dat999zx
Five failure modes around fact retirement in agent memory, with measurements
It's Trump's world, and he's accelerating its chaos
Democrats try to ride data center backlash to election victory rural US Midwest
Show HN: Chronocity – Watch your Git repo's history grow into a 3D city
https://dat999zx.github.io/chronocity/
This is timelapse of a 3D city simulation based on your Git history, files are buildings and height is based on LOC. Weather can determines if the repo is quiet or active during a span of time.Just a fun visual to get the momentum overview of your project. I've been working a lot in 1 repo for the past few months and got a handsome amount of commits, so I decided to make something visually...
Go advertise somewhere else, HN is not the place for you.
About the sync about 2 different sessions or even agents/subagents, Knowl notify to the query card what changed since the last query. Eg: 'KNOWL CHANGED: 3 items since you last looked' so they know to re-read and decide.
Open-source version at: https://github.com/dat999zx/knowl
The number is FactConsolidation single-hop at 262K context on MemoryAgentBench. On multi-hop Knowl scored 0.07 against an all-time ceiling of 0.14.
Write-up of the approaches that failed first: https://blog.knowl.cloud/the-story-of-knowl
Show HN: Knowl – agent memory with write-time supersession, 0.90 on MAB
To me when I code by hand, I never use {} after an if statement if it's one-line, it's just faster, look cleaner to me.
npm install -g @dat999zx/knowl
knowl init
That's it, fully working now.
Also all your atoms are stored inside of knowl.db file which is at root of your project after initializing. You can use knowl view command to watch the atoms and how they interact with each other.
Show HN: Knowl – CLAUDE.md hit 1000 lines, so I built memory that prunes itself
https://github.com/dat999zx/knowl
Using AI agents is amazing, but in long-term, I started to have more and more problems about the context of them. They keep forgetting what we are working on, opening a new chat session clears all their context, my CLAUDE.md had like 1000 lines.So I thought to install those agent memory that are on the market. Started out great, but I noticed that those memory only append and does not fix what...