Engram Hardware Turns AI Hallucinations Into Experimental Audio
Music startup Thoughtful Things has launched a Kickstarter for Engram, an offline sampler that uses custom AI models to mangle sound.
Key highlights · 3 min read
- Thoughtful Things has introduced Engram, a hardware sampler and groovebox designed to extract e…
- Evan King, the founder of Thoughtful Things, describes the device as a field recorder for laten…
- To encourage deeper technical exploration, the startup plans to make the firmware open for users to modify or load their own custom models.
The Scale ReportThoughtful Things has introduced Engram, a hardware sampler and groovebox designed to extract experimental textures from AI-driven audio hallucinations. Rather than aiming for high-fidelity pop production, the device invites users to engage in a process akin to digital circuit-bending by pushing localized, custom-trained models to their breaking points.
Localized AI Processing
Evan King, the founder of Thoughtful Things, describes the device as a field recorder for latent space. Unlike mainstream generative AI tools that rely on cloud connectivity, Engram operates entirely offline. The company emphasizes that its models were developed exclusively using audio datasets licensed for commercial use, a distinct pivot from the controversy surrounding larger foundational models.
Custom Firmware and Hardware
To encourage deeper technical exploration, the startup plans to make the firmware open for users to modify or load their own custom models. The Scale Report understands that the Kickstarter campaign currently offers the unit at a starting price of $675, with expected retail pricing likely to settle between $850 and $900.
By prioritizing user-controllable distortion over predictable outputs, Engram attempts to reconcile the burgeoning field of AI audio with the established tradition of experimental electronic music production. This approach highlights an emerging trend in the hardware sector where developers treat generative models not as black boxes for convenience, but as interactive instruments that rely on input manipulation to reveal unique sonic artifacts.
Reporting based on coverage from AI | The Verge.




