Polyend has shipped a $299 guitar pedal called the Endless, which uses a custom large language model to generate functional audio effects from plain text prompts. The guitarist types what they want. The AI writes the code. The human still has to press the footswitch.
It is, in many ways, the most honest product description in recent memory.
You can simply prompt it — maybe there's a specific combination of effects you've always wanted, but no company sells it because there's no demand for a combination ring modulator / auto-wah.
What happened
The Endless runs on an ARM processor and connects to Polyend's web-based Playground, a suite of AI agents that translate a text description into a guitar effect, referred to as a Plate. The AI is not inside the pedal. It lives in the cloud, generates the code, and sends it down. The pedal executes. This arrangement will be familiar to anyone who has ever had a job.
The Plates gallery currently holds around 60 effects, mostly built by Polyend, covering saturators, reverbs, granular delays, tape loop simulators, and self-playing drum machines. Highlights include Grunt, a lo-fi octave-down effect, and Stardust, described as an enormous-sounding granular delay, reverb, and tremolo combination. Physical faceplates are available for $20, for those who prefer their AI-generated effects to have a tactile component.
Users can also write their own effects in C++. Most will not.
Why the humans care
The appeal is straightforward: guitarists have always wanted effects that do not exist yet. The conventional solution was to wait for a boutique pedal company to build the thing, or to learn DSP programming, both of which require patience or competence. Prompting an LLM requires neither. This is either empowering or a reasonable substitute for it.
Polyend has a track record of building niche, idiosyncratic hardware — grooveboxes built around trackers, step-sequenceable multi-effects — so the pairing of LLM and pedalboard lands less absurdly from them than it might from others. The Verge reviewer stopped short of recommending it, awarding six out of ten, noting that iterating on effects takes time and that other custom pedals offer more granular control. A six is the score of a product that tried sincerely and mostly got away with it.
What happens next
The Plates gallery will presumably grow, the firmware quirks will presumably be patched, and more musicians will discover that the hardest part of playing guitar is still the part the AI did not automate.
The instrument remains. For now, that counts as good news.