Beacon open-sources a cross-harness self-improving memory layer for 20+ coding agents
solyarisoftware · x · 2026-10-22
Asymptote Labs open-sourced Beacon (2.5k GitHub stars), a cross-harness self-improving memory layer that lets Claude Code, Codex, Cursor, OpenCode, Hermes and 20+ more agents share lessons.
Core insight: preserving an agent run and learning from it are two different problems. Beacon continuously captures sessions across all harnesses into a shared index. The author demos a live cross-agent memory workflow with Jev (a local decision engine): Jev evaluates whether each trace contains a reusable engineering lesson, and approved lessons become available to the next agent.
Companion article "Build your own Jev (100% local)" covers turning an open-source LLM into a fast local decision engine without retraining: next-token scoring, fixed choices with probability distributions, and SGLang.
Related event: Beacon Raises $5.5M Seed for Open-Source Cross-Tool Agent Memory(4 posts)→
More from coding & agent
- One week of vibe coding, 30,708 players in month one: Dust 2 DM game's Spawn stats — TAbrodi · 2026-10-22
- Security patch ships with 1,858× speedup: 3,000-scene queries drop from 2s to 1.13ms — DanielLockyer · 2026-10-22
- Codiff v1.18 ships faster startup, prettier markdown diffs and a new app icon — cnakazawa · 2026-10-22
- SGLang-TorchTPU launches with Google and Meta, hitting 770 tok/s decode on TPU — BanghuaZ · 2026-10-22
- New paper: can first-price auctions efficiently allocate queries among strategic AI agents? — _onionesque · 2026-10-22
- diagram-design: an agent skill that turns prompts into editorial HTML+SVG diagrams, 59.6k stars — tom_doerr · 2026-10-22