Tag
A new optimization technique for open-source RL training engines introduces prompt caching during training, achieving up to 7.5x speedup on long-prompt, short-response workloads by reducing redundant compute.
Anthropic has published the Claude Cookbook, a curated collection of 81 practical developer guides spanning AI agents, RAG, evaluations, multimodal apps, and production workflows. The resource offers actionable code examples and best practices for building and deploying applications with Claude.
Former Anthropic scientist Shunyu Yao revealed details on the R&D of Claude 3.7 in a podcast, along with Anthropic's strategic shift to heavily bet on coding capabilities, and compared the differences in decision-making structures between Anthropic and OpenAI.
The author reflects on a visit to China's AI labs, comparing cultural differences between Chinese and American labs in building LLMs. Chinese labs benefit from a culture of collective work and student involvement, while American labs face challenges from individual ego and career ambitions.