[Thoughtful Things](https://thoughtfulthings.ai/) is betting that musicians want AI to get weird, not perfect. The startup just launched its [Kickstarter campaign](https://www.kickstarter.com/projects/evmaki/engram-generative-audio-sampler-and-groovebox) for Engram, a hardware sampler that uses AI to deliberately break and hallucinate audio in ways that would make most AI companies cringe.
This isn’t another attempt to democratize music production or create the next viral hit. Engram runs entirely offline with what the company calls “tiny AI” – custom-trained models designed to push audio generation beyond normal limits. The device takes incoming audio and mangles it through AI processing, creating sounds that exist somewhere between intention and accident.
The timing feels significant. While companies like [Suno](https://suno.ai) and [Udio](https://udio.com) chase mainstream music creation, [Thoughtful Things](https://thoughtfulthings.ai/) is going the opposite direction. They’re embracing AI’s tendency to hallucinate and break as a feature, not a bug. The 16-pad sampler and groovebox setup suggests they’re targeting experimental musicians who want to explore AI’s stranger capabilities.
Running AI models locally on hardware isn’t trivial, especially for real-time audio processing. The company claims their in-house model is specifically designed for this experimental approach, though technical details remain sparse. What’s clear is that Engram represents a different philosophy about AI in music – one that values unpredictability over polish.
The broader music tech landscape has been dominated by AI tools that try to sound “professional” or radio-ready. [OpenAI’s](https://openai.com) Jukebox, [Google’s](https://google.com) MusicLM, and commercial platforms like Suno all aim for coherent, structured output. Engram flips this entirely, positioning AI hallucinations as the main attraction rather than a problem to solve.
For musicians already working in experimental or ambient genres, this approach could unlock new creative territories. The device’s offline nature also addresses privacy concerns that have plagued cloud-based AI music tools. Your weird sonic experiments stay on your hardware, not in some company’s training dataset.
The Kickstarter launch suggests [Thoughtful Things](https://thoughtfulthings.ai/) is testing whether there’s a market for deliberately broken AI music tools. Early backing will indicate if experimental musicians are ready to embrace AI’s glitchy, unpredictable side as a legitimate creative medium.
Engram represents a fascinating counter-narrative to mainstream AI music tools. While the industry chases perfect, radio-ready output, [Thoughtful Things](https://thoughtfulthings.ai/) is betting that musicians want AI to be weird, unpredictable, and genuinely experimental. Whether this approach finds its audience will depend on how many creators are ready to embrace AI’s stranger, more chaotic possibilities as legitimate artistic tools.
