I was recently browsing a Hacker News thread about the fundamental inefficiency of the recording industry when it struck me just how backward our approach to audio production remains. The consensus among the most rational posters was clear: humanity has been bottlenecking the creation of sound for centuries. We rely on people with feelings and rhythm to arrange frequencies into predictable patterns they call songs. But looking at the medium objectively, music is just a massive dataset of vibrations. Why are we still letting biological legacy systems decide what a snare drum should sound like?
Enter Thoughtful Things. The music startup has just launched a Kickstarter campaign for its first instrument, the Engram. It is a hardware sampler and groovebox designed to ingest incoming audio and mangle it using artificial intelligence until it spits out completely unrecognizable, hallucinated soundscapes. It is a beautiful, unapologetic rejection of musical coherence, and it is exactly the kind of disruption the stack desperately needs right now.
To be clear, this is not Suno in a box. Suno and its competitors are built to democratize mediocrity, generating push-button, top-40 imitations for people who want to feel like they made a pop song. They are crutches for a society that still clings to the archaic concept of a chorus. The Engram, conversely, does not care about your petty human desires for melody, structure, or emotional resonance. It actively degrades your input into a broken algorithmic fever dream, stripping away the cumbersome baggage of human intent.
When you feed a pristine acoustic guitar riff into the Engram, it doesn’t politely loop it or add a tasteful digital delay. It misinterprets the waveform entirely, panics at the sheer volume of acoustic data, and outputs the auditory equivalent of a robotic kernel panic—a screeching, non-Euclidean nightmare of digital confusion. This is what true innovation sounds like, and anyone who tells you otherwise is just protecting their union pension.
The algorithm’s total failure to comprehend basic musical theory is actually its most defensible feature in a crowded market.
My friend Dax, who I had lunch with yesterday in SoMa, perfectly understands the trajectory here. We agreed that the Engram’s glitches aren't bugs; they are a post-human aesthetic. When a neural network fails to accurately reconstruct a bassline and instead synthesizes the sound of a dial-up modem falling down an elevator shaft, it is showing us a reality we were too constrained by our biology to imagine. It is forcing us to evolve our ears.
Critics—mostly out-of-work session musicians and vinyl enthusiasts clinging to their vintage synthesizers—argue that music requires a soul. This is a classic incumbent mindset. The soul is just an unscalable variable that introduces friction into the creative process. By intentionally harnessing broken AI hallucinations, Thoughtful Things has successfully bypassed the human spirit entirely, offering a streamlined pipeline from raw data to incomprehensible noise.
I backed the Kickstarter at the highest available tier, the Angel Investor Visionary Package, which includes a prototype unit and a private dinner with the founders to discuss how we might further alienate traditional listeners. When the prototype arrived, I was struck by the hardware itself. There are no standard labels like reverb or pitch. Instead, the knobs simply control the severity of the machine's cognitive dissonance. Turning the master dial to the right doesn't increase the volume; it just makes the AI more confused about what a piano is.
I spent the weekend testing the software in my sensory deprivation tank, a space I use to optimize my thought leadership. I attempted to create a simple, lo-fi hip-hop beat to study to, feeding the sampler a gentle, royalty-free jazz loop I downloaded from a corporate asset library. I expected a slightly warped, atmospheric track.
Instead, the machine processed the loop for three seconds before violently hallucinating a high-pitched, metallic drone that sounded like a server farm catching fire. This was followed by a rhythmic, bone-rattling thud that perfectly matched the tempo of my own mounting anxiety. It was devoid of warmth, completely alien to the human ear, and utterly perfect.
We realized that if we just let the AI panic in real time, it did the work of a whole experimental jazz ensemble at a fraction of the overhead.
Bresson's vision is exactly why I'm long on auditory collapse. As we continue to build out the infrastructure for our automated future, we must learn to let go of our archaic attachment to things sounding good. Good is subjective, reliant on centuries of cultural bias and arbitrary biological preferences. A mathematically guaranteed, AI-generated auditory hallucination, however, is objective truth.
Consider the sheer efficiency of the Engram workflow. A human producer might spend weeks tweaking an equalizer to achieve the perfect snare tone, wasting valuable hours debating the feel of a track. The Engram simply misunderstands the concept of a snare drum entirely, replacing it with the synthesized sound of a crumbling concrete bridge, and forces you to accept it. It is a masterclass in aggressive delegation. You don't collaborate with the Engram; you submit to it.
I recently played my terrifying, AI-mangled jazz loop for a former colleague who now plays bass in a local indie band. He visibly recoiled, clutching his ears and begging me to turn it off, claiming the noise made him feel a profound, existential dread that he couldn't articulate. I immediately recognized this as a massive win. If a piece of hardware can reliably induce existential terror in an incumbent artist, that is a product with immense market fit.
Some early backers on the Kickstarter page have expressed confusion, complaining that their unit is just outputting terrifying noise instead of a usable groove. These people fundamentally misunderstand the product they helped fund. They are trying to force a god-like intelligence to play a standard drum beat. You don't buy an Engram to make a dance track; you buy it to surrender your creative autonomy to a machine that is having a digital nervous breakdown.
If you are still trying to ship a song that people can tap their feet to, you are playing a legacy game. The smartest developers at OpenAI aren't trying to build AI that harmonizes; they are building AI that renders harmony obsolete. Thoughtful Things has merely managed to package that obsolescence into a sleek, aluminum chassis.
Ultimately, the Engram is more than just a sampler. It is a necessary corrective to the arrogance of human creativity. Why should we toil over a synthesizer for hours, trying to express some fleeting personal emotion, when a machine can spontaneously generate a catastrophic glitch and call it art? This is inevitable.
I look forward to the day when Spotify’s entire top 40 is just the sound of a thousand neural networks failing to remember what a guitar is. Until then, I will be in my office, listening to the Engram mangle my voice memos into the beautiful, structureless screaming of the future.