Port is about 80% done, but completely mechanical with AI and an oracle (the Dolphin emulator)

Comments URL: https://news.ycombinator.com/item?id=49964603

Points: 4

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=49923064

Points: 17

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=49022605

Points: 16

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48848783

Points: 17

# Comments: 0

Honestly the WASM part was pretty straightforward, other than some shader performance issues with Burn + CubeCL , but nothing this project can really change.

The hardest part was the weight porting honestly, since I had to perfectly recreate all of the digital signal processing and model, any deviations from the original PyTorch implementation, and you ended up with subtle loss of signal in the outputs. Added that folder `bench/` to step through the two implementations piece by piece to help debug where they were diverging -- and it's still not a perfect replication of the original after all the tuning!

Try it now: https://nikhilunni.github.io/demucs-rs/ (needs a WebGPU-capable browser — Chrome/Edge work best)

GitHub: https://github.com/nikhilunni/demucs-rs

It runs three ways:

- In the browser — the full ML inference pipeline compiles to WASM and runs on your GPU via WebGPU. No uploads, nothing leaves your machine.

- Native CLI — Metal on macOS, Vulkan on Linux/Windows. Faster than the browser path.

- DAW plugin — VST3/CLAP plugin for macOS with a native SwiftUI UI. Load a track, separate it, drag stems directly into your DAW timeline, or play as a MIDI instrument with solo / faders.

The core inference library is built on Burn (https://burn.dev), a Rust deep learning framework. The same `demucs-core` crate compiles to both native and `wasm32-unknown-unknown` — the only thing that changes is the GPU backend.

Model weights are F16 safetensors hosted on Hugging Face and downloaded / cached automatically on first use on all platforms. Three variants: standard 4-stem (84 MB), 6-stem with guitar/piano (84 MB), and a fine-tuned bag-of-4-models for best quality (333 MB).

The existing implementations I found online were mostly wrappers around the original Python implementation, and not very portable -- the model works remarkably well and I wanted to be able to quickly create samples / remixes without leaving the DAW or my browser. Right now the implementation is pretty MacOS heavy, as that's what I'm testing with, but all of the building blocks for other platforms are ready to build on. I want this to grow to be a general utility for music producers, not just "works on my machine."

It was a fun first foray into DSP and the state of the art of ML over WASM, with lots of help from Claude!

Comments URL: https://news.ycombinator.com/item?id=47234566

Points: 19

# Comments: 3

Cortex helps engineering teams understand and improve their services. By aggregating data from tools like Datadog and Okta, we help teams understand their architecture at a glance – everything from ownership to runbooks. Using this data, we enable engineers to build report cards that score services against initiatives and service quality standards, helping teams build high quality software and instill a culture of quality. The tools teams use today (excel sheets and confluence) are cumbersome and rudimentary - we’re on a mission to create a new category of engineering tooling, to solve the organizational complexity of microservices

https://getcortexapp.com/ Contact: recruiting@getcortexapp.com

Cortex helps engineering teams understand and improve their services. By aggregating data from tools like Datadog and Okta, we help teams understand their architecture at a glance – everything from ownership to runbooks. Using this data, we enable engineers to build report cards that score services against initiatives and service quality standards, helping teams build high quality software and instill a culture of quality. The tools teams use today (excel sheets and confluence) are cumbersome and rudimentary - we’re on a mission to create a new category of engineering tooling, to solve the organizational complexity of microservices