Thoughtful Things opened Kickstarter pledges on Sept 27, 2026 for Engram, a hardware sampler and groovebox that uses a locally run AI model to mangle incoming audio and hallucinate new sounds. Pledges start at $675, a price the company describes as a 30% discount against an unannounced retail figure that pencils out to roughly $850 to $900. Unlike cloud-based generative music tools, Engram runs its model on-device with no internet connection.
Founder Evan King pitches the device as a "field recorder for latent space" rather than a jukebox. In the campaign video, King asks the sampler for "piano" using his voice, and Engram returns a glitchy, vaguely piano-like sound rather than a clean instrument patch. The product is aimed at producers chasing uncanny textures, not at users looking for finished tracks.
The framing matters because the AI music category so far has been defined by the opposite pitch. Suno, currently being sued again by Sony and UMG, sells the promise of full songs on demand. Engram sells the promise of controlled failure — sounds that only exist because a small model got something wrong in an interesting way. Thoughtful Things even cites circuit bending, the analog tradition of deliberately short-circuiting toys and synths, as a design reference.
“It's about exploring the world of AI hallucinations and experimental sound design.”— Evan King, Founder of Thoughtful Things
Key facts
- 01Thoughtful Things opened its Engram Kickstarter on Sept 27, 2026, with pledges starting at $675.
- 02The $675 pledge is billed as a 30% discount, implying an eventual retail price of roughly $850 to $900.
- 03Engram runs a custom-trained model locally with no internet connection, and the firmware will be open to third-party models.
- 04Thoughtful Things says its audio models were trained only on CC-BY or similar commercially licensed audio.
Engram's model is custom-trained in-house, and Thoughtful Things plans to open the firmware so owners can tweak it or load their own models. That is a deliberate contrast with closed hardware and closed weights, and it lines up with a small but growing market for hackable music instruments.
On training data, the company is trying to inoculate itself against the lawsuits now hitting the rest of the AI music stack. Thoughtful Things states on the Kickstarter page: "We've trained our audio models on open datasets that only contain audio licensed for commercial use (CC-BY or similar). We have not trained and will never train our models on non-commercial, pirated, or otherwise stolen data."
“We've trained our audio models on open datasets that only contain audio licensed for commercial use (CC-BY or similar). We have not trained and will never train our models on non-commercial, pirated, or otherwise stolen data.”— Evan King, Founder of Thoughtful Things
That posture is a direct response to the legal weather. Sony and UMG are suing Suno again over training data, and UMG is separately suing DistroKid over alleged deceptive trade practices and copyright infringement. A hardware startup shipping a $675 to $900 instrument cannot absorb the kind of litigation that a venture-backed cloud service can, so a CC-BY-only training claim is both an ethical position and a survival strategy.
The device also sits in a different regulatory and business posture than software peers. Lunacy Audio's Nova platform recently launched as a marketplace for AI-powered music plug-ins, and Suno remains the reference point for one-shot song generation. Engram is neither. It is a physical box that produces raw material — samples, textures, malformed instruments — for a human producer to arrange.
The obvious risk for backers is the usual Kickstarter risk: a small hardware company promising a novel instrument at an aggressive early-bird price, with retail pricing that has not been formally set. The $675 tier is a limited run, and Thoughtful Things has not disclosed unit economics, delivery timelines, or a production partner. Buyers are effectively funding a first hardware run from a startup whose only public product is this campaign.
Engram is a small launch, but it points at where a chunk of the AI audio market is heading. The mass-market end — push-button song generation — is now defined by lawsuits and platform risk. The other end is small, local, artist-facing hardware trained on cleanly licensed data, sold to people who want AI as a texture generator rather than a songwriter. Whether that market can support $850 boxes at scale is the open question, but it is the corner of AI music that currently looks the most defensible.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
