Content Formats
The overall shape of a post. Hooks = the opening move. Formats = what the whole post IS. Most formats require different things from you — check requirements before committing.
Long-Form Essay
risk: medium5,000+ word essay making a strong, specific thesis. Rare format — most people don't write it, which means the ceiling is very high when it lands. Requires cultural timing to escape X.
Anatomy
- 1. Strong thesis in the first paragraph — no warmup
- 2. Specific analogies (historical, scientific, or cultural)
- 3. Evidence from lived experience or primary data
- 4. Counterargument addressed and dismissed
- 5. A specific prediction or call to action
- 6. Publish on personal site → X thread summary with key quotes → wait for pickup
Platform path
Personal site / Substack → X thread with hook quote → newsletter amplification → if it's good: press pickup (Fortune, Inc, TechCrunch) → CNBC / podcast → LinkedIn lag
Examples
Something Big Is Happening
Machines of Loving Grace
Requirements
- Strong contrarian thesis (or strong pro-thesis with unusually specific evidence)
- Cultural timing — the bubble must be ready to hear the argument
- 50+ early readers / amplifiers on day 1
- A hook quote that works out of context as a standalone tweet
Ceiling: 8+ figures in views, press coverage, Wikipedia entry, reusable frame
High effort, low frequency. Once-a-quarter at most. The 'February 2020 moment' frame is now overused — don't copy that specific comparison. New analogies only.
Shower Thought Drop
risk: lowShort post (1–3 tweets/paragraphs) with a coined term or provocative observation. High variance: usually nothing, occasionally a dictionary entry. The author's credibility and the precision of the phrasing are everything.
Anatomy
- 1. The observation in the fewest possible words
- 2. One clarifying sentence or specific example
- 3. No call to action — let it land
Platform path
X → quote-tweets and debate → newsletter citations → job descriptions / conference slides
Examples
Vibe Coding
"There's a new kind of coding I call 'vibe coding', where you fully give in to the vibes..."
Context Engineering
"I really like the term 'context engineering' over prompt engineering."
Requirements
- Genuine field presence (not commenting from outside)
- Crisp, precise phrasing — the term must be better than existing alternatives
- Real observation, not forced coinage
Ceiling: Standard vocabulary for the field (months to years of cultural shelf life)
Completely unpredictable. Can go nowhere or become a dictionary entry. Karpathy himself can't predict which of his tweets lands. Write these when you have a genuine observation. Don't force it.
Proof Thread
risk: lowSeries of specific data points from actual work. Leads with a number, walks through methodology, ends with a takeaway. The specificity is the product.
Anatomy
- 1. Lead: [exact metric] in [exact time period]
- 2. Context: what this is, why it matters
- 3. The surprising or non-obvious finding
- 4. Methodology: specific tools, process, steps
- 5. Takeaway: what others can apply
- 6. Optional: what didn't work
Platform path
X thread → screenshot resharing → newsletter curation → DEV Community / Medium writeup
Examples
100% Written by Claude Code
"259 PRs, 497 commits, 40K lines added — every line by Claude Code"
Boris Cherny Workflow Thread
"My setup might be surprisingly vanilla! Claude Code works great out of the box"
YC W26 Record Stats
"3X more companies at $1M ARR than W25. 14% WoW growth on average."
Requirements
- Real numbers from real work (no rounding, no approximations)
- A non-obvious finding that advances an ongoing debate
- Methodology specific enough to be reproducible or verifiable
Ceiling: Long-running citation in ongoing debates; reference in newsletters and investor memos
The absolute requirement: numbers that feel too specific to be made up. '$42,317 MRR' > '$40K MRR'. '259 PRs' > 'hundreds of PRs'.
Debate Response
risk: lowA direct, substantive response to an influential person's strong claim. Works as well as originating the debate if delivered with specific counterevidence.
Anatomy
- 1. Quote or reference the original claim precisely
- 2. Name your position clearly (agree / disagree / reframe)
- 3. Provide specific counterevidence or a missing dimension
- 4. Acknowledge what the original got right
- 5. Land on a clean resolution or escalated question
Platform path
X reply/quote-tweet → debate tree amplification → business press covers the back-and-forth
Examples
Sridhar Vembu (Zoho) responding to Garry Tan's vibe coding kills SaaS take
"Vibe coding just piles up tech debt faster until... If vibe coding is so powerful, why don't vibe-coded versions of email, spreadsheets, or accounting exist yet?"
Requirements
- Credibility in the domain of the original claim
- Specific counterevidence (not just vibes or disagreement)
- Willingness to name the thing you disagree with precisely
Ceiling: As much reach as the original post, plus the 'response' coverage frame
Responses often get as much engagement as the original if you're credible and specific. 'Hot take + named objection + resolution' is a known high-engagement format. Punching down (responding to low-credibility claims) doesn't generate useful engagement.
Archival Recirculation
risk: lowSurface an old, high-quality insight that resonates with a current felt need. Works because great advice is perennial — the audience discovers it as fresh even if it's years old.
