machine-learning A field note by Vikrant Sharma
Someone built a browser playground for ML-specific programming languages
A live playground for experimenting with domain-specific ML languages, no install required.
I found a browser-based playground for experimenting with ML-specific programming languages. It is open source and runs entirely in the browser. The interesting bit is not the languages themselves but the execution model. Most ML framework tutorials assume you have Python, CUDA drivers, and eight gigabytes of dependencies installed. This lets you run code samples for languages designed around tensor operations and autodiff without touching a terminal. The playground includes a few domain-specific languages built for ML primitives. Think languages where matrix multiplication is a first-class operation, not a NumPy import. The kind of thing academia builds to explore what ML code could look like if we designed syntax from scratch instead of bolting frameworks onto Python. I cannot tell from the demo whether these languages compile to something production-ready or if this is purely educational. The lack of comments on Hacker News suggests it is early. Either way, the browser execution is the clever part. WebAssembly presumably, though the source repository would confirm. If you have ever debugged a TensorFlow installation on a Friday afternoon, the appeal is obvious. Instant feedback loops beat environment setup every time. The question is whether anyone writes serious ML code in these languages or if this stays a teaching tool. I would use this to prototype tensor manipulation logic before translating it to PyTorch. The friction of switching languages is probably too high for production work, but for learning how autodiff engines think, it is faster than installing JAX.
Source: Show HN: Open source ML programming language playground
