At this point you might as well go Typescript → Rust. Just have an AI do the port and then dont touch the TS code anymore. Have the AI do a Rust → TS conversion if you need the code in a browser. I did this on a codebase recently and it was straightforward. And now I have total language portability.
@bendee983 OpenAI needs to focus on efficiency if they want to stay as competitive as it sounds like they do. So far I've been impressed with Sol-6, whereas Astra is basically unusable unless you like reaching your capacity limit fast.
I have an existing audio synth / editor (SASS subscription) in the works. Basically my idea would be to also be able to generate background music, and then export the recipe (sonic pi code as config data in a JSON). And then that recipe is portable as a procedurally generated audio format.
Basically I would need a wrapper around your SonicPi code could abide by AGPL, and that would also be AGPL open source.
Basically I want a command line executable:
sonic-music -f settings.json > music.wav
And then the rest of my software has that functionality.
Anyone who can switch LLMs on a whim when a new one comes out are probably not really doing anything significantly complicated to be a good judge of the models. The context knowledge ChatGPT has on my projects aren't exactly portable unless I also spend time exporting that out to markdown. A few % better on some benchmark and we got clowns declaring OpenAI already dead.
@mikerankin@aussetg Ah gotcha, thanks. I have a project where I need I display like this, but am now looking at making my own reusable display module with a Pi-style mezzanine connector.
@udiWertheimer Maybe this sounds very gen X but I haven't like a lot of music since the 90s. I still listen to all the same music I did back in high school.