need more knowledge; pick one business model, join a community of like-minded people, and relentlessly take action. In the AI era this is even starker: you can ask AI for a roadmap to any goal, so the bottleneck is no longer information but adherence. Failure is now a follow-through problem.
• Educational content is more dangerous than entertainment because it creates an illusion of progress while nothing gets done. People “contentmaxx” on business knowledge and stay broke, buying a body-language book instead of talking to people, watching Hormozi instead of selling. Information not put into instant action is forgotten fast; learning without application is disguised procrastination. Almost everyone could 10x their income by simply applying what they already know: the value of books is not in reading 500 of them but in reading one good one and applying it before moving on. Before buying any course, check whether you are the type of person who acts on information. If not, paid and free info are equally useless. And if you are broke, do not buy info products at all.
• Learn by building immediately, not by studying theory first. Two people learning AI: one studies the technology behind LLMs, one builds practical solutions for his business from day one. The builder wins. Marketing especially cannot be learned through theory; “this works, I don’t know why, but it works” beats deep technical understanding. Successful people are mostly improvising. They hunt experience, not knowledge; content can only spark a thought, and the real feedback comes from the action that follows.
• A dumb but relentless person outperforms a smart but lazy one. High IQ finds every possible way a good idea will fail; “caveman IQ” relentlessly executes, and execution is where the money is made. I have had hundreds of “million dollar ideas”; only the executed ones mattered.
• Plans never survive contact with execution. Nothing I planned unfolded as envisioned, yet my plans still became successful projects. The path only becomes clear as you walk it. Start before the plan is complete and adapt at each obstacle. The best way to figure out the best way of doing something is to do it.
• Don’t get “cucked by uncertainty.” If checking 100 boxes guaranteed $1M, you would attack the list relentlessly; the only difference in reality is that you must guess the steps. Act like the checklist is real and attack your best guess at it. When people tell me “you’re oversimplifying,” my answer is “no, you’re overcomplicating.” Uncertainty paired with learning is not gambling, it is a puzzle, and doing nothing because you are uncertain guarantees failure.
• Procrastination grows in proportion to how much a project benefits you. The brain avoids possible disappointment by delaying the highest-stakes work, so the urge to procrastinate signals importance, not irrelevance. Resistance compounds. The longer you delay, the harder the task becomes. Push through anyway.
• Difficulty is not a proxy for value. My best successes were often the easier projects; “everything worth doing must be hard” is conditioning. Work only feels hard when you don’t enjoy it, and easy wins count fully. Relatedly, a degree of laziness is an asset: it forces you to find the path of least resistance instead of brute-forcing an approach that clearly isn’t working.

Speed, Testing & Urgency

• The gap between idea and execution predicts failure. Successful operators go from idea to sales page to driving traffic within 24 hours; doubt-driven deliberation kills valid ideas over single uncertain aspects. Decisiveness and speed of execution are the biggest influences on income.
• High performers test immediately and return with data. When discussing an idea, the high performer says “will start testing right now” and comes back with findings; everyone else enumerates reasons it will fail and forgets it. Every Fortune 500 company had a thousand reasons to fail, so hyperfixating on what can go wrong is pointless. Bias toward cheap, fast tests over pre-emptive risk analysis.
• Blind consistency is a self-help psyop. I have never succeeded with anything that did not show results within one month. Use a roughly one-month feedback window: no signal of traction means change the approach, not double down. Blind patience on a bad vehicle just delays the inevitable quit. A shared trait of my very successful friends is urgency: if something does not work fast, it gets abandoned. Patience belongs inside a validated vehicle. Since failure is inevitable anyway, fail fast rather than postponing it with more studying.
• Two rules for ideas: never pitch an idea unless you’re ready to start acting on it immediately, and give away every idea you’re not going to act on. Hoarding unexecuted ideas has zero value; sharing them builds goodwill.
• Manufacture urgency with real stakes. My commitment-device format:
  1. Place meaningful money (e.g. $10k) in escrow.
  2. Define goal X and timeframe Y.
  3. If you miss the deadline, the money goes to an organization you would hate to fund. Loss aversion plus a repugnant beneficiary creates the urgency most people never generate voluntarily.

