A lightning talk from Blast Off Rails 2026. I demo a globally distributed feature flag system I built on open-feature + Flagd backed by S3, for under a dollar a month, surfaced via AI-driven code spelunking. Frameworks historically earned their keep by polyfilling the gaps between core technologies (Active Record over MySQL, Postgres, SQLite); the core technologies have since caught up (JSONB, native pub/sub) but frameworks often remain burdened by the polyfill. Batteries-included framing made the polyfill feel free, but the batteries aren’t free to replace — they’re ongoing support past their usefulness as a polyfill, and often drift toward implementations worse than the core they wrap. I argue AI makes fluency cheap, so the bet shifts from T-shaped to M-shaped skill stacking: stack depth in multiple cores — database, web spec, distributed systems, and cloud-native technologies (like the demo) — and let AI fill the gaps you can still evaluate. Ideally this lets us traverse more closely to the technology, taking advantage of what it does really well, and prevents reinventing tech that has outlived its polyfill usefulness.

Note: the screenshare in the video seems a bit out of sync with my talk.

Watch the talk on YouTube.