If you are looking at open models, check out Pi, as its very extensible and comes with a sane default. Maybe even roll your own.

Most harnesses (claude, codex, opencode etc.) assume that you use a cloud model. There’s no sense of optimization or finer control.

▲

theshrike79 4 days ago | parent [-]

I've been running pi.dev + codex (GPT-5.4) for a long time for my workhorse stuff.

Actually tried caveman mode yesterday and it made everything SO MUCH BETTER. GPT-5.4 has a habit of being extremely verbose to a ridiculous degree, it's like it's writing a report for a CTO or something and padding everything as much as possible to sound smart.

With caveman it just gives me lists of stuff in a compact format. Perfect.

▲

michaelcampbell 4 days ago | parent [-]

let's be clear(er) here that you like caveman's format and output, so that's one value of "better". I've seen at least a few (n < 10, so there's that) tests on token use and for THAT value of "better" it's not much different.

	▲	theshrike79 3 days ago \| parent [-]
		The token savings with caveman is completely secondary to me to be honest. What I appreciate is the terse answers it gives to me. I don't need a padded essay on what the issue is or what was fixed. Give me the bullet points.