Focus, Obsession & Lock-In

• Being “locked in” means neglecting everything else for one primary goal, not spreading yourself across a scattered self-improvement checklist. You cannot beat someone putting in 16-hour days on one thing when your focus is split across affirmations, cold showers, journaling, and content recording. Every successful online-business person I know has gone through obsessive lock-in phases: weeks in a room working on a single thing, voluntarily blind to everything else, typically cycling lock-in with decompression. My test: could you work on one thing in a room for 90 days? Consuming lifestyle content while broke is a red flag that you’re in it for the reward, not the process.
• Dedicate a 1-3 year period to one ultimate goal, but only with a concrete plan. Trading ages 21-24 buys a 100x better quality of life from 25-40; “do both at once” fails for people who can’t focus after dopamine-heavy activities. Aimless monk-mode solitude is as wasteful as endless partying. Cutting off fun with no plan or progress is the saddest failure mode. My stronger version: every young man should consider a 1-2 year isolation phase focused solely on getting his finances right, accepting the real social costs for what’s on the other side.
• For burst-wired (ADHD-type) people, manufacture 4-6 week periods of hyper-obsession around one goal, then move on. This produces far more progress than forced, sustained moderation.
• Obsession is malleable. Deliberately get addicted to things that benefit you. All of my successful ventures came from steering my addictive tendencies toward useful obsessions. If you’ll always be addicted to something and the object can change, choose it consciously; obsession is the moat in business, fitness, and any domain. Corollary: replace addictions rather than removing them. Quitting a dopaminergic habit cold turkey leaves a hole something will fill by default, so choose the replacement deliberately.
• Master one business model; refuse to opine outside it. I got rich without knowing most guru buzzwords because I never diverted attention from my one model. Buzzword density inversely correlates with income.
• Splitting attention across many projects caps progress. After realizing scattered attention was my bottleneck despite success, I began declining all new project conversations and DMs for the rest of the year. Deliberately refuse new projects and put undivided attention into current ones.
• Work-life balance is a luxury you earn later. Be “chronically online” and fully immersed until you’re making decent money with systems in place. The more I “touched grass,” the less I made. Balance across all facets of life is the hardest thing to achieve; don’t attempt it prematurely.

Dopamine & Environment

• Dopamine detoxes are misguided; the problem is the source, not the amount. Dopamine deficiency resembles depression, and successful people are flooded with dopamine, from work itself. Getting dopamine from watching your business grow versus doomscrolling is the same neurochemistry with opposite life outcomes. Never restrict dopamine; control which activities it comes from. The detox framing fights biology instead of using it.
• The anti-brainrot protocol - remove recommendation algorithms:
  1. Disable or ignore the For You page; use only the Following tab on Twitter/X.
  2. Hide the YouTube homepage; use only Subscriptions (browser extensions exist for both platforms). Because you follow a finite list, you quickly run out of content, breaking the infinite-scroll loop while keeping the useful accounts. Social media becomes a tool instead of a dopamine trap.
• Discipline problems travel with you. If you can’t focus in a boring hometown with zero distractions, moving to a distraction-rich hotspot like Thailand makes it worse. Choose environments by their absence of temptation, not their aesthetics as a “grind destination”, and never travel expecting to “lock in”. I have never seen it work.
• A cheap, distraction-free environment beats an expensive one. My math, splitting a place with friends: a nice apartment three ways at roughly $700 each per month, gym at $50/month. About $15,000 total funds a full year of zero-distraction lock-in.

