Lenny Rachitsky

Edwin Chen's Surge AI passed $1 billion in revenue with under 100 employees and no outside funding, supplying frontier labs with the data, environments and evaluations that teach models what counts as good.

OpenAI's head of product design on the practice behind ChatGPT and Codex: why designers feel displaced while engineers got 10x'd, what OpenAI screens for when hiring, a "just do less" rule for scope, and the argument for ChatGPT as a super app.

Dianne Penn joined Anthropic as its first technical product manager when the product team was five engineers, and has shipped every model since Claude 2 while helping incubate Claude Code, MCP, Skills, computer use and reasoning.

Chip Huyen covers AI engineering fundamentals, why many companies struggle to ship AI products, how hard coding-agent productivity is to measure, and how to improve apps via user feedback.

Alexander Embiricos, who leads product on Codex at OpenAI, on the agent's 20x growth since August, building a software engineering teammate that spans the whole development lifecycle, and why humans are becoming the bottleneck.

OpenAI's Codex product lead on building for model capabilities two to three months out, and what changes when agents take over execution: judgment and scoped ambition become the scarce inputs rather than throughput. A look at how a frontier lab plans product against a moving capability curve.

Sherwin Wu leads engineering for OpenAI's API platform, where roughly 95% of engineers use Codex and often run 10 to 20 parallel agents. On cutting code review from 15 minutes to three, and the widening gap between power users and everyone else.

Andrew Ambrosino, who leads the Codex desktop app at OpenAI where nearly every employee now uses it weekly, on what product work looks like when the thing you are building is an agent surface.

Amol Avasare, Anthropic's head of growth, on scaling from $1 billion to over $19 billion in ARR in 14 months, cold-emailing his way into the role, and why at this point Claude is growing itself.

Jenny Wen, who leads design for Claude at Anthropic after directing design at Figma, argues the classic discovery-mock-iterate process is dead: designers no longer have time for beautiful mocks, and the role shifts to helping teams execute.

Codex lead Andrew Ambrosino explains why OpenAI is merging Codex and ChatGPT, how internal use spread across the company, and how product team roles shift as implementation gets cheap.

Mercor's Brendan Foody explains why expert-authored evals and RL environments have become the lab bottleneck, and how a labor marketplace for professionals scaled around that need.

Fiona Fung leads the Claude Code and Cowork teams at Anthropic, after 11 years at Microsoft on Visual Studio and TypeScript and a stint at Meta starting Facebook Marketplace, on what professional work looks like once coding is solved.

Fei-Fei Li on how ImageNet set off the deep learning era and why she thinks the next one runs through spatial intelligence: what world models do that language models cannot, where Marble is already being used, and why robotics remains harder than text.

Simon Willison gives an AI state of the union: the inflection point has passed, coding agents changed the daily texture of programming, and the terms he coined — prompt injection, AI slop, agentic engineering — name where things now stand.

Lenny Rachitsky interviews OpenAI's Tibo Sottiaux at DevDay about Dots, the new personal agent platform, why most internet actions will move to agents, and why workflow scaffolding such as loops and graphs is temporary.

Benedict Evans argues AI is a 1997-scale shift — real, but only as large as the internet — and works through the consequences: where value accrues across the stack, why distribution outlasts capability as a moat, and why the task-versus-job distinction matters more than automation percentages.

Boris Cherny on Claude Code going from a terminal prototype to 4% of public GitHub commits in a year, the counterintuitive product principles behind it, and what happens to the profession once coding is solved.

Cat Wu, head of product for Claude Code and Cowork, on how Anthropic's shipping cadence went from months to days, the hard problem of building for the current model rather than an imagined stronger one, and the skills PMs now need.

Dan Shipper of Every, a year after correctly predicting people were sleeping on Claude Code for non-technical work, returns with the AI paradox: more automation is producing more humans doing more work, not fewer.

Lazar Jovanovic builds internal tools and customer-facing products at Lovable entirely with AI and no coding background, and argues that lacking one can be an advantage rather than a handicap.

Claire Vo runs her business, podcast and family life on nine OpenClaw agents across several Mac Minis and old laptops, after starting as a sceptic when it deleted her family calendar on the first attempt.

LinkedIn's CPO on scrapping the associate PM track for a full-stack builder role that ships end to end, the specialized agents built to critique ideas and probe for weaknesses, why generic coding tools stalled on their enterprise code, and the change management that made it stick.

Lenny's Newsletter interviews the engineer who took xAI's Grok Bot from first commit to public launch in about seven weeks, covering the decision to build it standalone rather than inside Cursor, the early product bets that drove adoption, and personally onboarding its first 300 users.