Claude Code now generates 4% of all public GitHub commits, with Anthropic’s engineering team shipping 10–30 fully AI-written pull requests daily—reviewed only for intent and correctness. Its success stems from counterintuitive principles: underfunding teams to spur creativity, granting unlimited tokens to encourage fearless experimentation, and designing around latent demand—where users repurpose tools in unexpected ways (e.g., using coding tools for genome analysis or gardening), guiding the fast 10-day build of Cowork. Boris argues coding is effectively 'solved' as a low-level task, shifting engineers’ role toward high-level problem definition, integration, and user empathy—mirroring historical shifts like the printing press. He emphasizes model-centric design: expose powerful, general-purpose agents (like Opus 4.6) with rich tool access, not narrow optimizations. While rapid innovation brings disorientation—requiring unlearning old debugging habits—it also boosts job satisfaction for 70% of engineers and product managers. Safety remains core, enforced via Anthropic’s three-layer framework and early public releases. Ultimately, Boris champions patience, iteration, and human-centered vision—even as AI agents operate autonomously for days.