Make rigidity optional.
2026-07-03 — origin: a recurring itch when co-drafting slides with Claude
Working with an AI on something human-facing — not software, something like talk slides — there’s a point where I want to reach for a function old MIDI sequencers used to have, called humanize. Not “fix this bullet point.” Something closer to: we’ve discussed enough now — stop patching the document, and let what we’ve actually built between us generate the next version. Not iterate on the draft in front of us as if it were the thing holding our understanding. It never was. It was just today’s readout of a memory the two of us are building together.
That impulse doesn’t have a name yet in AI-assisted creative work. It already has one in music — several, actually.
It’s called oral transmission, and it comes in two very different tempers. One permits drift on purpose: the folk process, where a song changes a little with every singer who learns it by ear, no version more authoritative than the last, ownership never settling on anyone. The other resists drift on purpose: Vedic reciters memorized tens of thousands of verses and transmitted them for over a thousand years before any of it was written down, using techniques — pada-pāṭha, krama-pāṭha, jaṭā-pāṭha, ghana-pāṭha — built specifically to make an error impossible to hide: reciting a passage forward, then in overlapping word-pairs, then in permuted chains, so a mistake anywhere breaks the pattern and gets caught by whoever’s listening. The result is text preserved letter-for-letter, accent-for-accent, across three thousand years — in places more faithfully than manuscript traditions that had paper the entire time and still accumulated scribal errors and edits for centuries. Both traditions skip the written draft completely. Neither loses what matters by doing it.
That’s the actual practice this essay has been reaching for — not “discard the draft and rebuild.” The draft was never supposed to be where the understanding lives. What gets built, session to session, is shared memory between collaborators, and the document is just today’s projection of it — safe to throw away, not because throwing away is the point, but because nothing that mattered was ever stored there alone.
Western instrumental music mostly didn’t take that bet. It chose paper — and once the paper existed, the paper’s limitations became the instrument’s limitations. MIDI’s piano roll is a rigid grid — quantized time, quantized velocity, not enough information to represent what a human actually plays. For decades the standard fix was cosmetic: humanize functions that jitter timing and velocity by a few percent to fake the feel back in. That’s fixing MIDI as a format, and it was never going to be enough, because the format was never the real constraint.
The actual break came from Bob Moog and Don Buchla — and even that’s a later chapter than the usual telling admits. Buchla’s own instrument shipped without a keyboard, on purpose. He said so himself: “I saw no reason to borrow from a keyboard, which is a device invented to throw hammers at strings and, later on, for operating switches for electronic organs and so on.” And, more pointedly: “A keyboard is dictatorial. When you’ve got a black-and-white keyboard there, it’s hard to play anything but keyboard music. And when there’s not a black-and-white keyboard, you get into the knobs and the wires and the interconnections and timbres; you get involved in many other aspects of the music, and it’s a far more experimental way.”
A keyboard is dictatorial. … When there’s not a black-and-white keyboard, you get involved in many other aspects of the music, and it’s a far more experimental way. — Don Buchla
Underneath the keyboard is the actual target: equal temperament. Twelve equal steps to the octave isn’t a law of acoustics — it’s a Western tuning convention, standardized because it’s convenient for keyboards and modulation between keys, not because it’s the only way pitch can work. Buchla’s resistance to the keyboard was resistance to the tuning grid built into it. Swap the interface for continuous voltage instead of twelve fixed pitches, and the cultural assumption stops being invisible. It becomes a choice you can decline.
Same move, one layer down, on MIDI: quantized time and velocity aren’t just insufficiently expressive, they’re inherited conventions standing in for “how music has to be represented,” the same way equal temperament stands in for “how pitch has to work.” Neither is natural law. Both are just the convention that happened to win — and conventions that win long enough stop looking like conventions. They start looking like the ground. The piano roll, the twelve-tone grid, the chat turn, the slide template, the way a draft is supposed to progress one polite edit at a time — same shape, different century. Buchla didn’t propose a better-tempered keyboard. He built an instrument that made the convention visible and optional. Make rigidity optional: not a better prompt template, just noticing which defaults are conventions wearing the costume of necessity, and choosing a process that can decline them.
Here’s the part of the story that usually gets skipped: the instrument on its own didn’t change the culture. The pioneers’ aim was still to replicate orchestral sound with synthesis — an assimilation move, new instrument, old grammar. Sounds familiar? What actually changed the culture was Hans Zimmer renting out his own studio and building film scores directly on modular synth systems, in a room with the filmmaker in it. Not a better piano roll. A different room, a different process: collaborators invited in, cranking sound out of the machine together in real time, instead of a composer disappearing to notate alone and returning weeks later with a score. (Source: Hans Zimmer – Der Komponist Hollywoods, ARTE — watched once, not yet checked scene by scene against the claim above.)
Notice the arithmetic: two humans, one instrument between them. The machine was never the collaborator. It was the shared ground two people met on to make something neither of them owned alone.
The same move is missing from AI-assisted creative work right now. Everyone is building humanize functions — prompting tricks, style transfer, “make it sound less like AI.” Cosmetic fixes to the grid, same category error as jittering MIDI velocity.
