Guide
Write a brag document from your work history
Everyone agrees you should keep a brag document, and almost nobody does. When review season arrives you’re trying to remember what you did in March. Here is a template, and how to rebuild one from the records you already have.
What it is
A brag document is a plain list of what you did and why it mattered, kept through the year. Julia Evans made the case for it in “Get your work recognized: write a brag document”: your manager doesn’t remember everything you did, and neither do you. Reviews reward specifics, and specifics fade within weeks.
A template
# Brag document: 2026
## Projects
- [Project]: what it was, my part, the outcome (number if any). [month, source]
## Decisions I drove
- Chose X over Y because Z; result. [month, source]
## Problems prevented or fixed
- What broke or would have, what I did, impact. [month, source]
## People
- Who I helped, mentored or unblocked, and how.
## Learned
- New skills, areas I now own.
## Smaller things that added up
- Reviews, docs, on-call, interviews.The [month, source] tag is the part most templates leave out. It makes every line checkable, by you now and by your manager later.
Why nobody keeps one
Writing it weekly takes five minutes, and five minutes on Friday afternoon loses to everything else. By November the document has three entries from January. The work still happened; the record of it is scattered across your calendar, your commits, your tickets and your calls.
Rebuild it from records
Go month by month and pull from what you have:
- Calendar: recurring meetings show which projects you were on; one-off meetings mark decisions and launches.
- Commits and pull requests: what you shipped, with dates.
git log --authorper repository is a fast start. - Tickets: what you closed, and what you filed that others fixed.
- Meeting notes and messages: where decisions were made and who made them.
- Work history, if something has been recording it. Cachejay keeps a page per project, notes from each call, and a summary of each day, week and month, and an AI app can read them over MCP. It only covers time since you started using it, so for this year it may fill the last few months; next year it can cover the whole year.
Prompts that keep the evidence
With your records available to ChatGPT or Claude (pasted in, uploaded, or through a connector), ask for evidence first and prose second:
From my work in [month], list what I shipped, decided, fixed or
helped with. For each: one line, the outcome if stated, the date,
and the source (meeting, PR, ticket, note). Don't infer outcomes
that aren't in the source. Mark anything that was a team effort.Turn these items into brag document entries using my template.
Keep each source tag. Don't merge separate items.Running it a month at a time keeps the answers grounded; a whole year in one prompt produces confident summaries of things that didn’t quite happen.
Check before you submit
- Every line has a date and a source you can open.
- Numbers come from the source, not from the AI.
- Team work says “we”, and names who else did it.
- Nothing confidential leaves where it should stay: client names, revenue, people issues. Our AI memory privacy checklist covers what to keep out.
Keep it going
Once it exists, keeping it is ten minutes a month: run the first prompt for the month that just ended, check it, paste it in. A weekly review from work history does the same at a finer grain, and the same records make standups a 30-second job.