Articles by tag: rlhf

AI Progress in Code and Mathematics: Karpathy's Theses

Breakdown of why AI surged in programming and RL but lags in texts. For developers: how to use Codex and Claude Code. Learn about the gap in model perception.

DeepSeek creates an AI agent better than Claude Code

DeepSeek is hiring 17 engineers for agentic AI, competitor to Claude Code and Cursor. Job details, RL techniques, DeepSeek V4 on Huawei chips. For IT specialists: study China's AI trends.