And the instrument for the actual practice already exists — inside programming tools, pointed at code. “Rebuild this whole module using everything we discussed” is an ordinary sentence to say to an agentic coding tool, precisely because nobody there thinks the file is where the understanding lives. The shared memory of the working session is; the file just gets regenerated from it, the way a Vedic reciter regenerates the verse from memory rather than checking a manuscript. Nobody’s done the Zimmer move with that yet: taken this treat-the-file-as-disposable practice out of the terminal and made it the ordinary way to work on a deck, an essay, a score. The gap isn’t a missing instrument. It’s a missing room — and at the scale this essay actually cares about, not one person’s workflow but a culture, that room can’t be a single owned studio either. Zimmer owned his. The room this needs is closer to a bonfire: something nobody owns, that people gather around because it’s lit and warm, that anyone can feed, that goes out and gets relit somewhere else without asking permission from whoever built the first one. A studio has a landlord. A bonfire doesn’t.
Which leaves an obvious question hanging: who’s the Moog and Buchla of that room?
Chasing that question turned up something that undercuts the question itself. Buchla’s own instrument didn’t start in a private studio either. Morton Subotnick and Ramon Sender didn’t build the tool themselves; they ran the San Francisco Tape Music Center, a nonprofit collective that pooled six oscillators and some tape recorders between its members, then used a Rockefeller Foundation grant to commission an engineer — Don Buchla — to build what tape splicing couldn’t do anymore.
But naming Subotnick and Sender instead of Buchla is not the same move as naming the actual commons underneath it — it’s the same erasure one level down, with a bigger cast. The Tape Music Center also had a composer, Pauline Oliveros; a visual artist, Tony Martin; a technician, William Maginnis — the people who kept the pooled oscillators actually running, who did the connective work a grant application doesn’t itemize. “Moog and Buchla” is a great-man retelling of something that was, underneath, already collectively held — but stop at “Subotnick and Sender” and you’ve built another great-man story, just for two men instead of one. Someone always ends up playing Buchla at whatever level you stop naming people. Its actual shape isn’t a shorter list of names. It’s refusing to let the story resolve into a list of names at all: shared equipment, pooled grant money, an instrument nobody there owned alone, kept running by people the retelling keeps forgetting to mention.
I didn’t take the analogy on faith. I went and checked two folders I already had open: a proposal called malleable.music, submitted to a research funder in May 2026 as part of a wider portfolio. Its own funding pitch, written months before this essay existed, already says the essay’s sentence — that the expressiveness of a genre like livecoding is shaped by tool restrictions that are “partly accidents of limitation.” Buchla’s keyboard-as-tyranny insight, arrived at independently, about a completely different set of tools, asking the same question this essay is asking about chat turns and slide decks.
Sitting in the same portfolio, submitted the same week: a proposal for a Socratic AI tutor — federated benchmarks, EEG-based cognitive load, sovereign infrastructure. Same hand, same month, same funder. But the two proposals ask for money in different registers. The tutor’s ask reads like engineering money: tokenizer R&D, benchmark infrastructure, hardware integration, milestone-gated deliverables. malleable.music’s ask reads like creative money — a fifth of its budget line-itemed as “workshop format and pilots,” another tenth as “convening, performances, festival appearances.” The kind of budget an arts grant expects, not an engineering one.
Read the essay’s own argument back onto that: the tutor project is trying to do a Zimmer-shaped thing — change the room a Socratic tutor gets built in, not just ship a better model — while asking for Moog-shaped money: build the instrument, hit the milestone. Maybe that’s the actual gap, not a missing technical component.
Doing a Zimmer-shaped thing with Moog-shaped money — that’s trespass.ing. Moog-shaped money has its own origin worth keeping in view: his first real customers weren’t studios or universities, they were traveling evangelists, who bought his theremin kits for the eerie wail they could summon in a revival tent to make a congregation feel the presence of God. (Glinsky, Switched On: Bob Moog and the Synthesizer Revolution.) The instrument got funded before anyone had theorized what it was for, by people using it for a purpose its inventor never aimed at. Money has never waited for the categories to sort themselves out first. Not a metaphor borrowed for the occasion: the exact mechanism this site keeps naming. The loop trespasses on your authorship because it produces things that feel like your decisions the moment you initiated them. The funding trespasses on the work the same way, in reverse — an engineering grant produces engineering-shaped deliverables the moment you accept its categories, whether or not the actual work underneath is a room, a collective, an attitude change wearing an instrument as its alibi.
Whoever does this next is going to have to trespass a funding category to get the room built at all — the same way Buchla’s collective had to call an engineer’s commission an “instrument” so a grant officer would sign off on what was actually a room.
None of this earns a punchline. A tidy line like “the answer is a commons” is exactly the kind of resolved-sounding claim this piece has spent the whole time distrusting. No bonfire has been built. Nobody’s lit it. What’s sitting at the end of this essay isn’t an answer — it’s the open task the essay is leaving exactly where it found it: undone, unfunded, and not mine to finish alone.
Companion piece to Trespassing — A Brief and Sprunginnovationspotential. See also The Champion Who Never Wrote a Line, on the same authorship question in a more literal, less-negotiable form. Sources: Don Buchla, Red Bull Music Academy lecture “Passing the Acid Test”; Hans Zimmer – Der Komponist Hollywoods, ARTE; San Francisco Tape Music Center history via Wikipedia and Bandcamp Daily; Albert Glinsky, Switched On: Bob Moog and the Synthesizer Revolution, for the theremin-evangelist detail — the Rockefeller Foundation grant detail is not yet cross-checked against a primary source.