Hacker News Reader

Comment by whazor | Hacker News Reader

Both Codex and Claude have trained their models to their harnesses.

There are some experimentations by Igor Warzocha to extract Codex shapes and put it in Pi: https://github.com/IgorWarzocha/howaboua-pi-stuff/tree/main/...

I'm expecting every model will have a fine tuned Pi extension at some point.

View on HN

I’ve been using both families of models inside pi/piclaw (https://rcarmo.github.io/projects/piclaw/) and I assure you they work _better_ in that environment than in the originals. The models are not trained to the harnesses, the harnesses provide cues that the models follow.

You can see the approach in his post: https://howaboua.dev/writing/how-i-gave-pi-17-tools-without-...

The result: 38% fewer startup tokens, 17 tools exposed through just three schemas, and 19 skills loaded only when needed.

The changing header is really distracting when trying to read your blog. Then it followed me when I scrolled and I just gave up.

Not my blog :)

It's pretty much what I do as well, piclaw has had selective tool activation since... March, I think.

in my experience pi does much better than codex for compaction and token use. And this is the main reason I switched to it.

migrating to the same compaction and exact tools as codex uses will make it at the same level as codex so what benefit will it have over codex? sure you can customize tui to your liking and add something on top, but the efficiency gains will be gone

Reply on news.ycombinator.com