NEWS | Los Angeles startup Thoughtful Things launched a Kickstarter this week for Engram, a hardware sampler and groovebox that runs a small AI model directly on the device, no internet needed. Instead of trying to generate clean, polished sound, Engram is built to mangle whatever audio you feed it and spit out new sounds the original recording never had. Backers can get one starting at 675 dollars, with a regular retail price expected around 850 to 900 once the campaign ends.
Most AI music tools right now compete to sound as close to a finished song as possible. Engram does the opposite. Founder Evan King demos it by asking the device for piano and getting something only vaguely piano shaped back. The company calls this audio model bending, comparing it to circuit bending, where musicians used to rewire broken electronics on purpose to get weird sounds nobody designed for.
BMN sees this as a small but telling shift in how some builders are treating AI. Instead of hiding the model's mistakes behind a clean interface, Engram treats those mistakes as the actual product. The firmware is also planned to go open source, so people could eventually load in their own trained models instead of being stuck with only what the company ships.
This will not replace a laptop full of music software for most people, and 675 dollars for a niche instrument is a real ask for something built specifically to sound broken. But it points to a different question than most AI products are asking right now.
While every other AI company races to make things sound perfect, is there a real audience for a machine built to sound wrong on purpose?
Your picture beside a comment comes from Gravatar, matched to the email you signed in with. Set one there and it appears here; without one you get your initials. How we handle this