yes honestly its fun and the early access is better than i thought it would be

sorry your package got divided by zero and promptly disintegrated

In my experience AI has been not very good when it comes to writing burn code even for small snippets.

I just came across this thread so feel free to ignore but I wonder how it would fare with implementing a recent LLM like qwen3 using a less popular library like Burn for rust. Maybe I just dont know enough about it but it seems difficult to get good tests for something like this.

Z-Image is currently one of the top open weights models. It easily runs on 8gb vram when quantized and I've seen someone run it on 4gb vram and even an android phone. It may even be possible to run fully in the browser. So the barrier to run it is really not that high.

I think your art is AI

(actually impressive)

kagi news if you dont click on the articles

hello my falkenstein neighbours

whitehouse-47.bsky.social? very credible username, immediately creates confidence in that this account is official

even emails work, its a miracle

there is light at the end of the very long tunnel