Most AI coding benchmarks still ask the question: did the agent produce code that passes the current tests? This is a useful ...
I made my own Google TV remote with an ESP32, and it's better than the actual remote.