How Boris Cherny Climbed Four Levels at Meta by Pursuing Side Projects

Don’t Just Do Your Assigned Tasks

Many engineers think completing their Jira tickets is enough, but Boris’s promotions were largely driven by what he calls “Side Quests.” Before joining Facebook, he built a weekend state‑management library called Undux because he found Redux cumbersome, then promoted it across teams with dozens of tech talks, demonstrating that solving a problem many face creates leverage.

He later spent a year writing a TypeScript book and organizing the world’s largest offline TypeScript meetup, further expanding his influence.

Spotting Latent Demand

Working on Facebook Groups, Boris identified “Latent Demand”: observe what users are already doing clumsily and help them do it better. This insight led to Facebook Marketplace (40% of group posts were buying/selling) and Facebook Dating (many users viewed non‑friend profiles).

He advises engineers to align technology with existing user behavior rather than forcing new actions.

Technical Decisions: Challenging Authority

For the Public Groups data‑model redesign, senior engineer Bob argued that members and followers should remain separate tables. The team spent six months on the migration, but Boris’s code audit showed merging them simplified the architecture.

He rolled back Bob’s plan, backed by data and investigation, eventually winning Bob’s approval and illustrating that respectful dissent can correct costly directions.

Zero‑Title Mindset and Regaining Trust

After moving to Instagram in Tokyo, Boris found the codebase still relied heavily on Python and ignored Meta’s Hack/HHVM infrastructure. Seniority titles meant little; he first contributed code, fixed bugs, and built personal relationships (even flying to New York for beers) before tackling the large technical‑debt effort.

The time‑zone difference gave him uninterrupted deep‑work periods, which he used to enjoy coding again.

Embracing AI: Programming’s Future Is Here

Inspired by ChatGPT, Boris sees large language models as “alien life forms” to nurture. At Anthropic he built Claude Code, where 80‑90% of the project’s code is now generated by Claude Code itself.

He describes his workflow as launching several AI agents each morning to run tasks, then reviewing the results—a shift from mere co‑piloting to full orchestration.

He warns that current AI programming is at its “worst” stage, implying rapid future improvement.

Advice for Young Engineers

When asked what he would tell his younger self, Boris answered with two words: Common Sense . In large companies, processes and politics can obscure basics.

Does the user really need this feature?

Is this a problem others also face?

Will the project fail without proper documentation?

Regardless of technology shifts—from React to AI, Web to Mobile—returning to common sense and solving real problems remains the foundation for engineers.