Headroom vs Jørnal

It also checks your driver for the subgroup bug that makes some in-browser LLMs produce garbage.

Yesterday's entry is on the page above, read it whenever; it can't be rewritten.

About Headroom

It also checks your driver for the subgroup bug that makes some in-browser LLMs produce garbage. One outbound request: ~0.4 MB of public model metadata to derive bytes per token, no run data attached. Projections are measured read bandwidth ÷ bytes touched per token, at empty context; derived manifests make bytes-per-token exact, estimates fall back to an honest band. Real engines land below the ceiling; the gap is engine overhead (see methodology). Runs the transformers.js WebGPU path on the exact model your ceiling is quoted for ( onnx-community/gemma-4-E2B-it-ONNX.

About Jørnal

Yesterday's entry is on the page above, read it whenever; it can't be rewritten. A backup that runs without you having to remember. Find what you wrote, even when you can't remember the exact words. Cancel any time; you will never lose what you've written.