AI daily use case · November 11, 2026
Running coding agents on large, well-specified engineering projects
Teams are handing whole codebases to swarms of coding agents and reviewing only a small share of the output. The AWS emulator was built this way, with tens of thousands of commits in three weeks. Google's agents wrote millions of lines of proof code and flagged about one percent for humans to check. A builder also used Claude Opus 5.5 to build a tool that recompiles the Black Ops II game into a native ARM64 Mac app, which is a clear, testable target.