Thanks! Nice findings.

Is the output, that others can use, the research and findings, or is there a tool or something that came out of this that I can plug in to one of the common harnesses?

I guess that brief note in the twit is that it'll be part of that harness you're opening up in the future?

Thanks! I've seen many developers use the entire tweet as a prompt to improve their harness. If you trace errors, you can find tool call failures per billion tokens. I've seen like many versions of this on pi, oc, or whatnot. This is already live in our hardness command code for almost two months now and we are now fixing one million tool repairs per one trillion tokens.

also decided to open source Command Code by v1 release, so hopefully end of the month. https://www.youtube.com/watch?v=laEzOCgtK6c (live stream on this)

here's a lil pi extension I made yesterday based on it https://gist.github.com/maxjustus/3dd4c83865b1c439e26fdd925b...

see that's what i was saying.

in v1 we'll have a super strong extension api — week out probably.

> he trigger was watching deepseek-flash fail on the simplest /review run, every shellCommand and readFile call bouncing back with a raw zod issues blob, the model unable to recover because the error wasn't in a form it could read. by the end deepseek v4 pro was beating opus 4.7 6/10 times on our internal evals.

I think this is why Xiaomi forked OpenCode and created their own agent hardness, to reduce friction between model and harness:

https://github.com/XiaomiMiMo/MiMo-Code

probably yep, i doubt they'll be able to maintain that well. we did a lot of work improving MiMo's cache thrashing, it was pretty bad early on. Now we're seeing 97% cache rate on their models after lots of prefix work.