Some days I regret not building in Python

All the RL stuff is there

But I've found it's pretty easy to vibe code an adapter between the two with IPC, UDP or TCP so it's fine

Anyways, the robot can see and talk, it just can't walk yet. Butter passing soon.

Engineer @farcaster, co-founder @bountycaster. Sharing what I learn about AI + development. Previously: Phantom, 0x. NYC 🗽

Yeah, the GIL is still around 😂 

I thought there were plans to remove it. Just curious, what kind of operations made you feel the impact of the GIL? Aren’t you mostly doing API calls or Tensor operations, which shouldn’t really be affected since they’re usually handled in C and don’t hit GIL?

Let us agree to disagree

I wrote my original agent project in Python and it was pretty good for agents except async and global lock were so bad but 3.13 has greatly improved matters

I miss writing Python so much 😂 JS/TS is such a pain and such an inferior language

Barista in the morning, AI engineer by day, and parent (with @shiwen) by night ||
Host at /clickcast /cpg

For sure. There’s no way I was going anywhere near considering any Flask lol. I think there’s some added benefit to properly packaging the tool use components like an armory to make it easier to interchange the parts across different agents and evaluate

If I built it in Python I'd still have to build a whole ts stack for building frontends and mobile apps. Practically, TS is a way better go-to for app development and websites I think, and as we auto-generate frontends that is a really handy feature. But Python rules ML due to pytorch and numpy really

Question. If you did build it in Python would you have kept it all Python or had some of the interfacing in TS along with the async handling?

Your Daily Shot of Ethereum News 🗞️ 
Perma-bull of God, Ethereum, hard work, and forgiveness