Who Is Boris Cherny, the Creator of Claude Code?

Boris Cherny is the software engineer at Anthropic who created and heads Claude Code, the command-line agent tool for software development.

At Layer3Labs, we build AI automations and developer workflows for businesses, so we follow how the people who build AI coding tools use them in their own work.

Before joining Anthropic in 2024, Cherny spent about five years at Meta, working on Facebook Groups and Instagram. He also wrote the O'Reilly book Programming TypeScript.

Boris Cherny in Brief

Boris Cherny heads the Claude Code team at Anthropic, according to the Computer History Museum page for his September 2026 CHM Live appearance. On his personal website, Cherny describes himself as a software engineer at Anthropic, where he created Claude Code, and says he lives in San Francisco.

Interviews across 2026 covered his parallel terminal setup and how much code Claude Code now writes. In a February 2026 discussion with Lenny's Newsletter, interviewers reported that Claude Code authored 4% of public GitHub commits.

As of October 2026, no named publication has published Cherny's net worth or compensation package.

• Current role: Creator and head of Claude Code at Anthropic.
• Location: San Francisco, California, per his personal blog.
• Known for: Creating Claude Code, authoring Programming TypeScript, and releasing open-source developer libraries.
• Previous experience: About five years at Meta (Facebook Groups and Instagram); earlier work in finance.
• In the news for: Publishing his parallel terminal workflow, discussing agent team management at industry conferences, and reporting high commit volume.

Origins of Claude Code at Anthropic

Claude Code began as a side project and grew into a core internal tool at Anthropic, according to The Pragmatic Engineer. Claude Code runs inside the developer's terminal.

Anthropic announced Claude Code as a research preview alongside Claude 3.7 Sonnet on February 24, 2025, per the official Anthropic announcement. The announcement described an active collaborator that searches code, writes tests, and pushes commits to GitHub.

The architecture avoids complex database indexes. Cherny's team found that simple glob and grep searches, driven directly by the model, outperformed vector databases and retrieval-augmented generation. This design allows the assistant to inspect codebases without pre-computed embeddings. Our guide to Claude Code for business deployments covers what that design means for a company repository.

Terminal Workflow and Parallel Execution

Cherny runs five Claude sessions in parallel in his terminal. On January 2, 2026, Cherny published his personal setup on X, which VentureBeat reported on January 4, 2026. He numbers the tabs 1 to 5 and uses system notifications to keep track of them.

In that thread he said he used Opus 4.5 with thinking for everything, calling it the best coding model he had ever used. His team logs every assistant mistake in a repository file called CLAUDE.md. This file instructs future sessions to avoid repeating identified errors. Cherny also uses slash shortcuts like /commit-push-pr, dedicated subagents, and a Chrome extension to check visual interface changes.

The Pragmatic Engineer podcast episode with Cherny, which aired March 4, 2026, reported that this parallel setup yields about 20 to 30 pull requests a day. His process starts in plan mode and moves to implementation once the plan is good: "once there is a good plan, it will one-shot the implementation almost every time." By March 2026, Cherny told Station F that he had not written code by hand since November.

Boris Cherny's Career at Meta and in Finance

Cherny worked in finance before Meta, according to his AI Engineer speaker bio. That bio describes him as a technical lead for Facebook Groups and Instagram at Meta. The Pragmatic Engineer gives his title there as Principal Engineer and his time at Meta as five years.

His personal website explains that he prototyped and scaled product and infrastructure ideas during his tenure at Instagram. His open-source projects include json-schema-to-typescript and flow-to-typescript, per the same AI Engineer bio.

In May 2019, O'Reilly published his 322-page book, Programming TypeScript: Making Your JavaScript Applications Scale, as indexed by I-Programmer.

Cherny briefly left Anthropic to join Cursor and returned after about two weeks, according to his February 19, 2026 interview with Lenny's Newsletter.

For how Claude Code compares with Cursor and other coding agents, see our Claude Code alternatives comparison and OpenClaw vs Claude Code.

Perspectives on Agent Systems and AI Safety

At a Computer History Museum conversation on September 23, 2026, Cherny remarked that building with modern models feels more empirical, comparing it to a social science.

At Fortune Brainstorm Tech in June 2026, Fortune quoted Cherny discussing the shifting scope of technical labor. Cherny said that on the morning of that talk he was managing "maybe a few hundred" agents, and that some days it is thousands or tens of thousands. He added: "I can do much more, and the role of a builder is just totally changing."

In a follow-up interview on June 10, 2026, Fortune published the three traits Cherny said Anthropic looks for in hires: generalists, empiricists, and low-ego candidates. He told Fortune, "Ego just gets in the way of stuff." In March 2026 he told Station F, "For us, the model is the product and the product is the model." Our profile of Chris Olah covers another Anthropic figure.

The Computer History Museum Conversation

Cherny's CHM Live conversation at the Computer History Museum on September 23, 2026, was held with curator David C. Brock. On safety, he said there: "It's no accident that everyone that works closely to model research takes safety very seriously." He also predicted: "Everyone will be able to write code as well as I can."

Internal Development and Rollout at Anthropic

The Claude Code team released an internal version for Anthropic staff in November 2024, according to The Pragmatic Engineer. In that September 23, 2025 deepdive, Cherny said his first terminal prototype was hooked up to AppleScript and could tell him what music he was listening to. Once he gave it filesystem and bash tools, the prototype became useful for exploring a codebase.

The Pragmatic Engineer reported that about 20% of Anthropic's engineering team used the dogfooding release on day one and 50% by day five. By September 2025, The Pragmatic Engineer reported that more than 80% of Anthropic engineers who write code used Claude Code day to day, including data scientists running queries. The same deepdive reported a 67% increase in Anthropic's pull-request throughput while its engineering team doubled in size, and credited the gain to Claude Code.

Anthropic debated keeping Claude Code internal as a competitive advantage. Anthropic chose to release Claude Code because building tools people use teaches the company about model safety and capabilities. In his March 11, 2026 talk at Station F, Cherny added that Anthropic debated selling Claude Code top-down to enterprises before adoption spread bottom-up from individual developers.

Claude Code is built in TypeScript with React, Ink, Yoga, and Bun, a stack the team chose to be "on distribution" for the model. Cherny told The Pragmatic Engineer that about 90% of Claude Code is written with Claude Code. Cherny said the team keeps Claude Code a thin layer over the model and deletes code with each model release. With the Claude 4.0 models, the team deleted about half the system prompt. Claude Code runs locally with no virtualization, which Cherny attributed to always picking the simplest design.

Boris Cherny's Workflow and Team Costs

Running several Claude Code sessions in parallel, as he does, multiplies usage, and our Claude Code pricing guide sets out what each plan costs.

The Pragmatic Engineer also reported that working prototypes replaced product requirement documents (PRDs) on the Claude Code team.

For another builder of AI coding tools, see our profile of Peter Steinberger, and the wider list of AI people to watch.