> Curator / blog
memory, resume-diff, workflow

Resume with what you know now

You're not the developer you were a year ago. Your old projects shouldn't pretend you are.

Open a project you haven’t touched in a year. Your AI assistant loads it up and cheerfully helps you… continue exactly where you left off. Same patterns. Same assumptions. Same shortcuts you’ve spent the last twelve months learning to avoid.

That’s the default, and it’s backwards. The most valuable thing you bring back to an old project isn’t the old context — it’s everything you’ve learned since.

Curator’s core move is the resume diff. When you return to a project, instead of “here are some old notes,” you get the delta between the engineer who wrote it and the one who came back:

> mem_resume_diff("code/legacy-api")

Last touched: 11 months ago.
Since then, your defaults shifted on auth, background jobs,
migrations, import boundaries, and test strategy.

This project still uses 4 patterns you now reject.
Apply current-you context before planning edits.

It isn’t recall. It’s migration history for your judgment — and it runs before your agent proposes a single edit, so the plan starts from current-you, not the version of you that’s a year stale.

Local. Encrypted. Yours. Exposed over MCP, so any agent you already use can pull it in.

Stop restarting from old context. Resume with what you know now.

Curator is coming. Get early access.

Was this useful?