Meta's Zuckerberg, YC's Garry Tan Return to Hands-On Coding Amid AI Shift

A quiet but significant shift is underway in Big Tech's executive suites: founders with deep technical roots are personally diving back into coding, driven by the rise of AI. Mark Zuckerberg, after two decades, is reportedly shipping code diffs at Meta. Simultaneously, Garry Tan, President of Y Combinator, is back 'knee-deep' in coding 15 years after his last hands-on engineering role. This isn't mere nostalgia; it signals a strategic, founder-level bet that mastering AI-assisted development is critical as the industry moves beyond the initial 'honeymoon' phase with these tools.

The push coincides with a turbulent week for AI coding tools, exposing underlying tensions. The source code for Anthropic's Claude Code was accidentally leaked via an exposed sourcemap file. The leak revealed the tool employs 'anti-distillation' techniques to protect its model from competitors and hinted at future features like an 'always-on background agent.' More provocatively, Anthropic's issuance of DMCA copyright strikes for AI-generated code raises a fundamental legal question: can a fully AI-generated codebase even be copyrighted?

These developments underscore a broader industry scramble to define the new rules of software development. Meta is internally setting specific targets for the volume of AI-generated code, institutionalizing the practice. Meanwhile, GitHub, a cornerstone of this ecosystem, faces renewed scrutiny over its own platform's reliability, which has been flagged with issues for six years. The convergence of executive hands-on experimentation, competitive leaks, and unresolved legal and reliability questions marks a pivotal moment where the practical implementation of AI in coding is being stress-tested from the top down.