Routines & Real Productivity

• The 5AM “millionaire habits” canon is a psyop. I tested weeks of 5AM waking and found zero productivity gain. An actual millionaire’s morning: wake up, check the phone, refresh analytics, work. The more locked in someone is on morning-routine theater, the less money they make. Copying successful people’s routines is cargo-culting. Meditation, reading, and sauna in a precise order made nobody a billionaire; obsessive problem-solving plus enormous work volume did. Bolting wellness rituals (ice baths, saunas) onto procrastination sessions is a band-aid unrelated to the actual bottleneck. Judge routines by output, not aesthetics.
• Optimize your schedule around your market’s timezone, not your own. If you sell to the US from elsewhere, waking at 5-6AM can mean sleeping exactly when your market comes online. The “get after it before your competition” crowd is asleep during peak hours for engagement, sales, and networking.
• Measured honestly, even successful people do only 1-2 hours of truly focused, needle-moving work per day (use a timer that pauses whenever attention drifts). Output comes from those focused hours, not total hours at the desk. Nobody is actually “locked in 24/7”. Even elite performers scroll, game, and watch shows; converting five hours a day into measurable progress puts you ahead of virtually everyone. Measure days by progress produced, not hours performed.
• Open devices only with a stated intention; close them when it’s done. My best work: open the laptop with one specific task in mind, work 2-4 focused hours, close it for the day. My worst days: 12 hours flicking between 20 tabs where “research” and project-management tools masquerade as work. Three hours through a checklist one item at a time beats a full day of false productivity. 30 minutes of clear intent beats a week of fake work.
• Confine each day to three truly needle-moving tasks, each written down with an explanation of how it advances your ultimate goal. This forces alignment instead of busywork and beats a 100-item to-do list. The disguised time-wasters to cut: “researching” on Twitter, “business talk” in group chats, and consuming “educational” material. Cutting them compounds roughly 10x faster progress.
• Audit yourself daily with one question: what did you actually DO yesterday, not plan, research, or think about, that moved you directly toward your goal? Asked every morning, it exposes fake productivity; break goals down to daily granularity rather than 30-day sprints.
• Automate non-business goals into the schedule so they happen without conscious effort: want to lose weight, eat the same three meals every day; want muscle, train first thing in the morning. Preserve willpower and focus for work.
• 16-hour days are only worth it as sprints that build systems which work without you. Grinding is working hard enough that you no longer have to grind. My end state: about 30 minutes of administration each morning sustains the lifestyle. Year-round 16-hour days for personal gain are a failure of system design, not a flex.

Systems, Tracking & Goals

• Success largely comes down to identifying the metrics that matter, then tracking and optimizing them. Anything untracked gives you no sense of progress or direction. If it’s not tracked, you’re winging it. The habit transfers across business, health, and skills.
• Protocol + tracking removes decision fatigue. My 20-day fitness experiment: identical meals daily, progressive overload, 1 hour of daily cardio. Lost 4kg and 4cm off the waist, resting heart rate 58 to 48, HRV 70 to 135. The management rule is second-grade math: if average weekly weight is down, keep macros; if not, add cardio or cut calories. Most goals are a mental battle, not an information problem. Create a protocol, track the data, turn your brain off, and execute.
• Systemize habits rigidly with zero leeway. I can stick to anything I “autistically systemize” with visible data tracking; building in flexibility (“cheat meals”) is what kills adherence.
• Turn goals into a concrete checklist and daily habits. Most people know what they want but never map what achieving it requires, so they drift while feeling ambitious. If you can’t break a goal into tasks, that breakdown is itself the next step. Afterwards only discipline can be blamed. My AI prompt for this: an “elite execution strategist” clarifies starting point, timeline, resources, past attempts, and obstacles one question at a time, pushes back on vague statements like “make more money,” then outputs a goal breakdown, engineered habits, daily/weekly/monthly checklists, time blocks with habit triggers, and a measurement system.
• Set 8-week goals instead of yearly goals. People massively overestimate a year (and coast on the illusion of time) while massively underestimating 8 weeks.
• Break every problem down to its root cause with repeated “why” questions. Most people’s many surface problems trace to the same one to three root causes. Example chain: “I’m broke” → “I live paycheck to paycheck” → “I rely on a dead-end job” → “I can’t identify profitable opportunities” → “I’ve never studied where money flows online” → “I consume information but never implement; I have no systems for taking action.” Now attack the actual cause. My Root Cause Analysis AI prompt interviews you one question at a time, probes at least 5 levels of “why,” flags inconsistencies and avoidance, distinguishes symptoms from causes, and outputs the cause-effect chain plus initial fixes.
• The bare minimum to claim you’re “trying” to build an online business: actively scouting potential products, researching Reddit/X/TikTok for the market’s biggest pain points, having a funnel and sales assets live, and spamming content on YouTube/X/short-form to feed the funnel. 99% of people who say they’re trying do none of these. “I’m trying” without daily action is cope.

