Keel-opencore runs a bounded agent swarm for about 15 days of self-improvement
Efficient-Cap6662 · reddit · 2026-07-27
A long-horizon agent swarm with hard brakes
Keel-opencore is built for about 15 days of uninterrupted agent-swarm self-improvement, but with explicit containment controls: promote, killactive, budgetexhausted, driftregression, and maxiters.
- The project is framed as a bounded self-recursive ImproveLoop rather than an unconstrained autonomous demo.
- It is designed to stay mediated, killable, and auditable even during long-running iteration.
- The repo includes example runs such as selfimprovereal.py and adversarialdemo.py.
The core claim is that most agent demos last minutes, while this system is meant to sustain a much longer self-improvement loop without losing operator control.
More from coding & agent
- Community Buzz fork adds native Hermes Agent ACP integration and BYOH support — Teknium · 2026-07-27
- Agent loops get pricey because every call repays the whole conversation history — Future_AGI · 2026-07-27
- Claude Code experiment says Pangram's AI detector can be beaten after 17 rewrites — _akpiper · 2026-07-27
- NEO for VS Code and Cursor adds @ file tagging to streamline coding context — Nilofer_tweets · 2026-07-27
- A Stanford AI engineer launches a free 10-day AI basics series — HarperSCarroll · 2026-07-27
- Claude workflow can be deployed to run automatically every morning at 7 — TawohAwa · 2026-07-27