Tuned: The Two Seats Left to People

Writing the last few posts, I was quietly a little proud. Agents wrote the firmware, built the app, ran the ML pipeline, and spun a PCB. On my own.

Then I got curious. What does it actually take to put a product in front of real people? I sat down and made a list. Firmware, app, backend, PCB, enclosure, power, certification, tooling, marketing, finance, legal… by the time I stopped, there were twenty-eight. Only twelve of them were things I would call “development.”

I looked at the list again. Every seat made of code, an agent already held. That was when I understood why the whole company had felt handled. Code is all files anyway. The other sixteen I had been quietly passing over as I counted. Certification, tooling, insurance: I had pushed those to the bottom of the list as “later problems,” and left them blurry there.

The earlier posts all looked through a development lens. But development is only twelve of the twenty-eight seats. I had mistaken those twelve for the whole company. I counted the seats I know how to sit in, and treated the ones I don’t as if they weren’t there.

The real axis was ‘is it a file?’

At first I asked the same thing of each discipline: can an agent do it or not? I soon saw that this was the wrong line. The boundary between what an agent can and can’t do did not match development-or-not. The real split was somewhere else. Does the output end as a file, or not?

Everything an agent produces is a file. Code is a file, and so are a doc, a CAD model, a config. So the seats whose output is a file turned up well outside development. A positioning doc isn’t development, but in the end it’s a text file. So is the product roadmap, the landing-page code, the first draft of a certification file. Even security hardening is mostly editing config files. None of them look like development, yet each one lands as a single file. So development or not, all of these went to an agent.

The backend changed the most. I had planned to stand up a dedicated server and keep running it, which meant a whole discipline attached to the product. Then I moved to a BaaS and store billing, and the “run a server” discipline went away. What was left was a few config files. One person’s job shrank to a few lines of config.

The seat that only takes you to the doorway

Not every seat ended this neatly in a file. For some, I finish the design, the paperwork, and the modeling as files, and then there is always one last step that leaves my hands.

The RF antenna is one. Where the module sits, how much metal to keep clear around it (the keepout), I draw all of that. Up to there, it’s a file. But whether the antenna actually passes spec only comes out when the board goes in front of a certified lab’s instruments. That measurement is not a file.

Tooling is the same. How to design the mold, whether it will injection-mold cleanly (the DFM review), I do as files. But cutting the actual steel to make the mold is the mold shop’s work. Certification too: I write the whole technical file and the application. The physical test at the lab is the one thing I can’t do in their place.

A shape showed up here. A file carries me to the doorway. What gets me across the threshold was always one of three things: a physical measurement (real instruments, a real board), a purchase (buying an insurance policy, securing a part allocation), or the judgment of a licensed person (a patent opinion from an attorney, say). An agent stacks the paperwork right up to the door. It can’t open the door.

The two asterisks on ‘mine’

When I counted it up, twelve are code and twenty end as a file. The eight in between have nothing to do with development, but they are files, so they’re mine. Six make it to a doorway; two are fully outside. This is how the twenty-eight split.

In-house · 20 (end as a file): firmware, PCB, power/battery, enclosure CAD, on-device ML, mobile app, iOS, backend, web/landing, security, PM, UX/UI, marketing, signal-integrity bench, sensor bring-up, data collection, QA/reliability, program management, planning/strategy, demand validation

Doorway · 6 (I design it, one last step is external): RF/antenna, tooling/molds, incoming inspection, certification, business/legal, finance/supply chain

Human · 2 (stay a person): industrial design, domain expert

But “mine” comes with two asterisks. A job that ends in a file still takes time and carries conditions.

One asterisk is physical. Six of those twenty (the signal-integrity bench, sensor bring-up, QA/reliability, and the like) need a real board in hand before I can touch them. However much I prepare in files, execution is zero until a cheap fab run puts a board on my desk. The signal bench in particular is my first kill gate. If the signal doesn’t come out clean there, the product stops there. An agent can’t sit in this seat. I know that from having been blocked, not from reading a doc.

The other asterisk is order. Any cash that goes outside (a lab quote, tooling, insurance) gets spent only after the demand gate passes. I finish the design in files now, but the wallet opens only once “is anyone actually going to buy this” is settled. In the last post I admitted I had built it backwards. Here that regret became a rule.

So “twenty are mine” is only half right. It means no outside person is attached, and nothing more. Six of them wait on a board; all twenty wait on validation.

The two seats a person ends up in

The two that are fully outside are industrial design and the domain expert. Why those two? When I think about it, for both of them the output never quite lands as a file.

Take industrial design first. Ask an agent for an STL or a render and it will make one. That’s a file. But the real value of industrial design shows up when the thing is in your hand: the feel, where the weight sits, the line your eye catches. Numbers on a drawing don’t hold that. What says “this is a good form” is, in the end, a trained person’s eye. A file only holds the shadow of the form.

PCB layout sits on the same line. I later handed a board to an agent and then paid a contractor to finish it, and seven of the sixteen PCB routing techniques that came back need a human hand. The file is the deliverable; the judgement that produced it is not.

The domain expert is a little different. What that person offers is trust. Knowledge can be copied into a document, and that’s a file. But the fact that people believe a product because that person’s name is on it, that reputation doesn’t carry over even across a hundred pages. Trust doesn’t fit in a file.

People often say AI will replace coding first. My experience matched that. Coding went over first, because it is files. What stayed in human hands to the end was the eye for form and a person’s name. The seats in the company that are hardest to replace turned out to be exactly those.

These days, when a new discipline shows up, the first thing I check is whether it ends in a file. If it does, I hand it to the harness and stop thinking about it. If it doesn’t, that is where my own job starts. A physical act, a purchase, someone’s trust.

Twenty-six seats went to the harness. The two that are left are still empty, and that is where my time and money go now: finding the right industrial designer, and finding someone whose name the product can borrow. Because the harness bought me the time, that search is the work I do now.