Who to Learn From

• Learn from people one or two levels above you, not billionaires. Someone making $10k-$50k/month online is far better positioned to teach a beginner how to reach $10k/month than Bezos or Musk. People who’ve operated at the top for years are foreign to how the ground level works, which is why billionaire advice reduces to platitudes like “work hard, focus on changing the world.” People slightly ahead remember the terrain, their tactics still apply at your scale, and their help costs less. Proactively source advice from that tier rather than waiting for it.
• Work inside a successful operation - the cheapest education available. My roadmap for a broke beginner:
  1. Study a popular, in-demand skill (e.g. popular video editing styles) and learn it.
  2. Browse Twitter/forums and ask people with successful operations to work for free.
  3. Ask for compensation once you are integral.
  4. Observe and study how the operation runs behind the scenes.
  5. Use what you learned to build something similar of your own. I credit much of what I know to running a short-form marketing agency, a behind-the-scenes view of profitable businesses that cannot be transmitted in words, only lived. Nine out of ten people would get further by helping someone established than by insisting on being the face of their own brand from day one; ego, the need to be the star, is a common cause of failure.
• The mentorship price filter: nobody worthwhile mentors one-on-one for under about $5k. A $250 one-on-one call from someone claiming wealth is a red flag, because the opportunity-cost math doesn’t work for a genuinely successful operator. Cheap productized info (documents, communities) from legit operators is fine because it scales; cheap personal time is the tell.
• Formal education cannot keep pace with technology. Curricula run roughly a decade behind (a fourth-year data analytics student who had never heard of Cursor or the term IDE; design classes teaching hand-written raw HTML). Treat school as insufficient and build a continuous self-education loop, using AI itself as the tutor.
• Re-read formative books periodically. Concepts only fully land after you have lived them. Think and Grow Rich’s specialized knowledge, mastermind, and sixth-sense ideas felt understood at first read but only became real after success. Deeper understanding on re-read is a measure of your growth. In the LLM era, though, reading cover-to-cover to “acquire knowledge” is inefficient: have an agent with a knowledge base about you extract only what applies to your case, then read those parts. Time applying information beats time consuming it roughly 100 to 1.

Network & Community

• 90% of success is network, and the formula is mechanical: become great at something that benefits powerful people’s business or life, be useful to them, and the network forms. Networking is never one-sided extraction. Nobody helps someone who offers no value. If you want help from somebody, figure out how to help them first; and helping people without expecting anything back is the most effective long-term strategy. The network that drives most of my income came from freely giving help and insights first.
• Proximity to successful people is the highest-ROI activity. It’s much easier to succeed when you regularly talk to people no smarter than you who are hitting your goals. Proximity normalizes the goal and transfers tactics.
• Get into group chats and niche communities where operators exchange unfiltered ideas. The next big companies are being plotted in random Telegram groups. This is my version of Napoleon Hill’s mastermind principle. People who succeed online are invariably active in niche communities (I credit forums like BlackHatWorld; being active on X counts): daily participation supplies information intake, idea iteration, and immediate feedback you can’t get alone. One friend who has already made it online provides real-life proof that massively boosts your odds; communities are the substitute if you lack that friend.
• Run two social circles. Cutting off old friends is over-glorified hustle advice and usually a mistake. You’ll never click with anyone like the people you grew up with, so keep them unless they’re an actively bad influence. But build a separate circle of business peers who raise your baseline of what seems achievable: if your circle makes $100k/month, $25k/month feels easy; if everyone lives paycheck to paycheck, ambition atrophies. To enter valuable circles you must first become valuable yourself. And remember nobody on money-Twitter is your friend by default. People talk to you when there’s mutual benefit; don’t confuse networking with friendship, though real friendships can grow from shared incentives.
• Cold outreach compounds absurdly. A single DM on a random forum led to a partnership that made me $250k in a year. The expected value of well-targeted outreach far exceeds its cost. When reaching out, present competency: profitable operators have piles of tasks to delegate but can’t trust the 95% of applicants who show none. A portfolio, previous work, and a precise offer are the entire filter. With sophisticated buyers, skip scripted persuasion: I instantly block “let’s hop on a call to discuss pricing” pitches. My preferred format: “yo, I offer X, let’s work”. State the service and price upfront; nobody reads long AI-written paragraphs.
• Publicly posting niche expertise is a living resume. While “autistically posting AI sauce,” I received numerous job offers in my DMs despite being a 16-year-old dropout with no technical background. Demonstrated knowledge in public outperforms credentials.
• When helping friends get into online business, qualify them first. Despite five-figure-day screenshots and repeated explanations of the full model, only one of my real-life friends had the discipline to build an income stream from it. Most will not follow through no matter the proof. Still try to put people on, but learn to identify who actually wants the assistance.

