Using uvicorn, Hypercorn, or Gunicorn today? h2corn is a drop-in replacement with the HTTP hot path in Rust — 60–95% lower latency on the same FastAPI, Starlette, or Django app. Better performance on the same hardware.
@thsottiaux I love the efficiency, Tibo! It's finally fixed, and usage limits feel so much better. It's so much quicker at solving tasks, but sometimes the agents are a little too conservative.
From my personal experience, the hate around GPT-6.2 seems overblown. I personally really like the new models; they feel smarter, faster, and more cost-efficient. It's an all-round upgrade for me.
@FLUFISH_@thisjenlewis@Spotify Aren't those GIFs free, though? How are they making money on them? It seems like they are free to use, just like the author's—just their own spin on the idea. It's common in art to recreate someone else's piece of work.
@TokenGremlin I wish it wasn't over, even without the resets. Having something to look forward to every day with the updates was really nice. And I'm pretty sure OpenAI could keep it going if they wanted to. There are still so many things to improve and fix.
@Beamsters1 This feels like a good direction towards deranking benchmaxed models without any deeper capability. The ranking changes do make sense from my testing. Gemini and Muse being in the top was some kind of unfunny joke.