Matthew Boston

I built an AI tool to write my brag document

March 29, 2025

Key takeaway: Let a tool draft your review from your GitHub history, then add the glue work in your own words.

I built Reflect so I’d stop rebuilding six months of work from memory at review time. It pulls your merged pull requests and closed issues from GitHub and has an LLM turn them into a first draft of your brag document.

Every review cycle goes the same way. You open a blank document and scroll through PRs and tickets to jog your memory. You remember the last few weeks and forget the critical fix from four months ago.

Keeping a running brag document is great advice, and I’ve tried it. The problem is discipline. Deep in a sprint, the last thing on your mind is a side document about your impact. So it stays empty until review season.

But the data already exists. Every PR and closed issue is sitting in GitHub, scattered across dozens of repos with nothing connecting them. Reflect fetches them for a date range and writes three markdown documents: a detailed contribution log, a technical summary, and an achievement summary for review season.

What it can’t see matters just as much, and I feel strongly about this. Mentoring someone through their first production incident, pushing back on a bad architecture decision in a meeting, writing the design doc that got three teams aligned: Tanya Reilly calls this glue work, and none of it shows up in a pull request. Treat Reflect’s output as a starting point, and add that part in your own voice.

It came from the same itch as git ai-commit. The raw material was already there, and an LLM only had to read it. The AI was the easy part. Most of the work went into deciding which filters, date ranges, and output formats would actually help during review prep.