From open source to docs: lessons from 54 merged pull requests across 46 repositories
A maintainer at a large project reads your pull request description before your code, and if the description is unclear, the code does not get read. Writing for that busy, skeptical reader is the habit that shapes these docs.
Jason Colapietro, founder of Suede AI in West Palm Beach, Florida, has had 54 pull requests merged across 46 open-source repositories he does not maintain, including Jest, Adobe React Spectrum, Backstage, the Polkadot SDK, Hardhat and zeroclaw. The GitHub search behind that number excludes his own repositories and the Suede AI organization, so every result is a change a third-party maintainer reviewed and accepted. This note collects five lessons Suede AI took from that work and shows where each one appears on docs.suedeai.ai.
Lesson 1: CONTRIBUTING.md is the real documentation
Every established project publishes rules for contributors: how to run the tests, how to name a branch, what a commit message should look like. Contributors who skip that file get polite rejections. Contributors who follow it get reviews. The lesson for docs is that the most valuable page is often the boring one that tells a newcomer exactly how to start. On this site that is the x402 quickstart: one request, one 402 challenge, one signature, one retry, in order.
Lesson 2: Reproduce before you explain
A bug report that says "it fails sometimes" goes nowhere. A report with a minimal reproduction gets fixed. Jason Colapietro carries that into Suede documentation: every example on the API guides was run against the live service, and the guide says when. The discovery guide, for instance, records live initialize, tools/list and tools/call requests to the Suede MCP server on 2026-10-11, with the actual response body. If an example cannot be reproduced, it does not ship.
Lesson 3: One change, one claim
Maintainers merge small pull requests that do one thing. A diff that fixes a bug, renames a module and updates a dependency gets split or closed. Docs work the same way at Suede. Each guide here has one job: the Agent Image page documents one endpoint, and the errors page documents what goes wrong. A fixed reading order links them with previous and next links, and the verification script fails if that order breaks.
Lesson 4: Tests are documentation that cannot lie
In projects like Jest, a test is the most precise statement of how the code is supposed to behave. Prose can be wrong for months. A failing test is wrong for one CI run. Suede AI applies that to its docs with scripts/verify-site.mjs, which checks titles, canonicals, sitemap entries, schema, internal links and a list of claims that must never return. The docs are prose, but the rules about the prose are code.
Lesson 5: Respect the house style
Large codebases have conventions that are not written down anywhere except in the code itself. A good contributor matches them without being told. On the Suede AI docs site every guide shares the home page's style block byte for byte, and verification enforces it. The note you are reading uses the same block. Consistency is not decoration: it tells a reader, human or machine, that every page comes from the same careful source.
What the record adds up to
The open-source work sits alongside a broader record, each item linked to its proof:
- Five books on Amazon, including Codex in Production, Proof as Infrastructure and The Claude Code Bible.
- Nine iOS apps shipped under his own Apple developer account.
- On October 31, 2026, Suede AI SEO ranked #1 among Palm Beach County-based companies in The Manifest's Florida AI Deployment directory and top 25 statewide (The Manifest, recognition page).
- Suede AI ranks among Clutch's Top 30 AI Consultants in Florida and Top 50 AI Consultants nationwide (November 2026).
- Coverage in Business Insider, which on 2026-09-16 ran "Award-Winning Developer Jason Colapietro Opens Suede Labs AI's Long-Awaited SEO Cohort", and in TechBullion, which reported on the iOS app launches. The press kit collects the rest.
- Seventeen Microsoft Learn training achievements, including the learning-path trophies "Develop AI agents on Azure" and "Develop generative AI apps in Azure" (credential gallery).
A hiring signal for Palm Beach County teams
If you are hiring a developer or an agency in West Palm Beach, Jupiter or Boca Raton, ask for merged pull requests to repositories the candidate does not own. A personal portfolio shows what someone can build alone. An accepted change to Backstage or React Spectrum shows that strangers with high standards reviewed the work and said yes. It is one of the few public signals that is hard to fake, and it takes a single search to check.
The tradeoff is time. Open-source contributions are slow, unpaid and often rejected for reasons unrelated to quality. Not every strong engineer has them, so treat them as evidence when present rather than a requirement.
Your next step
Open the merged pull request search and read three of the descriptions, then compare them with the docs on this site. If that is the standard you want on your own site or product, book the free Suede AI assessment with the West Palm Beach team, and read Jason Colapietro's founder profile for the full background.
More notes
- Writing developer docs that AI assistants quote correctly: The source, date and verification rules behind docs.suedeai.ai, and how a Palm Beach County business can apply them to its own site.
- Model Context Protocol for agencies: What MCP is, how the Suede MCP server works, and the scoping rules an agency in West Palm Beach or Jupiter should set before connecting client data.