The hardware descriptions here seem on the light side. I'd want to be confident that it can handle intense time based effects.

It's promising that they seem to allow arbitrary write to the device, and only charge for tokens for the people that require the prompt playground.

Looking forward to see where this goes.

As an aside: building an ear-pleasing FDN reverb on an obscure-ish board with intense hardware optimization needs has been one of my favorite barometers for the abilities of new LLM models.

A pedal you can define with code? Kinda cool, definitely already exists, but kinda cool.

A pedal where you buy tokens to feed the ai monster to generate code to customize your pedal? Ugh. I want off this ride.

Edit - other hackable pedals:

https://www.electrosmash.com/pedalshield

https://www.op-electronics.com/en/dsp-multieffect/696-diydsp...

https://clevelandmusicco.com/hothouse-diy-digital-signal-pro...

I don't like these kind of products, what I get in breadth I lose in depth, it's like having a enormous Steam library but only play the first half an hour of each game because I have limited time to invest and too many things. I'm already overwhelmed with my Katana 100.

Like i said i don't hate the idea. I think it's just a difficult market for this kind of idea.

I play guitar, I've built many pedals, i worked in music retail for the better part of a decade.

I think "one thing does all" pedals are hugely attractive to beginners, with good reason, and I even used to recommend such things to beginners specifically. When i was young, a digital "i can try on every effect ever made!" was an amazing value proposition--assuming the product was cheap enough for my parents to buy it for me. That was usually a Zoom 606 or DigiTech GP50 -- not anything anyone would want to gig with, but an amazing birthday present for a junior guitarist.

Anyway, the reason i never pursued building guitar gear as a job, even though i built plenty of gear i used personally, was mostly because of what i mentioned:

1. Reliability above all else. Charging professional prices for guitar gear means that shit better survive being stomped on 30x/week and a couple a three beers being spilled on in its lifetime.

2. TPB, no question. Use high quality switches, switching caps as needed. Silent, un-loaded transition between on and off is huge -- though, if you're offering delay effects, tails might be desired, the user should have an override, which would require buffered bypass instead. Just be sure to communicate which you use so people know (publishing raw input/output impedance is good; designing to work with vintage impedance-sensitive fuzz, even better)

3. I think most high end guitar players are hyper-picky, and a "Jack of all trades" unit doesn't appeal--the market for that kind of device is decidedly mid-market, and must be priced appropriately to succeed.

It's cool, though. I remember using some GNU audio real-time FX app and plugging my guitar straight into my SoundBlaster and having a great time, despite the unusable lag it all induced.

I think you could capture serious players by offering presets, IR import/copying a given input sound, of course durability, and multi-functionality: is this can replace one effect at a time? Cool I guess, but i have a pedalboard. Can this replace ALL my pedals, including routing and stacking? Vastly more appealing. If it's all software, stacking effects is probably already in the codebas3, and routing would be trivial.