- “context is king. the more context you give to the ai, the more it can get close to putting out what you actually want it to put out.”
- “good ideas are the fuel for content, but usually good ideas come from sifting through hours of youtube videos on a walk or reading a good book — this is just a way to kind of force that to an extent and still maintain the quality without having the llm actually write it for you.”
- “doing things this way isn’t like outsourcing your agency or your cognition to ai. it’s actually getting you closer to being high agency — the best way to learn is by building projects, failing, iterating.”
dan koe’s content system is built on a single weekly anchor — a newsletter — that gets repurposed across every platform (twitter/x, linkedin, instagram, youtube, podcasts) with minimal additional effort, using twitter as the real-time litmus test for which ideas have legs. rather than asking ai to write content for him, he uses llms as a structured research and deconstruction tool — summarising long-form videos, extracting building blocks from high-performing posts, and generating interview-style prompts that pull context from him before producing first drafts. the core meta-skill he teaches is a two-step prompt engineering process: first have ai break down any piece of great content into its psychological and structural principles, then use a prompt-creation prompt to turn those principles into a reusable, context-rich prompt that writes in your voice.
- repurposing is not laziness — it is leverage. one deeply considered idea, written well once, can cascade across twitter, linkedin, instagram, youtube, and podcast with minimal friction. the quality bottleneck is the idea itself; once that is solved, distribution across platforms is nearly automatic. most creators exhaust themselves producing mediocre content for each platform separately when they could produce one excellent thing and let it travel.
- the llm’s job is to give you building blocks, not finished output. asking ai to write your tweets produces the worst tweets you have ever read. asking ai to deconstruct three tweets you love into their psychological patterns, structural archetypes, and core paradoxes gives you raw material that sparks your own thinking — and the writing that follows sounds like you, not a machine.
- experimentation plus spin-offs is the entire growth formula. post consistently, identify which specific posts bring in followers (not just likes), then dedicate roughly 30% of your output to deliberate spin-offs of those proven ideas while using the remaining 70% to keep experimenting. rinse and repeat as new winners emerge.
dan koe’s core intention is to show that high-volume, high-quality content creation is not a talent game but a systems game — and that llms, used correctly as thinking partners rather than ghostwriters, can give any person with genuine ideas an unfair distribution advantage right now. the deeper message is that ai accelerates learning-by-doing, getting you to your first failed iteration faster so you can refine faster.
- single anchor content model — everything originates from one weekly newsletter; all other platforms are downstream derivatives of that one piece
- twitter as litmus test — short-form posts act as low-cost idea validation before investing in long-form content (newsletter, youtube, podcast)
- idea density and novel perspectives — the differentiator that builds genuine audience authority; at least one mind-expanding insight per piece that most people do not know
- swipe file and brand synthesis — collecting examples of writing you admire, then synthesising them under your own worldview and voice rather than copying
- llm as deconstruction engine — using ai to reverse-engineer the psychological patterns, structure, and techniques behind great content rather than to generate content directly
- two-phase prompt architecture — phase 1: context gathering (ai interviews you for the inputs it needs); phase 2: execution (ai uses those inputs plus a structural guide to produce output)
- prompt-creation prompt — a meta-prompt that takes raw instructions and context requirements and produces a comprehensive, reusable prompt; the core productivity multiplier in dan’s system
- spin-off growth loop — once a post type reliably brings followers, systematically create variations of it while keeping a portion of output for fresh experimentation
- paradigm shift marketing — a framing dan discovered by having ai deconstruct a landing page he admired; illustrates how the deconstruction process surfaces transferable principles
- failing faster with ai — using ai-generated first drafts not as final output but as a way to accelerate the iteration cycle; get to failure and refinement sooner
- weekly content rhythm — write one newsletter section and three social posts every morning during a fixed 2-hour writing block; record one youtube video one day per week; everything else is repurposing
