Boris Cherny

Boris Cherny is a Ukrainian-born engineer who created Claude Code at Anthropic, the AI coding tool that crossed $1 billion in annual revenue. His career spans building Facebook Groups and Marketplace at Meta, leading Instagram infrastructure from rural Japan, writing the O’Reilly book “Programming TypeScript,” and pioneering AI-native software engineering. He dropped out of UC San Diego at 18 to start companies, and his family includes three generations of programmers going back to punch cards in the USSR.

Key Patterns

• Forced constraints (broken arms, moving to Japan, rejected PRs) consistently opened his most important career doors
• Side projects built for personal use became his biggest professional wins (Undux, Claude Code, TypeScript book)
• He leads through influence and trust-building rather than authority or directives
• Simple, repeatable systems beat complex setups every time (vanilla AI workflow, common sense)
• He watches what users actually do rather than what they say they want

AI and Engineering

Claude Code - The Side Project That Got 2 Likes — Claude Code started as a solo side project that got 2 likes internally but grew to $1B in revenue 259 PRs in 30 Days, 2 Bugs - The Vanilla Setup — Boris shipped 259 pull requests in 30 days using a vanilla AI setup with only 2 bugs Morning Agents from the Phone - Boris’s Full Daily AI Workflow — His full daily workflow: morning phone agents, 15 parallel sessions, voice dictation, and plan-execute-verify Daisy’s 20-Claude Swarm Weekend — A teammate launched 20 parallel Claude instances to build plugins over a weekend Three Modes of Working With AI — Three distinct modes for AI work: vibe coding, pairing, and by hand The Manager Who Codes Again — A manager who hadn’t coded in a decade started building again using Claude from her phone Brandon the Data Scientist Who Taught Himself — A data scientist self-taught Claude Code and sparked adoption across non-engineering teams The Rejected Pull Request — His hand-written PR was rejected at Anthropic, forcing a shift to 100% AI-written code The 80% Kill Rate on Prototypes — 80% of AI-generated prototypes never ship because taste and judgment are the real bottleneck Compound Engineering - Only Document Mistakes — The CLAUDE.md file that only documents real failures, building institutional knowledge over time Automate Your Code Reviews - The Lint Rule Spreadsheet — A spreadsheet tracking repeated code review comments that became automated lint rules

Side Projects and TypeScript Journey

Broke Both Arms, Discovered TypeScript — A motorcycle accident forced him into TypeScript, leading to an O’Reilly book and the world’s largest TS meetup Someone Should Write About This — He saw nobody was writing about TypeScript and filled the gap himself with a book and community The Undux Bicycle Campaign — Built a simpler Redux alternative and biked across Meta’s campus giving 20-40 personalized tech talks Weed Review Site - First Startup at 18 — His first startup at 18 was a weed review site; earlier he sold Pokemon cards on eBay

Leadership and Management

Never Tell Anyone What to Do — He scoped side projects for 5,000 engineers by matching opportunities to what people already wanted Delegate What You Love, Not What You Hate — Delegate the thing you know well so you can evaluate whether it’s going right Disagree, Commit, Then Reverse — He committed to a plan he disagreed with, executed it, then reversed it with evidence Beer with the Skeptics Before Pitching the Idea — He flew to New York and bought beers for the engineers most likely to resist his infrastructure plan Always Present Three Options (The VP Picks the Middle) — Always present three options to leadership; 80% of the time they pick the middle one The Design Competition That Solved an Impossible Problem — He split senior engineers into competing teams to solve an impossible architecture problem in 3 hours Unshipping Features at Instagram — Instagram’s culture of deleting features that didn’t meet a usage bar, keeping the product simple

Forced Constraints and Origin Story

Moved to Rural Japan for Love — Moving to rural Japan for his girlfriend eliminated meetings and made him top-1% productive at Instagram Three Generations of Programmers - From Punch Cards to AI — Three generations of programmers from USSR punch cards to AI-generated code The Greg Egan Sci-Fi Lunch at Anthropic — His first lunch at Anthropic confirmed he’d found his people when everyone had read the same obscure sci-fi novel The Advantage of Being Under-Leveled — Joining Meta under-leveled gave him freedom to explore instead of pressure to perform

Product Thinking

Latent Demand - How Facebook Marketplace Was Born — 40% of Facebook Groups posts were buy/sell activity, leading directly to Marketplace Guerrilla User Research in the Cafeteria — Testing new features on cafeteria workers instead of waiting for a user researcher Don’t Watch the Competition — He doesn’t use competing products and tells his team to focus on user problems instead Just Use Common Sense - Advice to His Younger Self — His advice to his younger self: just use common sense and ignore misaligned incentives