Motivation, Identity & Belief

• You only take actions that correspond with your identity, so income growth follows identity change. I 10x’ed my income when I started believing I was “the type of guy who prints money out of thin air.” Total self-belief is the prerequisite because belief drives behavior: someone who believes he is worth $10M starts making the moves of a person worth $10M. That, not magic, is why the “law of attraction” works. If you truly believe your product is superior, selling it becomes easy; teams that believe they are the best play like it. My practical reframe: build the belief through a pipeline of live projects. Having so many promising projects running that you subconsciously believe you’ve already made it produces relentless action; affirmations alone do nothing.
• Approach every business with delusional confidence: if it has worked for somebody else, it will work for you. I doubted everything yet succeeded at almost everything I tried. The doubt contributed nothing but delay. Nothing kills doubt like winning, and confidence should also be earned by doing the things you tell yourself you will do.
• Material desires are weak motivators; a survival-grade “why” is a competitive advantage. I know no successful person driven primarily by wanting nice things. People motivated by purchasable rewards treat business models as interchangeable vehicles and quit. It is hard to compete with someone who has to provide for the people depending on them, while grinding only to buy flashy things is measurably harder to sustain.
• A different difficulty appears after $10k/month: comfort kills ambition. Once money covers a good life and time freedom, money-as-feedback stops motivating. People doing big numbers are driven by something alongside money: status, validation, or the feedback of building things people use. Diagnostic question: “What would my day-to-day look like with 8 figures in the bank?” How you’d behave once work became optional exposes your real intentions.
• Ego is why 99% of people who want to get rich never do. They want to be rich AND cool AND respected, but making money is mostly an unappealing process of selling boring-but-valuable solutions. The people willing to look uncool doing boring things get rich. As a beginner, deliberately focus on the boring, unsexy work and ignore anything that sounds attractive or cool. You’re not “building an empire” yet, you’re trying to quit your job, and sexy-sounding advice is optimized for engagement, not your stage.
• Luck is positioning. You are always one idea, connection, or decision away from a snowball event; maximize the odds with four behaviors: keep learning like a beginner, talk to people inside and outside your industry, relentlessly act on your best ideas, and make decisions fast. “Is luck really that random if you have to position yourself to get lucky?”
• Treat action itself as positive feedback, not just revenue. The observable traits of successful people: obsession over one profitable skill; enormous communication about the right things (I exchanged 300k DMs with my business partner in a year); curiosity about obscure topics like history and psychology; interpreting taken action and validated ideas as wins rather than treating revenue as the only signal (the revenue-only mindset is why people quit); and a bias for rapid testing.

