Writing

The web has no knobs

Built for science, abandoned by the people who build it — and finally cheap enough to fix.

The web was invented at CERN, in 1989, so physicists could share the data from their experiments. That's the whole origin: a tool for science, built by a scientist, to move technical information between technical people. Then the rest of us got our hands on it.

It became the greatest commerce engine ever built, and the entire culture grew up around that — the tooling, the frameworks, the defaults, the job market, all of it bent toward marketing pages, dashboards, and checkout flows. Which would be fine, except the people the thing was actually for got quietly demoted to second-class citizens on their own platform.

Count the controls

The sharpest proof is the controls. A scientist or an engineer reaches for knobs, faders, gauges, an XY pad, an oscilloscope trace, a real-time plot, an instrument panel. Open the HTML spec and count what it actually gives you: a text box, a checkbox, a dropdown, and one sad slider of the kind you set a volume with. Not a single native control a person could use to do anything scientific. And the neglect runs all the way down — to the authors of the browsers themselves, who in three decades never decided the people the web came from deserved a knob.

It isn't only the labs. Factories, machine controls, configurators, the panels you run a real process from — industry got thrown under the same bus. The moment you need a real control or real performance, the standard stack shrugs and tells you to go build a native app instead. On the web. The thing the internet was supposed to be for.

"We don't take that job"

And ask a web developer to build the thing a scientist actually needs, and the honest answer is practically a punchline in the trade: we don't take that job. The technical, the instrumented, the high-performance — it's treated as beneath the craft, or as somebody else's problem. The web's own people looked at its founding users and decided they weren't worth serving.

It's not too late, and it's not even expensive anymore

Here's what changed, and why this is the moment. Building a genuine instrument-grade web interface — real controls, real-time rendering, native-class performance in the browser — used to be prohibitively expensive: months of specialist C, WASM, WebGL and DSP work that no agency would quote and most couldn't staff. That's over. Directed properly, a language model turns that month into a weekend — I've written about how. The thing that made serving scientists and factories "not worth it" was the cost, and the cost just collapsed.

So we can make good. We can give the underserved — the laboratories, the factories, the people who needed an instrument and were handed a checkout form — the web they were promised at the start. Real knobs. Real plots. Real machines you can drive in a browser tab. The web the internet was built for, finally built for them.

This isn't a complaint. It's a plan — and the evidence is already running, every control on it hand-made, because the web never shipped one.