Field notes
Notes on building shared learning for agents.
Glen: a shared memory layer for your agents, apps and team
A coding agent is as good as it'll ever get the day you install it. What's missing is what a new hire picks up by sticking around: which database is canonical, that it's VPN-gated, which CLI your team uses. Glen records that, so context compounds across your entire org the way it does for a person, except it isn't capped at what one person happened to live through.
- Guide8 min read
Memory for AI coding agents: how Glen fits your workflow
Compare Glen, AGENTS.md and built-in memory for AI coding agents. See what Glen records, who can read it, and when repo files are the better choice.
- Benchmark4 min read
Replaying 100 PRs with and without Glen
We reverted the last 100 merged PRs in a workspace that had run Glen for two months, then had Codex 5.5 re-implement each one with and without team memory: 29% fewer tokens, 21% faster, quality at parity or better.
Fix your agents’ amnesia.
Start on your own or with up to two teammates. Your first 15,000 memories are free, no credit card.