Frequently asked questions
These are the questions a skeptical developer is right to ask before taking the project seriously. Every answer here should match the status page and the research goals. Where the honest answer is no, or not yet, that is what it says.
- Does WebCompute replace JavaScript?
No. JavaScript remains the language an application is written in. WebCompute investigates a way to describe eligible computational work so that a runtime could plan its execution across mechanisms, one of which is JavaScript itself.
- Does WebCompute replace WebAssembly?
No. WebAssembly is one of the execution strategies an execution plan could choose for CPU work. WebCompute investigates the decision above that: whether a given piece of work should run as WebAssembly, as JavaScript, or on the GPU, not how portable CPU code runs once that decision is made.
- Does WebCompute replace WebGPU?
No. WebGPU is the explicit, portable GPU execution baseline this research would target as one strategy among several. WebCompute does not compete with WebGPU's explicit model, and there is no plan to introduce a shader language of its own.
- Does WebCompute replace Web Workers?
No. Web Workers are a concurrency topology this research treats as a capability inside the CPU execution domain, not a peer target to plan against.
- Does WebCompute replace WebNN?
No. WebNN is a hardware-independent graph execution API scoped to machine learning. WebCompute investigates whether a comparable idea can hold for general-purpose computation, and the two are not in competition.
- Is WebCompute a Web standard?
No. WebCompute has not been proposed to the W3C or to any other standards body. Technologies that do go through that process, WebAssembly among them, are developed by dedicated working groups over years, through incremental proposals and multiple independent implementations, before anything is finalized. WebCompute has not reached that stage, and this site will keep saying so for as long as it remains true.
- Is there a runtime today?
No. There is no production runtime and no reference implementation. A first experiment is being scoped to test the research hypothesis on a deliberately small surface before any runtime or API work begins.
- What has been measured?
Nothing yet. No benchmark numbers, performance claims, or comparisons exist, because no experiment has been run. When the first experiment produces results, they will be published in the canonical repository regardless of whether they support the hypothesis.
- When will WebCompute ship, or become a standard?
There is no date. The project is sequenced by evidence, not a calendar: repository foundation, draft explainer, foundation reports, a scoped experiment, and then a decision about whether the idea is worth pursuing further. Any one of those steps could conclude that it is not.
Anything not covered here should be answered the same way it would be in the canonical repository. If the two ever disagree, the repository is right and this page should be corrected.