‹ BackHN Continuity

Thread

Show HN: Lossless-memory – a personal AI memory that never summarizes

67 points · 34 comments · aru-labs

  1. firstline-dev · · focus · HN ↗
    Congrats on the launch. The "never summarizes" choice matches what I've seen from the other side: I build a tiny local-only context saver for devs with ADHD, and the failure mode of summaries is that they drop exactly the detail you need to restart — the half-finished function name, the thing you were about to check. Lossy compression is fine for recalling facts, but bad for resuming action. Resumption needs pointers, not prose. Curious: how do you handle stale memories when the codebase changes under them?
Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.