An newsletter aobut AI-built businesses and running a portfolio.

📩 Subscribe now and get my free guide: 101 Prompt Engineering Templates

Oct 16 • 1 min read

Claude Code kept forgetting what I was working on.


Claude Code kept forgetting what I was working on.

I would set something up on the desktop, move to the laptop, and it would have no idea what we had just decided.

I assumed I had one memory that was failing. I had five.

↳ Five separate stores across my projects and machines, none of them talking to each other
↳ One machine sitting eight commits behind the other, so half of what I told Claude never left the room
↳ A sync hook I had already written, which turned out not to be the thing that was broken

The fix was boring, which is usually the sign it is the right one.

One store. One file per fact. A single index that loads into every session, with the individual files pulled in only when they are relevant. 273 files, 1.2MB, all plain markdown I can open in Finder.

Then a hook that commits, pulls and pushes after every turn, so the other machine picks it up next time I start a session. Nothing manual.

There is a memory plugin in the marketplace. Claude told me not to use it, because it recreates the exact problem I had just fixed. I uninstalled it.

The part I keep relearning: when AI forgets something, the interesting question is not how to make it remember. It is how many places you were storing the answer.

Watch it on YouTube

Grammarly is my grammar checker of choice.
Lovable builds a working app from a prompt.
Both affiliate links.

Write on,
Bryan Collins

Landscape House , Baldonnell, Dublin 22 D22 P3K7
Unsubscribe · Preferences


📩 Subscribe now and get my free guide: 101 Prompt Engineering Templates


Read next ...