Right. Sonnet 3.5, yesterday. Converted Woodpecker pipeline into Forgejo, then modified to use buildah instead of docker. That's baseline right now.

The tough questions are when one asks about what shadow shapes from a single light can be expected on faces inside a regular voxel grid. That's where it just holds the duck.

Am I reading that correctly? You're using GPT 4o-mini? Why not try the more impressive o1 model?

Not that I think any of these would be worth it for me.

I am using o1 Pro in that chat