Before Fable 5 was shut down, it pushed Gemma 4 to 255 tok/s on WebGPU. Some didn't believe it was real.
Today we're releasing the demo and kernels it wrote for you to see yourself. Run it locally in your browser.
Agentic kernel optimization is the future of on-device inference
Reads like AI psychosis self-convincing confirmation-bias nonsense. “I told it to make it fast, and holy shit, one day it became really fast!” yeah probably because it ripped out all non-essential code like, oh, say, all the safety and security and reliability and compatibility features. Optimizing for a specific task is not that hard until you start having actual real-world requirements and responsibilities which none of these sloppers ever seem to understand.
Reads like AI psychosis self-convincing confirmation-bias nonsense. “I told it to make it fast, and holy shit, one day it became really fast!” yeah probably because it ripped out all non-essential code like, oh, say, all the safety and security and reliability and compatibility features. Optimizing for a specific task is not that hard until you start having actual real-world requirements and responsibilities which none of these sloppers ever seem to understand.