Anatomy
- 1. The clip or quote (as specific as possible)
- 2. Why it resonates right now (tie to current discourse)
- 3. The new angle or context you add
Platform path
X → screenshot resharing → newsletter curation → LinkedIn lag
Examples
@StartupArchive_ daily PG / Sam Altman clip curation
"'Sam Altman on the Paul Graham advice that saved OpenAI: Always make an API'"
PG aphorisms recirculated during hiring/layoff waves
Requirements
- Real historical source (not paraphrased or invented)
- Contextual relevance to current discourse tension
- Add something new: a current example, a new angle, or your own observation
Ceiling: Consistent mid-tier reach; builds reputation for curation
StartupArchive as a model: volume + quality curation = sustained reach without original ideas. Works for Seva when there's a genuine current relevance hook.
Agent Workflow Reveal
risk: lowA specific, step-by-step reveal of how you actually use AI agents in your work. Not 'AI is amazing' — but the exact configuration, tools, prompts, and results. The methodology IS the content.
Anatomy
- 1. The outcome first: 'Here's what I now do in 30 minutes that used to take 3 hours'
- 2. The setup: exact tools, models, configuration files (WORKFLOW.md, program.md, etc.)
- 3. The non-obvious part: what surprised you that others won't expect
- 4. What still breaks: honest failure points in the workflow
- 5. The takeaway: what others can replicate immediately
Platform path
X thread → developer community amplification → newsletter curation → DEV Community / Medium writeup
Examples
Boris Cherny 'My Setup Is Surprisingly Vanilla'
"My setup might be surprisingly vanilla! Claude Code works great out of the box"
Karpathy autoresearch program.md pattern
"program.md is a complete research methodology document — a coding agent reads it and executes indefinitely"
Peter Steinberger 'Running 4-10 Agents Simultaneously'
"6,600+ commits in January 2026 alone, running 4-10 Claude Code agents simultaneously"
Requirements
- Actual numbers from actual work (agent count, time saved, commits, experiments run)
- Specific tool names and versions (not 'an AI tool' but 'Claude Opus 4.6 via Claude Code')
- At least one non-obvious finding or failure
- Something reproducible: a WORKFLOW.md pattern, a program.md template, a prompt
Ceiling: Newsletter citation as canonical workflow reference; replicated by hundreds of developers
This is the highest-value format for the engineering/builder audience in 2026. The Boris Cherny model: specific numbers + surprising result + reproducible method. The anti-slop quality is demonstrated automatically by the specificity. 'I ran 4 agents simultaneously' beats 'I used multiple AI assistants.'
Competitive Landscape Honest Essay
risk: mediumA specific market — coding tools, AI search, agent platforms — with an honest, data-backed assessment of who's winning and why. Requires genuine product experience. The definitive reference that others bookmark.
Anatomy
- 1. The question: 'Which [category] tool should you actually use in [current year]?'
- 2. The criteria: specific dimensions that matter for real work
- 3. Per-tool honest assessment with concrete data points
- 4. Your actual recommendation with the 'it depends on' caveats
- 5. What will likely change in the next 6 months
Platform path
Personal site / Substack → X thread with key verdict → newsletter curation → developer community bookmarking → LinkedIn arrival (4-8 week lag)
Examples
AI Coding Tools 2026 (Pragmatic Engineer)
"Claude Code is now #1 by developer love (46%), overtaking Cursor"
Zach Lloyd 'The IDE Is Dead'
"I don't believe the 'Cursor is dead' memes, but 'The IDE is dead' is real"
Requirements
- Real experience using the tools you're comparing
- Specific benchmark or metric comparisons (not just vibes)
- Current data (this market moves every 6-8 weeks)
- A clear recommendation with honest 'when this is wrong' caveats
Ceiling: Definitive reference bookmarked by hundreds; cited in investment memos
The catch: this market moves so fast that a '2026' essay can be outdated by July. Either publish with a clear 'as of April 2026' timestamp and plan to update, or frame it around timeless evaluation criteria that survive model updates.
IRL-to-X Pipeline
risk: highPhysical or real-world action that creates a screenshot-worthy moment, generates X discourse, then escapes to mainstream press. High engineering overhead; high ceiling.
Anatomy
- 1. The physical action (billboard, protest, stunt, event)
- 2. First X moment: photo/screenshot of the IRL artifact
- 3. Discourse phase: takes, counter-takes, outrage
- 4. Press pickup: TechCrunch / KQED / local press
- 5. Second X wave: response to coverage
- 6. Optional: LinkedIn arrival (4–8 week lag)
Platform path
IRL artifact → X screenshot → press → second X wave → LinkedIn
Examples
Artisan 'Stop Hiring Humans' SF billboards
'March for Billionaires' SF (Feb 7, 2026)
Requirements
- A physical artifact that looks ridiculous or provocative as a screenshot
- Clear connection to a current cultural debate
- Ability to withstand negative coverage (not everyone will agree)
Ceiling: Press coverage, cultural reference status, Wikipedia page
Very high engineering overhead. Works best for companies seeking brand-level awareness, not individuals. The unintentional version (March for Billionaires) shows that IRL spectacles can go viral without intent.