Harmix/open-stanley / open-stanley/skills/study
study
VerifiedLearn how a specific writer wins on LinkedIn or X — Paul Graham, Dharmesh Shah, Yamini Rangan, Yuriy Zaremba, or anyone the user names — by reading their recent posts through the user's browser and extracting reusable *shapes* (openers, structures, endings, specificity moves, cadence) into a shapes catalog, never their sentences. Use when the user says "write like Paul Graham", "learn from X's posts", "why do their posts work", "study my competitor", or when write needs a shape for a lane. Ships with a catalog of the creators studied in September 2026.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.8 KB |
| references/creators-2026-09.md | markdown | 5.0 KB |
| references/creators-latest.md | markdown | 7.3 KB |
| references/shape-vocabulary.md | markdown | 2.3 KB |
references/creators-2026-09.md
View on GitHub# Creator shapes, September 2026
Derived from live reads of each feed on 8 Sep 2026 (see the parent repo's `docs/research/live-feeds-2026-09-08.md`) and, for Paul Graham, from his X feed as read by the user and the paid product in Aug 2026. Shapes, not sentences. Quoted fragments are pointers under 15 words.
## Paul Graham — X (@paulg)
- `opener:claim` / `body:single-observation` / `ending:flat`. One thought per post, stated as if said at dinner, in plain words. No setup, no list, no conclusion, no "why this matters".
- Sentences are ordinary length and uneven; he will write one 30-word sentence and one 6-word sentence. Never three short punches in a row.
- Vocabulary is simple on purpose; technical ideas get a household analogy rather than a term.
- Hedges are honest, not defensive ("I suspect", "I'm not sure, but"), and rare.
- `never:` hashtags, emoji, questions to the audience, bold, threads for single ideas.
- For comments/replies in this shape: one observation, ideally one the original poster hadn't considered, in 1–3 sentences. The user's own accepted comments matched this ("how do you know which project context is still relevant? most projects are dynamic…").
## Dharmesh Shah — LinkedIn (1.19M) and X (~500K)
- Best-performing shape (626 reactions, 90 words): `opener:claim` on a naming/labels idea → `body:single-observation` with a then-now step → `ending:flat`. Short beats long for him.
- Second: `opener:lost-argument` → `body:argument-lost` (states the counter-argument fairly, says he loves A/B tests, then gives the intuition) → `ending:intuition` ("over the fullness of time, lower friction always wins"). 145 comments: a fair argument invites disagreement.
- Product shape: `body:product-episode` ("How I Grow Better With YouSpot: Episode N"), always with a concrete use from his own life, the price, "cancel anytime", and "let me know what you think". Runs as a series so each post compounds.
- Signature moves: "Woo hoo!" for milestones; "BREAKING NEWS:" used for a joke; `ending:ps` where the p.s. carries the real point; parenthetical asides; **an explicit "no AI was used to write this" p.s.** on personal posts.
- Cadence: ~2 LinkedIn/week, X daily-ish; X gets the rawer, later-night version; the 1.35M-view X post was a plain "Woo hoo! Thrilled to announce…".
- `never:` hashtags on personal posts (only on product posts), engagement-bait questions, one-line paragraphs.
## Yamini Rangan — LinkedIn
- Fixed shape: `opener:past-self` ("Early on in my career…", "I started my career in sales…") → `body:then-now` (what she saw then, what she sees teams doing now) → what it means → "that's what we're focused on at HubSpot" → `ending:flat` values line.
- 150–300 words, text-only, paragraphs of 2–3 sentences, one number per post (62% more deals), no hashtags, no question.
- Her historically strongest posts are admissions (being wrong, a bad quarter). `specific:own-mistake` outperforms product news by an order of magnitude.
## Yuriy Zaremba — LinkedIn (AiSDR)
- Near-daily, every post with a screenshot or photo (`media:screenshot`), short lines, `specific:exact-number` unrounded, `body:steps` with a result ("Three steps: 1… 2… 3… booked a qualified meeting").
- `opener:event` from his own inbox or day ("Someone tried to hack my GSuite today", "Got a new profile picture today") and `ending:open-question` or `ending:teaser` ("more details 09.09") that runs across posts.
- Names his failures ("3 biggest screw-ups") and publicly documents what tools do wrong (the AI-poster post: repeated stories, fake stories, feedback ignored). Comments run high relative to reactions.
- `never:` hashtags, generic advice without his own data.
## Vitalii Dodonov — LinkedIn and X (Stan CTO)
- `opener:number` with real business figures ("I made $3.42M in August 2026", "20,000 sign-ups a month with 11.4% free-to-paid, ~24% churn") → `body:numbers-breakdown` or `body:story-lesson` → one operating lesson → `ending:cta-price`/CTA to a concrete thing (Launchpad, Stanley) or `ending:open-question` ("what has to change before someone says $1,000 is obvious?").
- A quoted customer sentence as the pivot ("Not in its current state.").
- 100–250 words, plain paragraphs, one contrarian operating line per post ("Indecision is the real enemy. Not bad decisions.").
- On X: the revenue breakdown got 10x his usual views; everything else is one operating thought per post, `body:single-observation`.
## Cross-creator constants (what `write` should treat as defaults)
1. The claim is something only the author could know: a number from their dashboard, a mistake, a lost argument, an email they got.
2. Openers are plain ("One of…", "Ran a…", "Someone…"); the best-performing posts are the least formatted.
3. Endings don't summarize. They land on an intuition, a flat line, a price, or one real question.
4. Series compound (Episode N, 09.09). One-off insight posts don't.
5. The biggest AI builders now signal human authorship explicitly. Whatever the plugin writes must survive that context.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |