Init – a Rust agent harness that recompiles and hot-swaps its own core mid-task
Posted by nagi_builds 2 days ago
Comments
Comment by blixt 2 days ago
I built something for this and also found that the idea of a "guardian" core that can detect misbehavior (crashing etc) and roll back a messed up state really helped. But it's a lot of work to get a harness to behave well, and each model reacts differently to the harness limitations, so I found more success in building a meta-harness instead (just run existing harnesses inside a common UI and inject all capabilities).
Comment by nsagent 2 days ago
[1]: https://maki.sh/#lua
Comment by jauntywundrkind 2 days ago
Comment by tmaly 2 days ago
Comment by ComputerGuru 2 days ago
Comment by whereistejas 2 days ago
Comment by athrowaway3z 2 days ago
Do i have it right that its some prompts injection for reflection and a composition scheme for the prompt/context where each element is version controlled?
If so, my experience with all the SOTA models is that they're pretty bad at finding the 'clean' solution to self improvements; eg they still always add text instead of taking it away, and somebody needs to remember to tell them (eg AGENTS.md) to delete stuff.
I suppose there is a real demand for a harness that does get better over time for average users, i'm just a bit skeptical if that actually works out over a longer time.
Might check it out later.
Comment by jonstewart 2 days ago
This is the sort of nonsense up with which I shall not put!
Comment by osbull 2 days ago