Money First, Passion Later

• If you’re young and even slightly lost, chase money before passion. Very few people have a durable genuine passion. Most claimed passions are status desires that can’t fund a life, and most are brutally hard to monetize (imagine building a table tennis offer with zero business fundamentals). Money-earned skills plus covered survival make a later pivot into passion far more feasible; Elon got rich first and chased passion projects second. Money buys the freedom to explore passions without the crippling stress of a 9-to-5 and bills.
• Passion is malleable: enter an industry because it is profitable and let the reward loop create the passion. A common trait of successful people is picking profitable problems, then falling in love with the process because solving them gets heavily rewarded. Elon did not found PayPal out of passion for payment processing. The ideal stack: make a lot of money, help people (the best way to make money), and build things you enjoy. Hit all three and you’re fulfilled.
• Chase impact, not arbitrary revenue milestones. Focus on becoming a positive influence on the operations of rich people who want to save time, and on delivering good service and pushing good products. Income darts past “$10k/month”-style goals without you tracking them.
• Avoid entertainment paths - pure survivorship bias. Only about 1% of millionaires come from entertainment (athletes, musicians, content creators), yet these are the most pursued careers because the winners are the most visible; even “making it” usually requires promoting scams to earn significantly. Choose paths where the odds are structurally better even if nobody films them.
• Anonymity is viable and often preferable. I made over a million online while fully anonymous. No sales calls, no clients, no public-figure costs. Most of the biggest earners on X are anonymous; “super affiliates” avoid publicity to protect their operations. Faceless brands work when the content is genuinely valuable: detached from ego, the account just posts value and the only reason to follow is the quality of ideas. Running with your face is easier, but anonymity preserves sanity and operational security, and you lose little.

Skills, Career Strategy & AI

• Build a foundation of skills before buying any business-model course. 95% of biz-op courses fail buyers because they’re marketed to beginners while silently assuming a foundation: creative skills (video editing, design, copywriting), technical skills (programming, AI, logic), understanding where money flows online, platform algorithms and attention, hiring and project management, payment processing, all marketing types (email, paid, organic), offer crafting, funnels, and sales. Courses teach one part of a massive system; build the foundation through many small projects first. More fundamentally: most people can’t answer what value they bring to the marketplace. The first step is not monetization but picking a skill, validating that its market is profitable, and building specialty; monetization follows.
• Skill acquisition is a sprint, not a years-long process. Any capable person can reach expert-level understanding of any subject in under three months of living and breathing it. My operating loop: learn a new thing, apply it instantly, repeating every 30-60 minutes.
• Surface-area-max: a broad stack of seemingly unrelated skills is one of the strongest predictors of success. Treat every idea as an asymmetric bet that stacks skills (project management, distribution, copywriting, coding, editing, design) even when the venture fails. My own stack: Minecraft servers, selling digital services on internet-marketing forums, a short-form agency, content-locking CPA campaigns, YouTube automation, selling info and software, and co-founding one of the biggest trading communities. Skills fuse multiplicatively. Generative AI plus video editing is worth roughly 10x either alone. Failed businesses are never wasted: each teaches transferable skills that compound into future ventures, and first-hand experience across verticals is why some people generate endless good ideas. The inverse loop is why most people stay stuck: idea → failure to act → zero output → zero experience acquired.
• Fully lean into natural strengths; outsource or refuse everything you’re naturally bad at. Years of grinding without results are useful mainly for uncovering what your real strengths and weaknesses are. Then allocate all effort to the strengths rather than grinding to fix weaknesses.
• If you’re broke, don’t learn to code as your path to money. It takes too long to get good enough to charge substantial amounts. Learn skills that directly sell products or services: content creation, media buying, SEO, outreach, sales. Coding is fine if you genuinely love it, but it’s not the path of least resistance to cashflow; overly technical founders also often lose their product edge. Know enough tech to hold the vision, offload implementation.
• “Learn AI” as a goal is guru propaganda; AI skills should be a byproduct of solving profitable problems. I learned Python to auto-upload YouTube Shorts, LLMs to generate copy fast, and coding agents to build internal team tools, each because a money problem demanded it. If you don’t routinely discover profitable problems to solve, no amount of AI skill will make you rich. The real leverage comes from fusing domain expertise with AI: only a skilled copywriter who also understands prompt engineering can build the best copywriting prompt. Learn one or two high-value skills deeply, then stack AI on top.
• AI is repeating the machinery playbook - position yourself as the overseer. Machines replaced repetitive physical labor and created machine-operator roles; AI is replacing repetitive mental labor and creating roles for people who design and supervise AI workflows. Businesses will always choose the faster, cheaper AI option over hand-crafted work. Put yourself in an owner’s shoes choosing between paying an artist and waiting days versus a finished product in 30 seconds. Creatives must integrate AI or reposition; passion art can continue as passion, but as a career the workflow must change.
• AI empowers genuinely creative people rather than replacing them. Creativity has always been about ideas, not execution mechanics. AI removed the technique barrier between having an idea and producing it, which is why creative people who never learned craft are suddenly producing strong work. AI cannot invent new things, only remix, so creativity stays scarce; and output quality is bottlenecked by how you structure prompts, an extension of logic, making creativity and logic the two most important skills. Career implication: designers should pivot from executors to ideators (thumbnail designer → thumbnail strategist), and companies will consolidate creative roles into one “creative prompt-engineer” generalist covering graphics, video, and web.
• Curate inputs, don’t eliminate them. Creativity is a balance of consumption and creation. The “stop consuming, only create” advice is skewed: I know no creative person who doesn’t consume enormous amounts of content, and dopamine-detox isolation would cut my idea flow to a tenth. Output quality is capped by input volume. Someone raised in an empty room has nothing to write about. Cycle deliberately between heavy consumption (books, videos, podcasts, tweets, curiosity-driven rabbit holes) and creation in a loop of “learned something interesting → post about it.” If output feels dry, the fix is usually more and better inputs.
• Facebook is an underrated marketing channel. Never getting into Facebook marketing is one of my biggest regrets. Dismissing it as “a platform for old people” was precisely wrong, since older users are one of the easiest demographics to sell to.

