Governed Memory for AI Agents

magnemo.ai

11 October 2026 18:23

Losing the count is a sharp test, because the work itself still looks fine. Did keeping the count outside the chat, in a file it updates each pass, hold up better than any rewording of the instruction?

Peter Ritchie

peterritchie.com

06 October 2026 14:15

The "Why" part of the user story is where most of the value comes from, it helps developers as well as AI.

Mark

denovich.com

26 September 2026 20:41

I’ve been trying to explain how to use SSH tunnels so we can do on-mobile device testing against remote servers which have egress-only firewalls. I’ve been doing it for years, but I guess it was too much unfamiliar setup for anyone else. Now /connect automates the whole thing.

Mark

denovich.com

26 September 2026 19:57

Just getting prepared to start the work took many hours setting up databases, web servers, development environments, etc. /bootstrap does it all. Same as before but my coworker’s jaw dropped when it installed Docker and spun up a database and populated it, resulting in a fully configured system.

Mark

denovich.com

26 September 2026 19:57

I’ve started doing this for other tasks we do frequently. Ex: Porting custom customer code from one version of our platform to another, which was a Herculean effort. Pointing claude at the first workbench repo for inspiration, plus relevant sources and references and it created a new workbench.