Anthropic's Boris Cherny Hasn't Written Code Manually in 8 Months, Meta Conference Reveals

At Meta's AI & Data @Scale 2026 conference, Anthropic's Head of Claude Code, Boris Cherny, disclosed he has not written a single line of code manually in eight months, with every commit generated by Claude Code. The conference centered on AI-native transformation, highlighting a shift in the primary bottleneck from model capability to infrastructure, governance, and evaluation. Security researchers from Meta warned that traditional rules-based security systems break down in agentic environments, identifying four key failure modes including identity confusion, entitlement creep, recursive leakage, and prompt injection via data payloads. Meta presented an architectural response called DataVM, a trusted execution environment designed to structurally constrain an agent's operational scope from the moment it is instantiated. The broader takeaway from the event was that the developer's role is evolving from hands-on execution to high-level direction and oversight of autonomous agent systems.