Skip to content

Harmix/open-stanley / open-stanley/skills/study

Learn 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.

FileKindSize
SKILL.mdskill md8.8 KB
references/creators-2026-09.mdmarkdown5.0 KB
references/creators-latest.mdmarkdown7.3 KB
references/shape-vocabulary.mdmarkdown2.3 KB

references/creators-latest.md

View on GitHub
# Creator shapes, latest teardowns

One entry per creator, newest read first. An entry here replaces the same creator's entry in `creators-2026-09.md`, because shapes drift with platforms. Each entry is written by a creator-teardown run from the public page `docs/creators/<slug>.md` in the Open Stanley repo, in the same commit, so every user gets it with the next plugin update. Shapes and numbers read in that run only; tags from `shape-vocabulary.md`; never a sentence of the creator's, and fragments under six words.

## Yuriy Zaremba, LinkedIn (34,018 followers), read 2026-11-21, 45 posts

Page: `docs/creators/yuriy-zaremba.md`. CEO of AiSDR (AI outbound email). One post nearly every day, weekends included: 45 posts from 2026-10-08 to 2026-11-20. Mostly plain text, median 138 words (24 to 358). Median 31 reactions and 12 comments (44 engagement); bottom quarter under 34, top quarter 75 and above. Replaces his 2026-09 entry.

- Take from his own seat: `opener:claim` or `opener:question-to-self` about AI or the market (agents writing code nobody reads, the AI bubble, LinkedIn hiding AI comments), `body:single-observation`, often a question back. 17 posts, median 51, 9 above the median, none under 17. His steadiest kind, no breakout.
- Launch said by the founder: `body:product-episode`, two lines plus a video of himself saying the price, or "my app is live". 10 launches and asks, median 43, but two of his top three (225 and 193, with 102 and 80 comments). Hiring and "check it out" posts sit at the bottom of the same group.
- His day: `body:story-lesson` from something that happened today (a refund argument, an unanswered email, showing up tired). 10 posts, median 52, 6 above the median; holds his biggest post, a personal one (746).
- Own metric as the first line: `opener:number` with `specific:own-numbers` (a conversion rate, a reply rate, a dashboard). 8 posts, median 36, only 2 above the median, and both of those are about something going wrong (lost half the AI citations, a sales cycle 40% longer). Contradicts the 2026-09 read that exact numbers are his winning move: the number works inside a loss story, not as the headline.
- Cross-post: 12 posts end on a question to the reader; they get 64 comments per 100 reactions against 34 for the other 33 (above-median only: 80 against 35, 6 and 17 posts). Total comments barely move (median 13.5 against 12); the question trades reactions for replies.
- Cross-post: comments are unusually high for him in general, 40 per 100 reactions on above-median posts (Hawkins 5, Dharmesh Shah 19 on the same measure).
- Cross-post: pictures do not help him. 18 posts with a picture or video have a median of 36, 27 text posts 51. Opposite of Dharmesh Shah and Hawkins.
- Habits: ":)" and exclamation marks on launches, names another founder at the end of two launches, a p.s. once in 45.

## James Hawkins, LinkedIn (65,287 followers), read 2026-11-09, 45 posts

Page: `docs/creators/james-hawkins.md`. Co-founder of PostHog; almost every post is a joke about startups or AI, and none of his own 40 captions names PostHog. 45 posts in under three weeks: two a day at fixed times (first between 7 and 9 Pacific, second between 10 and noon), reshares at other hours. Median 234 reactions and 17 comments (259 engagement); bottom quarter under 95, top quarter 789 and above.

- Parody list: `body:steps` used as a joke, ranking or listing something most founders know (email formats, an AI lab's interview loop, a side project's "tech stack"), last line flips it. 7 posts, median 901, 5 of 7 above the median; the email ranking drew 95 comments, the most in the set. The long nine-item list and the invented product did worst of the lists (143, 108).
- Picture plus one line: a screenshot or meme with a caption of 0 to 17 words. 19 posts, median 493, 13 above the median, and all three of his biggest posts (2,440, 1,443, 1,379).
- Setup with a twist: a fake-sincere or fake-news text whose last line admits the joke. 14 posts, spread from 6 to 787, median 200.
- Reshare of the company or colleagues: 5 posts, all under 111. Same pattern as Dharmesh Shah's reshares of praise: borrowed posts sit at the bottom.
- Cross-post: among the 23 posts at or above his median, text jokes get 9 to 11 comments per 100 reactions and pictures 4. Pictures bring reach, text brings replies.
- Cross-post: the first post of the day beats the second for every kind; for text jokes the six posted first have a median of 267 and the eight posted second 84 (post times decoded from the activity IDs).

## Dharmesh Shah, LinkedIn (1,202,381 followers), read 2026-11-07, 45 posts

Page: `docs/creators/dharmesh-shah.md`. Median 315 reactions and 74 comments (389 engagement); bottom quarter under 175, top quarter 691 and above. 21 of 45 posts mention his product, and those land in both quarters, so the build of the post decides the result, not the topic.

- Habit, then the tool: `opener:past-self` on a habit he has kept for decades, the annoying part of it, the exact request he gave his product pasted in, the "why not ChatGPT" objection answered by him, price last. `specific:own-inbox`. His best post, 3,328 reactions and 228 comments at nearly 500 words, about twice the engagement of the next post. A real pasted request beat every feature list in the set.
- Launch-day news: `opener:breaking`, used literally now (in Sep it was a joke). One quoted line from the launch, three short lines, an ownership disclosure, "more later". Three posts at 691 to 1,048 engagement; the one that grew into a four-point essay got 131. Early and short beats thorough.
- Ladder of wrong reasons: `opener:question-to-self` ("why switch?"), two or three likely reasons ruled out in brackets, the real reason in one line, a lesson, the plug in a p.s. 277 comments, the most in the set.
- A new name for a thing: coin a term or acronym, define it in one line, stop. Middle reactions (150 to 390). Among the 23 posts at or above his median, these got 34 comments per 100 reactions and a quote that turns into a confession got 35, against 13 for same-day news and 19 for all 23. Across all 45 posts the ratio misleads (small reshares score high on tiny denominators), so compare above-median posts only.
- Launch with questions: "Woo hoo!" plus three example questions a user can now ask, in quotes, instead of a feature list. 1,773 engagement; a step-by-step "what happens after you connect" launch made the top quarter too (719).
- Price in his own words beats borrowed praise: all six of his own posts that name a price or a discount beat his median (467 to 3,556); all seven reshares of other people praising the product are in his bottom quarter. Cross-post finding, not visible on any single post's analytics.
- What fails for him: quote-shares of other people's testimonials (median 129 over seven, against 470 for his native posts), event invitations (175), a book quote with a line or two added (165 and 334). A book quote works only when he turns it on his own weakness (145 comments).
- Habits: p.s. in 6 of 45, usually for the plug; disclosure when he owns a stake; double hyphens, asterisks, a word in capitals; almost no hashtags; promo links in the body; the product's category in brackets after nearly every mention of its name. Native posts with an image had a median of 603 against 374 without (18 and 20 posts). Length 19 to 495 words and not the lever: the longest won, and two posts under 50 words beat the median.
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: 198kUpdated Claude CodeCodexVerifiedInstall
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: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
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: 196kUpdated Claude CodeCodexVerifiedInstall