We Benchmarked Jev, Nimble, & a Simulated Fly Brain Playing DOOM on Matt and Ryan have a chat

In this episode, Matt and Ryan catch up on a whirlwind of AI developments, developer controversies, and hardware experiments. They kick off by dissecting OpenAI DevDay announcements—including the new Decisions API and models like Astra and Soul—before diving deep into the fast-emerging world of ultra-low-latency classification models like Jev, Nimble, and TEV1, exploring how sub-50ms local inference via Ollama can revolutionize LLM gateway routing and prompt-injection defense. The conversation takes a wild turn through unconventional compute benchmarks, contrasting AI models with a simulated f