Money Management

• Do not blow your first real money on status purchases. I blew my first few hundred thousand dollars, and being broke in a nice car is more miserable than being broke without one. Internet income is volatile. Bank the surplus until income is stable and diversified before any lifestyle inflation.
• Make internet money fast and convert it into security. The game changes fast. You can lose it all within a month, and it is a statistical trend that most young online earners will be broke within 7 years because flashy accounts spend every dollar. My split: 50% enjoying life, 50% into property and investments. If you’re young and printing online, buy property and invest before lifestyle inflation eats it. Don’t be the guy with a supercar and no assets.
• The real metric of financial security: if you stopped working today, how long before you run out of money? My own answer was 5+ years; $100k liquid provides life-changing security in a low-cost country. Use this question to test anyone’s “passive income” claims, including your own.

Filtering Advice & Spotting Grifters

• Ask about incentives, not claims. Instead of “why does this influencer say X is bad?”, ask “what incentive does this influencer have in me believing X is bad?” This single reframe protects you from most engineered narratives online.
• The one-rule filter that eliminates 90% of gurus: never take advice from anyone who puts materials or status above taking care of his own family. Supercars-and-girls marketing signals someone optimizing to impress strangers, not a model worth following.
• Flexing is an inverse signal. Judge a money-Twitter figure’s legitimacy by how often they post revenue screenshots (most circulating ones are old); if someone feels the need to flex something, assume it’s not part of their everyday life. Nobody downplays themselves, and some people literally empty their bank accounts to buy a supercar because it drives impressions. A $200k car in this space is a marketing expense, not evidence of wealth. Genuinely successful operators print money quietly.
• Distrust certainty theater. Nine times out of ten a “know-it-all” persona is a snake-oil setup for selling you something later; the smartest and most honest operators openly admit how much they don’t know. Intellectual humility is a trust signal.
• Distrust branded frameworks with a single cult-leader figure (named diets, renamed biz-opps). Blindly adopting a framework “made for the masses” because the guy “feels legit” outsources your judgment. Take in all inputs, run your own experiments, track your own data, and find your own way.
• No advice is a universal law. Money Twitter generalizes from single experiences: “it failed for us so it never works” and “it worked for us so nothing else does.” Most contradictory guru claims are both true for their specific circumstances and false as universal rules; apply this filter to everything you consume, including growth tactics.
• Never take life-affecting advice from internet personalities. The Belle Gibson case: an Australian wellness influencer claimed she cured terminal brain cancer with natural remedies, built an app, cookbook, and large following, collected over $300,000 in charity donations she mostly kept, and never had cancer; followers died after abandoning chemotherapy on her advice. It doubles as a marketing lesson in how powerful, and dangerous, transformation narratives are.

Payment Processing

Coding Agents

Social Media Tools

Analytics

Forms