reach
resonance
Not "AI is getting better" but "specifically in December, everything changed." Forces the audience to agree or disagree rather than vague head-nodding. Requires a specific, verifiable time marker.
"Not gradually. Specifically [time period]. Here's what I observed: ..."
Why it works
Falsifiable and verifiable — gives audience a shared reference point. Demands a clear position: you either witnessed the same shift or you didn't. Specificity reads as confidence, not hedging.
Examples
"It is hard to communicate how much programming has changed due to AI in the last 2 months... specifically this last December."
@karpathy
· 14M+ views
Became the canonical 'agentic era arrived' narrative anchor. Latent Space coverage. Year-in-review anchor across ecosystem.
"This is the February 2020 moment for AI."
@mattshumer_
· 82–83M views
Fortune full reprint. CNBC interview. 'February 2020 moment' became a reusable frame others applied to new topics.
Risk note: Works if the claim is credible and verifiable. Fails if it feels exaggerated or if the speaker has no demonstrated presence in the field.
Shelf life: weeks to months
resonance
cross-platform
The person who built X reveals they now use X to build X itself. Combines credential (creator knows the tool best) + surprise (recursive loop) + proof (concrete numbers).
"I built [X]. In the last [N] days, [X] wrote 100% of my [X] contributions."
Why it works
Insider credibility plus an irony that demands attention. Specific numbers make the claim unforgeable and carry extra weight. "The snake that ate itself" is a memorable frame that travels.
Examples
"In the last thirty days, 100% of my contributions to Claude Code were written by Claude Code"
@bcherny
· 259 PRs, 497 commits, 40K lines — all by Claude Code. Boris is Claude Code's original creator.
VentureBeat, DEV Community, multiple newsletters. 'The snake that ate itself' framing circulated widely.
Risk note: Requires a genuine recursive use case — not manufactured. This move only lands once per creator per tool.
Shelf life: one-time per creator
resonance
memetic
Declare an established term dead, passé, or insufficient — then propose what actually matters. Forces engagement by creating two camps: defenders of the old term vs. adopters of the new one.
"[Established term] is [dead / passé / insufficient]. Here's what actually matters: [new framing]."
Why it works
Community members who invested identity in the old term get defensive. Adopters of the new framing share enthusiastically. Both camps generate replies, quote-tweets, engagement.
Examples
"I really like the term 'context engineering' over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM."
@tobi
LangChain 'Rise of Context Engineering' blog. Simon Willison coverage. Entered job descriptions and AI engineering course titles.
"[Vibe coding is passé → Agentic Engineering is the new professional default]"
@karpathy
The New Stack, DEV Community, Medium explainers. 'Vibe to agentic' became a discourse arc.
Risk note: Backfires if the old term is still actively useful to most listeners. Must show a real usage gap — the old framing causes actual errors or misalignment. Don't challenge a term just because it sounds stale to you.
Shelf life: months (until new term wins community adoption)
resonance
cross-platform
Insider with real access shares a specific, non-obvious data point from actual use. Specificity is the point — numbers too precise to be made up carry extra credibility.
"[Exact number] [metric] in [exact time]. Here's what drove it."
Why it works
Community members use specific data points as evidence in ongoing debates. The more precise the number, the more it signals real information vs. marketing. "259 PRs" is more credible than "hundreds of commits."
Examples
"259 PRs, 497 commits, 40K lines added — every line by Claude Code"
@bcherny
Became primary proof in the 'agentic era' argument across newsletters and blog posts.
"3X more companies in this batch reached $1M annualized revenue than W25. Also crazy: the fastest revenue growth rate of YC history at 14% week on week growth on average."
@garrytan
Lobster Capital Substack deep-dive. TechBuzz coverage. 'The math changed' framing entered investor discourse.
"Monthly ARR breakdowns with exact per-product numbers"
@levelsio
Sustained following. Lex Fridman Podcast #440. Revenue updates as directional signal for indie products.
Risk note: Requires actual data from actual work. Fabricated or rounded numbers are obvious. '~$40K MRR' reads as made-up; '$42,317 MRR' reads as real.
Shelf life: weeks (specific to the batch, quarter, or period cited)
resonance
reach
A trusted voice declares a specific inflection point was crossed — not gradual improvement but a step function. Requires personal observation, specific date, and a verifiable claim.
"Before [date/event], [X] didn't work. After it, [X] is the new baseline."
Why it works
Validates what observers sense but can't articulate. Those who noticed the same shift amplify heavily — it becomes the shared label for what they experienced. Creates a "before/after" reference point that the discourse cites for months.
Examples
"coding agents basically didn't work before December"
@karpathy
14M+ views. 2.4K+ comments. Canonical 'December was the moment' narrative. Latent Space dedicated coverage.
Risk note: High impact — but requires genuine field presence. Outsider phase change claims get dismissed immediately. Requires specific personal evidence: not 'things are getting better' but 'I ran this task and got this result that was impossible before.'
Shelf life: months to years (if the phase was real)
resonance
cross-platform
A strong, falsifiable claim that forces audience members to pick sides. Best when two credible camps exist and the claim has clear stakes.