- youtube niche research — go to accounts in your niche, filter videos by most popular, list 10 topics you could cover from your own perspective; use the topic and angle, never the content itself
- llm research workflow — paste long youtube videos or pdfs into gemini (for large context windows) or claude/chatgpt; use the summary as a research reference rather than note-taking during consumption
- deep post generator prompt — feed newsletter or video into a prompt that extracts: 5 compelling post ideas, core paradoxes, key quotes, transformation arc, core problems, key examples, action steps; use outputs as raw building blocks, not copy-paste tweets
- content ideas generator — prompt that produces 60 idea starters categorised as: harsh life advice, counterintuitive truths, core pain points, key insights, big ideas; used as starting points, not finished posts
- youtube title generator prompt — trained on your 15 best-performing titles; fed the newsletter to produce 20–30 title options following the same psychological patterns; cross out weak ones, test the best
- three-tweet deconstruction → unified guide — pick three posts you admire, ask ai to break each one down (structure, psychology, techniques, context needed), then combine all three into one master writing guide
- two-phase prompt creation workflow — take the master writing guide plus a list of context requirements; feed both into the prompt-creation prompt; output is a reusable interview-then-write prompt personalised to your voice
- superx / tweet hunter x browser extension — view any twitter account’s top-performing posts; use to study structure and ideas from writers you aspire to emulate
- structure-idea swap exercise — take a high-performing post structure and swap in a new idea; or take a strong idea and rewrite it in a different structural format; one idea becomes two posts
- offer creation via ai — ask for a detailed guide on a specific expert’s method (e.g. alex hormozi on offers), then ask what context it needs from you, then use the prompt-creation prompt to build a two-phase offer-building prompt
- what is my single anchor content format — and am i currently building everything downstream from one strong weekly piece, or am i creating separately for each platform and exhausting myself?
- which of my existing posts or pieces of content have actually brought in new followers or meaningful engagement — and have i systematically created spin-offs of those, or just moved on?
- what three pieces of content (posts, scripts, landing pages) do i genuinely admire and want to write like — and have i ever deconstructed them properly to understand why they work?
- what context would an ai actually need from me to write in my voice — and have i ever taken the time to define that comprehensively?
- am i using ai as a ghostwriter (which produces generic output) or as a thinking partner and deconstruction tool (which produces leverage)?
- people: dan koe (x, youtube, newsletter — all worth following); alex hormozi ($100m offers — referenced for offer creation framework); zach pogrob (referenced as a writer worth deconstructing)
- tools: gemini 2.5 (large context window for long videos/pdfs); claude with extended thinking (for prompt creation); superx browser extension (top tweets by account); tweet hunter x; ideabrowser.com (dan’s own startup idea tool)
- implied reading: study of human psychology as it applies to content (no single book named, but the framing of “algorithms are based on human psychology” points toward persuasion and behavioural science literature)
- establish a daily 2-hour writing block dedicated solely to one newsletter section and three social posts — protect this time before anything else
- maintain an active swipe file: every time you encounter a post, script, or piece of copy that stops you, save it immediately with a note on why it worked
- before writing, spend 5–10 minutes reviewing the swipe file or running the content ideas generator prompt to have a starting point ready
- at the end of each week, review which posts performed best and tag them as candidates for spin-off content the following week
- build the core prompt library: (1) youtube title generator trained on your best titles, (2) deep post generator, (3) content ideas generator, (4) the two-phase tweet-writing prompt built from deconstructing three posts you admire
- identify three posts or pieces of writing you genuinely want to emulate; run the full deconstruction process on each and combine into a personal writing guide
- define your content anchor — one format, one platform, one day to produce it — and map every other platform as a downstream derivative
- set up superx or tweet hunter x and identify five to ten accounts whose writing style you want to study; begin visiting their top posts weekly
- run the two-phase prompt creation process at least once end-to-end this week on a real piece of your own content to internalise how the system works