I built a mobile app that puts 100+ AI models in your pocket — images, videos, audio, music & chat

I tried building a memory-first AI… and ended up discovering smaller models can beat larger ones

Dataset Model Acc F1 Δ vs Log Δ vs Static Avg Params Peak Params Steps Infer ms Size Banking77-20 Logistic TF-IDF 92.37% 0.9230 +0.00pp +...

[D] Howcome Muon is only being used for Transformers?

Muon has quickly been adopted in LLM training, yet we don't see it being talked about in other contexts. Searches for Muon on ConvNets tu...

[P] Run Karpathy's Autoresearch for $0.44 instead of $24 — Open-source parallel evolution pipeline on SageMaker Spot

TL;DR: I built an open-source pipeline that runs Karpathy's autoresearch on SageMaker Spot instances — 25 autonomous ML experiments for $...

Improving AI models’ ability to explain their predictions