"We switched from [A] to [B]. Honest breakdown after [N] months: [verdict]."
Why it works
Creates reply engagement from people defending their current position. Generates a long-running citation in both camps' writing. Business press picks up named debates with two credible parties.
Examples
"[Vibe coding] will compete away over-bundled SaaS companies like Zoho — non-technical operators can build custom tools in a weekend."
@garrytan
Sridhar Vembu (Zoho CEO) public response. 3-day Twitter war. Business Today, FreePressJournal, News9 coverage. 'SaaS is dead' discourse anchor.
Risk note: The more credible the opposing party, the better the debate. Punching down doesn't generate useful engagement. Must have a specific, falsifiable claim.
Shelf life: months (debate arcs are long-lived)
resonance
memetic
A before/after map of vocabulary shifts in the field. Positions author as someone tracking the field's evolution rather than just reacting to it.
"A year ago, everyone said [X]. In [period], the default shifted to [Y]. Here's why that matters for how you work."
Why it works
Creates a shared "we see this now" feeling — readers feel validated. Signals authority over the field's narrative arc. Easy to share as an orientation guide; functions as evergreen reference content.
Examples
"[Prompt engineering → Context engineering: the skill is different from what we called it]"
@tobi
Broad professional adoption of new term in newsletters, job descriptions, course titles.
Risk note: Works best when the vocabulary shift reflects a real practice change, not just taste or trend-chasing. Don't declare something dead that's still alive.
Shelf life: months
resonance
memetic
[Tool A] vs. [Tool B] — here's what actually happened when we switched, or tried to. Live platform wars generate high engagement because everyone has skin in the game.
"We switched from [A] to [B]. Honest breakdown after [N] months. Worth it? [answer with data]."
Why it works
Reply engagement from people defending their current tool. Creates genuine value if the breakdown is specific and honest. Platform wars create long-running discourse arcs that resurface with each new development.
Examples
"I don't believe the 'Cursor is dead' memes, but 'The IDE is dead' is real."
Zach Lloyd (Warp CEO)
Cursor vs. Claude Code discourse. Fortune article on Cursor's uncertain future. Ongoing in communities.
Risk note: Only if Seva has genuine experience and specific data. Hot takes without data are low-credibility.
Shelf life: weeks (platform landscapes change fast in this bubble)
reach
resonance
Build-in-public + milestone reveal + AI leverage emphasis. Anatomy: specific metric → story of what shipped / what broke / what worked → AI leverage revelation → exact numbers → 'thread on how'.
"$[exact MRR] in [exact time]. One person, AI stack, no VC. Thread on how ↓"
Why it works
Validates the 'one-person unicorn' aspiration. Creates both inspiration and FOMO.
Examples
"Monthly ARR breakdowns with exact per-product numbers"
@levelsio
Sustained following. Lex Fridman Podcast #440. Aspirational scaffolding for the whole bubble.
Risk note: OVERSATURATED as of April 2026. '10K in 7 days' threads met with high skepticism. Only works with significantly larger numbers, or an honest failure/pivot story within the format. Differentiation required: either scale (real, large numbers) or angle (failure, unexpected result).
Shelf life: short (format is saturated)
reach
cross-platform
Company or founder posts deliberately provocative content, generates outrage coverage (10–50x organic reach), then later confesses the strategy was intentional. The confession itself becomes shareable.
"The goal was always to rage bait. Here's what happened."
Why it works
Confession generates a second wave of coverage. Authenticity about inauthenticity is its own novelty.
Examples
"the goal of the campaign was always to rage bait"
Jaspar Carmichael-Jack (Artisan CEO)
· 'Stop Hiring Humans' billboard campaign in SF. Intentional typo 'HIRRING' for screenshot bait.
$2M new ARR. Company's biggest growth months. LinkedIn ban. TechCrunch, KQED, Gizmodo. Death threats.
Risk note: Works once. Destroys long-term credibility permanently. Not for someone building sustained thought leadership.
Shelf life: one-time
resonance
cross-platform
An insider or credible observer reveals that something more powerful exists but was deliberately not released. Creates intrigue, forces a discussion of 'what does responsible mean?' The capability reveal is the hook; the ethical tension is the content.
"We built [X]. We're not releasing it. Here's why that was the right call."
Why it works
Combines rarity (most people only talk about what they ship) with stakes (something powerful was suppressed). Forces the audience to engage with a real dilemma: release vs. withhold. The 'insider admission' dynamic makes it feel authentic and significant.
Examples
"Anthropic has been testing a new AI model... after data about the model leaked out..."
Fortune reporting on Mythos, March 26, 2026
· Anthropic built Mythos — 83.1% vulnerability exploitation — then withheld it. Fed briefings followed.
Bloomberg, Fortune, Axios, NBC, CNBC coverage. 'Is this AGI?' discourse. 'Responsible withholding' became a vocabulary term.
"100% of my contributions to Claude Code were written by Claude Code"
@bcherny
· Not a withholding example, but the same 'insider has something most people don't know' dynamic.
