The New Bottleneck
AI has made it cheap for almost anyone to build a public-information tool. It has not made it any cheaper to check one. A small radon project in Spain shows where the pressure moves when building is cheap and expertise is not.
In May I built a radon lookup for Spain. Radon causes around 838 lung-cancer deaths a year here. The official data is public, but nothing turns it into a plain answer for one household. The tool searches a municipality, shows its official classification, and points to the one thing that settles the question for a particular home: measuring it.
It took a few hours. Five years ago it would have taken me months, or I wouldn't have attempted it.
Before promoting it, I wanted someone who actually knows radon to read it. It's public-health information, and I'm a developer. So I picked researchers whose work the tool directly depends on. I wrote a one-page sheet listing every claim alongside its source, so a review could take ten minutes rather than an afternoon. Then I asked. It hasn't come together, and that's an entirely reasonable outcome. A stranger's side project is not a specialist's responsibility.
But it points to something bigger than one tool.
Where the pressure moves
AI has collapsed the cost of building. The cost of checking hasn't moved at all. The number of people who can tell you whether a radon map, a medication explainer or a benefits calculator is right is roughly what it was. Their time was always scarce. What's changed is the ratio: many more people can now build something useful-looking, and they all need the same few readers.
That changes the economics on both sides:
- For the expert, every well-meant request is unpaid work, arriving in larger numbers, often from people they can't vet. Silence is the rational default, not a failing.
- For the builder, the hard part is no longer making the thing. It's deciding what it's allowed to claim when no one qualified has checked it. I wrote about that in When Building Becomes Cheap, Responsibility Moves Upstream.
- For everyone else, the risk is a growing layer of plausible tools that nobody competent has read.
I don't think the answer is fewer tools. Small, honest public-information tools are one of the better things cheap building makes possible. But review needs a route that doesn't depend on a cold email landing on a good week: paid review, institutions that take this on as part of their remit, or professional bodies that treat it as public service. Until then, builders should make review as cheap as possible, state their limits plainly, and not promote what hasn't been checked.
What I'm doing with this one
- The tool stays up, with a plain caveat: it has not been reviewed by a radon specialist. For decisions, use the CSN's official maps and measure your home.
- I won't promote it until someone qualified has read it.
- If you work on radon in Spain and have half an hour, I'd value your criticism.