Agentic orchestration tools are disguising themselves as "coding agents" (Claude Code, Antigravity, Open Code).
It’s a decoy.
They are the foundation for the personal AI OS of the future. Developers are the guinea pigs, and code is just the safest go-to-market strategy.
@chamath - to get better software factories you need better agents. To get better agents you need better models and benchmarks. Someone must combine execution based benchmarks (SWE-Bench Pro) with craftsmanship benchmarks to kick things to the next level. Built right > Built