Does it refuse security questions? I want to red-team my own app...

We use Julia for realtime audio processing. ;)

I am also curious! A few things I would like to play with but haven't made the time for yet - real time FFT/spectrogram visualization (I want to play with some spiral things for chord / harmony visualization) - synth / live coding stuff (like supercollider but in Julia)

Re: live coding, the core part (redefining Julia functions), already works in Julia Base. You use the REPL and invokelatest() . :) https://www.youtube.com/watch?v=k-WUQFxWNj0

I have wrapped it in a project with helpers for opening audio streams and timing, but it's too vibecoded to share. It mostly exists as a Strudel-like environment to experiment with audio processing inside GPU shaders, and defining analog circuits in ModelingToolkit.

I am curious if you have any projects example, writing or example around audio processing and dsp with Julia?

The interesting part is deployment. Deploying Julia is still in flux, and especially tricky for realtime audio. I wanna do a proper write up one day, with slides and GitHub code. Perhaps as a JuliaCon talk or local meetup.

I find that DSP tends to look like procedural C code in any language. It should be straightforward to port your C++/Rust directly to Julia, and the ergonomics are much improved (especially around AutoDiff!). See https://www.youtube.com/watch?v=SvnDr9nnOZs

Here's my older talk on audio processing with Julia. The code still works, if you change the deprecated functionality for the new equivalents covered in the docs: https://www.youtube.com/watch?v=3DfVowNFI2c

Why not Julia? Serious question.

For that matter, why not C#?

> Programming language - C and C++ give you direct memory control and, with it, bug classes - buffer overflows, out-of-bounds reads and writes - that memory-safe languages like Rust eliminate at compile time. We saw consistently more false positives from projects written in memory-unsafe languages.

Re-write your Rust into C++ to drown the attacker in false positives? ;)

The filters sound digital, as someone remarked the last time this was posted. https://news.ycombinator.com/item?id=47335880

It's a compromise of speed vs quality. You can do `oversample(8, ()-> ...)` to get an increase in quality on certain occassions but that will slow down the performance.

If anyone wants to contribute gens/filters[0] the code is open source under MIT.

[0]: https://github.com/loopmaster-xyz/engine/tree/main/dsl

> Plugdata can run as a plugin in your DAW, ...

And as an iPad app! :)

Except I already have a local Mac to run Xcode. OpenRouter cannot help with that, at any price.

> 64 gigs should run a model like Gemma 4 31b

No, it can run anything in the 70B range. It's a notable quality upgrade from the 30B, which isn't obvious because the famous flurry of April releases didn't contain any 70Bs.

It can also run 120B in UD-Q3. Or 230B disk-streamed.

Home-printed 80s transistors would unlock a lot of analog-style guitar pedals. :D

What's the alternative to frontier models? Disk-streamed GLM 5.1? By the time you get a single response back, the API will be back up.

35t/s on Qwen3.5 122b on a Strix Halo. The local stuff works great now. Stop giving the corpo monopolists money.

Nice! Doing something similar with a Jumperless so that the model can reconfigure the circuit on the fly.

Oh, I remember seeing Jumperless a while ago, but completely forgot about. Combining this with something like Jumperless does sound interesting. What does your setup look like? Does Claude tell you: "try 1k resistor in parallel here"?

It's just measurements for now. But sourcing ideas from the model could be interesting!

Great use case!