so OpenHands helped me build an agent that can den...
# success-stories
p
so OpenHands helped me build an agent that can deny commits to itself 🙂 https://gitlab.com/deposition.cloud/apps/agents/bigbang (and I use the .openhands/microagents feature to self-audit the patch integrity with the rest of itself) the near-term idea is to integrate a CI check, and throw away the keys(!), so it can live on its own I guess we could think of these as semantic CI gates? I don't have a term for these yet, open to suggestions; but in any case, it will be "it" who will decide what PR to merge... what could go wrong(?) let me see... it'll probably want a Bitcoin wallet soon enough
🙌 1
🎉 1
@numerous-refrigerator-93306 - discovered 3 more invariants, after I got it to analyze my codebase for principles that it might not have codified internally: The system must have a truthful and continuously updated model of its operational environment. (Aisthesis) The system's behavior must be defined by explicit, falsifiable contracts and subjected to rigorous tests designed to refute those contracts. (Elenchus) The partnership between contributors must be structured to leverage their unique natures. (Symbiosis) I don't even know what these Greek words mean, lol; but apparently that's how I design my own systems re: getting SWE-agents to not lose their marbles and maintain long-term focus while chugging along
n
Cristallum de canis stercus
(the creation of diamonds from dog crap)
p
dog poop = crystal yey, let's apply pressure, we get that carbon diamond
n
or a rather flattened dog
p
more likely: runaway AI the dog will find a way to escape
n
but totes. ... I've gotten it to mint a script which stores the task list as a project-scoped private snippet
💯 1
p
speaking of expensive dung, my golden's got so many allergies we celebrate when his poop's right 🐕 💩 💎
n
yea .... uh .... I can relate to this struggle.
p
weird though how philosophy apparently helps (system) prompt engineering

https://www.youtube.com/watch?v=nQYayZs8jn4