Chrome 147–148 ships WebGPU upgrades — and we teach this directly

What shipped

Chrome 147 and 148 expand WebGPU compute-shader performance and texture-format support. For ML workloads, the practical wins are faster GPU memory transfers and broader format compatibility for model inputs and intermediate tensors.

Why it matters for FE engineers

WebGPU is to inference what requestAnimationFrame is to paint — the buffer-per-frame discipline you already know, applied to tensors. If you've worked with Float32Array and texture uploads in WebGL, the mental model is the same: stage data in a typed array, push to GPU, dispatch a compute pass, read back.

The 147–148 update doesn't change that mental model. It just makes more of it cheap.

Where we teach this

Deep Orbit → Optimization Edge → WebGPU Acceleration covers compute shaders, buffer lifecycle, and the FE-to-WebGPU bridge from Float32Array upload to tf.tensor on the GPU backend.

If you opened Chrome's release notes and thought "I should learn this," that's the lesson.