I made 100 programming languages talk to Convex

I asked a ridiculous question: Could AI build working Convex clients in 100 different programming languages? That means getting Fortran from 1957, COBOL, x86-64 assembly, Verilog, LOLCODE and dozens more to call Convex and subscribe to realtime updates. Every client had to earn its badges by passing the same shared conformance tests against both a local backend and a real hosted deployment. No mocked results, no quietly wrapping an existing JavaScript client, and no accepting an agent's word that it worked. In this video I break down the rules, the army of coding agents, the strange language history, the implementations that fought back, the ones that never made it, and what I learned getting all 100 across the finish line. Explore all 100 languages: https://the-convex-100-language-challenge.mikecann.app/ Source code: https://github.com/mikecann/the-convex-100-language-challenge Project statistics: https://github.com/mikecann/the-convex-100-language-challenge/blob/main/STATS.md The languages that didn’t make it: https://github.com/mikecann/the-convex-100-language-challenge/blob/main/INFEASIBLE.md Everything I learned: https://github.com/mikecann/the-convex-100-language-challenge/blob/main/LESSONS.md How the build fleet worked: https://github.com/mikecann/the-convex-100-language-challenge/blob/main/BUILD-FLEET.md Learn more about Convex: https://www.convex.dev/ These are educational client-side experiments, not official Convex SDKs, server runtimes or production-ready packages. Which language should I turn into a proper full-featured client next? Let me know in the comments. #Convex #AI #Programming