Tobi Lütke ran Karpathy's autoresearch loop against the Liquid templating engine he created 20 years ago, producing 93 commits from 120 experiments that cut parse+render time by 53% and allocations by 61%.
METR found maintainers would reject roughly half of AI PRs that pass SWE-bench automated grading, with a 24-point gap that suggests benchmark scores substantially overstate production readiness.
Italian-Legal-BERT is a 110M-parameter domain-adapted BERT model for Italian legal NLP, trained on 3.7GB of court decisions from Italy's National Jurisprudential Archive.
Researchers at Scuola Superiore Sant'Anna in Pisa built Italian-Legal-BERT, a 110M-parameter model trained on 3.7GB of Italian court decisions that outperforms general Italian BERT on legal NLP tasks.
NVIDIA releases Nemotron 3 Super, a 120B-parameter open model with only 12B active at inference, combining Mamba-2 and Transformer layers for agentic AI workloads with a 1M token context window.
Real costs of self-hosting Llama, Mistral, Qwen, and DeepSeek models on cloud GPUs vs. API access - with break-even analysis and hardware pricing.
