'The Karpathy Loop': 700 experiments, 2 days, and a glimpse of where AI is heading

kept by eddie mar 24

Andrej Karpathy's 'autoresearch' AI agent ran 700 experiments in 2 days, discovering 20 optimizations that sped up language model training by 11%.

Autoresearch on an old research idea | Blog
Using Karpathy's Autoresearch loop with Claude Code, the author reduced mean rank by 54% on an old research project, primarily through bug fixes and hyperparameter tuning.
also on autoresearch, Andrej Karpathy, #ai-research, #ai-agents
A new software engineering paradigm – Blog
Using formal verification, AI agents can generate code and machine‑checked proofs, removing the need for human code review and overcoming the AI‑generated‑code bottleneck.
also on autoresearch, Andrej Karpathy, #ai-coding
It's Hard to Feel the AGI
Leading AI researchers argue that current LLM scaling will hit a ceiling, requiring new insights before AGI emerges, pushing timelines back 5-20 years.
also on Andrej Karpathy, #ai-research, #ai-agents
LLMs are complicated now – Ian’s Blog
Modern LLMs have grown complex with many attention variants and mixture-of-experts, echoing the messy evolution of recommendation systems.
also on Andrej Karpathy, #ai-research