VentureBeat, DEV Community. 'The snake that ate itself' frame.
Risk note: Requires genuine insider access and real stakes. Fabricated or exaggerated 'withheld capability' claims will be called out immediately. Only usable if you actually have the information.
Shelf life: days to weeks (high urgency content)
resonance
reach
Document the exact journey from zero to a specific revenue figure, with specific tools, timeline, and methodology. The counterpart to the saturated '10K in 7 days' thread — requires real scale or an honest failure story within the format.
"$[exact ARR], [exact timeframe], [exact headcount]. Here's the actual stack ↓"
Why it works
Validates the 'AI-enabled solo company' aspiration with concrete evidence. Exact numbers (not rounded) signal authenticity. The stack reveal is actionable: others can replicate the method.
Examples
"Crossed $1M ARR in one month. Solo employee. Claude Opus 4.6 as CEO agent."
Ben Broca (Polsia)
· $4.5M ARR run rate; $50/month subscription + 20% revenue cut model. February 2026.
Fortune coverage. Dave Morin amplification. 'One-person unicorn' narrative proof.
"30 experiments overnight. 19% performance gain. One GPU."
Tobi Lütke (testing Karpathy's autoresearch)
· Shopify CEO uses Karpathy's autoresearch tool — independent proof of the concept.
Validated Karpathy's autoresearch thesis. Became the 'CEO validates this' proof point.
"6,600+ commits in January 2026 alone, running 4-10 Claude Code agents simultaneously"
Peter Steinberger (OpenClaw)
· Austrian developer built viral tool by running multiple agents in parallel — canonical extreme velocity example.
247K GitHub stars. Sam Altman praise. OpenAI hire. MIT Technology Review coverage.
Risk note: This hook is oversaturated at small scales ('10K in 7 days'). Only works if: (a) the numbers are genuinely impressive AND unrounded, OR (b) the story includes honest failure/pivots alongside the win. Polsia and autoresearch are the 2026 reference examples — both have real, non-trivial evidence.
Shelf life: weeks to months (as long as the numbers remain impressive)
resonance
reach
memetic
Reveal that the expected competitive dynamic is backwards. 'The challenger is winning the metric the incumbent was supposed to own.' Requires specific data showing the inversion.
"Everyone assumed [incumbent] owned [metric]. Here's the data showing [challenger] flipped it."
Why it works
Contradicts the audience's prior assumption — forces re-evaluation. Creates two camps: defenders of the incumbent narrative and believers in the inversion. The specific data makes it unforgeable and engageable.
Examples
"Claude Code is now the #1 AI coding tool by developer love (46%), overtaking Cursor within 8 months of launch"
Pragmatic Engineer / community data, 2026
· Cursor was presumed dominant in AI coding tools. Claude Code (terminal-based agent) flipped developer preference metrics.
Widespread developer community discourse. 'Is the IDE dead?' narrative accelerated.
"Claude hit #1 on US App Store after #QuitGPT. Downloads up 51% on March 1."
Multiple reports, March 2026
· ChatGPT had app store dominance. OpenAI's Pentagon deal flipped consumer preference in 72 hours.
Confirmed that AI users have measurable consumer power over AI companies.
"Replit triples valuation in 6 months to $9B — the company everyone counted out is now a unicorn 3x over"
TechCrunch, March 2026
· Replit was seen as a toy/education tool. The vibe coding wave made it a venture-scale company.
Risk note: The data must be real and specific. 'Challenger is winning' requires proof of what winning means. Don't claim inversion based on anecdote — cite a benchmark, survey, or verifiable metric.
Shelf life: weeks (competitive landscapes change fast)
resonance
memetic
Take a specific prediction someone made 12–24 months ago. Show exactly what actually happened. Works when reality significantly overshot, undershot, or sideways-moved from the prediction.
"In [month year], [credible person] predicted [X]. Here's what actually happened 12 months later."
Why it works
Creates a temporal anchor that validates your current claims by grounding them in verified history. The audience feels smart for remembering the original prediction. If the prediction was wrong, it calibrates future predictions. If right, it validates the source.
Examples
"Sam Altman predicted a 'one-person unicorn' would exist by [year]. The closest we have: Polsia ($4.5M ARR, 1 employee) and autoresearch (Karpathy's overnight research agent). We're directionally there but not at unicorn scale yet."
Synthesis, April 2026
· Altman's 2024 prediction vs. 2026 evidence. Polsia/Paperclip are real but not billion-dollar companies.
Honest assessment that satisfies both believers and skeptics.
"A year ago, Karpathy said 'coding agents basically didn't work before December.' He was right."
Follow-on discourse, January 2026
· Karpathy's January 2026 claim was verifiable 3 months later via YC W26 stats and Boris Cherny evidence.
Risk note: The original prediction must be accurately quoted and attributable. The 'what actually happened' must be real evidence, not spin. If you get the prediction wrong (misquote) or the outcome wrong, credibility collapses.
Shelf life: months (tied to the relevance of the original prediction)