It's been 2 weeks since Karpathy released autoresearch. The community is trying to apply it to everything with a measurable metric, here are some successful attempts:
- Shopify's Liquid template engine - 53% faster, 61% fewer allocations - GPU kernels - 18 to 187 TFLOPS - Voice agent prompt engineering - 0.728 to 0.969 eval score - Baseball pitch speed prediction from biomechanics data - Ancient scroll ink detection (Vesuvius Challenge)
We're curating a list of use cases and open-source implementations. PRs welcome, ideally with traces so the community can verify.
URL: https://github.com/WecoAI/awesome-autoresearch
Comments URL: https://news.ycombinator.com/item?id=47493896
Points: 2
# Comments: 0