Just want to point out -- some recent technologies are aimed at making this a lot better. For example, as WASM is more widely adopted, it will enable developers/companies to code a lot of things in a fast, performant way, bypassing a lot of the abstraction of JS. WebGPU, similarly, is the successor to WebGL and WebGL2 and exposes GPU APIs in a way that is more similar to how the hardware actually runs, rather than in a way that is supposed to be convenient to the developer.

Of course, these are iterative solutions that are going to take a long time for people to adopt. WASM is already fairly far along -- https://caniuse.com/wasm shows that WASM has a 92% support rate globally! And it's gaining traction with developers too -- for example, Microsoft's recent web stuff is apparently all about C# and Blazor.

Anyways, I just think it's encouraging because it shows that even big companies are moving towards a more performant web.

japhib
WebGPU [...] exposes GPU APIs in a way that is more similar to how the hardware actually runs, rather than in a way that is supposed to be convenient to the developer.

That's not really true.
It gives more control over resource allocation, but it's nowhere close to how the hardware runs. It is as far as OpenGL, D3D or Vulkan is from actual hardware.
Also it is just a wrapper over another wrapper API's, so it cannot be closer to hardware if it's implemented just as another wrapper over existing GPU apis.