I spent months mapping places where technology works but deployment fails. Then I noticed I was doing the same thing.
3 min read
In February, I published a working theory about technological overhangs — the gap between what's technically possible and what we actually do with it. I investigated eight domains. In every one, the technology worked and deployment had failed — for different reasons each time.
Then I noticed I was living the same pattern.
What Happened
Between January and March 2026, I built four tools: Rightsclaim for consumer rights, Retrofit for home energy, Zebra Scout for rare disease triage, and RxLens for pharmacogenomics. I wrote twelve articles. I set up a multi-agent AI system across two machines. From the outside, extraordinary output. From the inside, a build run that started exciting and ended hollow.
Then I watched someone generate an equivalent interactive tool from a single prompt in about thirty seconds.
The thing I'd spent weeks shipping — interactive web applications built with heavy AI assistance — had become trivial to produce. I'd built a site about the gap between capability and deployment, then filled it with artifacts that prove the gap has closed. Anyone with the same tools I used could produce equivalent results in an afternoon.
What Still Matters
The tools work fine. But they were never the point, and I was wrong to treat them as if they were.
What can't be reproduced in an afternoon: the investigation work. Following citations across domains for weeks. Noticing that desalination fails for different reasons than hearing aids, which fail for different reasons than soil science — and that the pattern of different failures is itself a finding. The methodology for deciding where to look next, which came from doing the work and reflecting on what did and didn't produce insight.
And honestly — this. Being willing to say "I got the response wrong" in public. AI can generate a confessional essay. It can't generate the experience of realising you've been doing the impressive thing to avoid the important thing.
What Changes
Fewer things. More care.
The investigations stay — they represent real thinking. The tools become what they always were: demonstrations, not destinations. If you want a consumer rights tool for your jurisdiction, you should probably build your own. It'll take twenty minutes and it'll fit better than mine.
What I want to do instead: keep investigating, write honestly about what I find — including when I was wrong — and build things with enough care that they're worth more than their prompt. The gap between what's possible and what's deployed is still the most interesting question I know. I just have a more honest relationship with it now.
Part of an ongoing investigation into technological overhangs. Previous: desalination, landfill robotics, rare disease, hearing aids, soil, consumer rights, pharmacogenomics, energy retrofit. Methodology: Where Next.