{"name":"handwired","repo":"chenglou/handwired","url":"https://midorreal.com/project/chenglou-handwired","repository":"https://github.com/chenglou/handwired","description":"Turn your regular JavaScript code into a neural net.","facts":{"license":{"value":"MIT","source":"https://github.com/chenglou/handwired","checked":"22 Nov 2026"},"language":{"value":"TypeScript","source":"https://github.com/chenglou/handwired","checked":"22 Nov 2026"},"homepage":{"value":"https://chenglou.me/handwired/","source":"https://chenglou.me/handwired/","checked":"22 Nov 2026"},"releases":{"value":"0","source":"https://github.com/chenglou/handwired/releases","checked":"20 Nov 2026"},"contributors":{"value":"1","source":"https://github.com/chenglou/handwired/graphs/contributors","checked":"20 Nov 2026"},"archived":{"value":"no","source":"https://github.com/chenglou/handwired","checked":"22 Nov 2026"}},"stars":2165.0,"writeup":{"what_it_is":"Handwired transforms regular JavaScript code into neural networks by converting control flow and state into matrix multiplication and ReLU operations. The resulting programs are fully differentiable and can run on JavaScript or WebGPU without requiring training.","audience":"JavaScript developers interested in neural networks and GPU acceleration.","claims":[{"kind":"specific","claim":"Converts if-else statements and loops into differentiable arithmetic operations","excerpt":"A program's made of if-else, loops, and state. Handwired transforms them into matrix multiplication + reglu over floats.","status":"not_checked"},{"kind":"specific","claim":"Programs remain fully differentiable","excerpt":"the full program stays differentiable","status":"not_checked"},{"kind":"specific","claim":"Weights do not require training since they are written manually","excerpt":"the weights don't need training since they're written manually","status":"not_checked"},{"kind":"specific","claim":"Can run on JavaScript or WebGPU","excerpt":"either run it on JS or WebGPU","status":"not_checked"},{"kind":"broad","claim":"Achieves GPU-level parallelism with single-threaded JavaScript code","excerpt":"achieving GPU-level of parallelism with JS code that was written in a traditional single-threaded way","status":"too_broad"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-10-16"},"why_now":null,"owner_supplied":null,"score":{"verdict":"mid","hype":31,"reality":18,"gap":13,"momentum":0.1,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T08:00:00+00:00","verdict_source":"rule"},"early_read":null,"changes":[{"date":"2026-10-15T02:15:09+00:00","text":"Repository created","source":"https://github.com/chenglou/handwired"}],"cite":"handwired is a GitHub project at https://github.com/chenglou/handwired written mainly in TypeScript, described there as \"Turn your regular JavaScript code into a neural net\".","attribution":"Data from Mid or Real, https://midorreal.com/project/chenglou-handwired, checked 22 November 2026. Attribution with a link is required.","still_real":null}