Lenny Rachitsky has released his entire newsletter archive, roughly 350 posts and 300 podcast transcripts, as AI-ready Markdown files at LennysData.com, along with an MCP server and a GitHub repo. Paid subscribers get the full dataset. Free subscribers get a subset. This is the raw material: years of product, growth, and career interviews, structured for immediate use with any LLM or agent pipeline.
The release follows a prior experiment where Rachitsky dropped podcast transcripts publicly and got back over 50 community-built projects, including parenting advice derived from PM frameworks, user research script generators, and an episode infographic tool built by Manus. The standout was LennyRPG, a video-game-style interface built by Ben Shih, a non-technical growth designer at Miro. The full piece walks through exactly how Shih built it, which is the real reason to read: a non-engineer shipping a functional, polished product from a cold dataset.
Rachitsky built LennysData.com itself using Codex and Claude Code, with beta feedback from Tal Raviv, Claire Vo, and Este Lopez. He is running a build challenge with a deadline of April 15th: ship something using the data, post it in the comments, win a free one-year subscription. The more interesting signal here is what the dataset enables at scale, fine-tuning, retrieval-augmented generation, agent memory. The article hints at the surface but the full piece gets into the construction details.
[READ ORIGINAL →]