There’s a threshold every cost eventually crosses where making something cheaper stops just giving you more of it and starts giving you a different thing. When long-distance calls got cheap, people didn’t place more “long-distance calls” — the category dissolved, and they just called. When compute got cheap enough to be interactive, nobody ran more overnight batch jobs; the batch job stopped being the unit of work. Below a certain price, the old noun quietly stops describing what you’re doing.
We argued elsewhere that the cap on user research was never curiosity — it was the calendar, the tax of recruiting and scheduling and moderating and writing up, one conversation at a time. That piece was about the cost. This one is about what the cost buys, because the interesting part isn’t that AI User Interviews lets you run a hundred interviews instead of five. It’s that once the hundredth interview is as cheap as the first, “an interview” stops being the thing you’re doing at all.
When research is expensive, it’s an event
Watch how a team treats research it can only afford a little of. It becomes an event. You plan a round. You schedule it, you run it, you present the findings, and then it’s over — the study is a thing that happened, with a start and an end and a slide at the end of it. And because it’s an event, it has to be scheduled somewhere, which is why, as we said before, it almost always lands late: after the spec is written, once the decision is mostly made, as confirmation rather than input. Not because anyone wants it there. Because you get one shot and you spend it defending a call you’ve already made, instead of making a better one.
Everything expensive works this way. It clusters into occasions. You don’t casually do the thing that costs a lot; you save it up, you make it count, you turn it into a milestone. The expense doesn’t just limit how often you do research — it dictates where in the process research is allowed to sit, and the answer is always “at the end, once, formally.”
When research is cheap, it’s an instrument
Now take the price down by an order of magnitude and watch the event dissolve. If asking a hundred users what they think costs about what asking one used to, you stop saving it up for a milestone. You ask before you build the thing, to find out if it’s worth building. You ask in the middle, when two designs are tied and the argument won’t resolve itself. You ask again after, not as a victory lap but because the answer might have changed. Research stops being a round you schedule and becomes something that’s simply on — an instrument you read, not an occasion you hold.
Cheap enough, and “talk to your users” stops being a project with a kickoff and a readout. It becomes a dial on the wall you can glance at any day of the week, the way you glance at whether the site is up.
This is the shift that actually matters, and it’s easy to undersell as “faster.” It isn’t faster research. It’s research that has moved to a different place in the work — off the calendar and into the process, from a gate you pass once to a signal that’s always live. The team that can ask cheaply doesn’t run better studies than the team that can’t. It runs fewer studies and more listening, which turns out to be a different relationship with your users entirely.
The same loop, closed at the other end
If this sounds familiar, it should, because it’s the studio’s oldest instinct pointed at the front of the process instead of the back. We built code-results because teams instrument everything right up to the merge button and then go dark the moment code reaches real users — the feedback loop dies at the boundary. This is the same dead loop, on the other side of the build. Before you write the spec, you’re supposed to know what users need; in practice the instrumentation stops there too, because asking was expensive, so you guess and hope.
code-results keeps the loop running past the deploy. Cheap interviews keep it running before the spec. Between the two, the whole arc — what should we build → build → did it work — finally has a live wire at both ends instead of a formal, expensive, once-a-quarter event bolted on at each edge. That’s not two products so much as the same conviction, twice: the loop is supposed to stay closed, and the reason it usually doesn’t is that keeping it closed used to cost too much.
The seam we watch for
There’s a discipline this demands, and it’s a new one — different from the one the first piece named. That one warned that cheap interviews invite lazy questions. This is subtler: an instrument you never read is just noise on the wall. When research was an event, its scarcity forced attention — you’d paid for it, so you looked. Make it continuous and it’s suddenly possible to run discovery constantly while hearing none of it, the same way a dashboard nobody opens is worse than no dashboard, because it lets you feel instrumented while flying blind. The tool makes asking cheap. It cannot make you listen. Cadence is now free; judgment about what’s worth watching, and the habit of actually watching it, are the parts that stay expensive, and we think they should.
Why this is the shape of the studio
The move under all of it is the one we keep making: find the cost everyone treats as a law of nature, question whether it has to hold, and notice that when it falls far enough the whole activity reorganizes around the new price. The headline is “talk to a hundred users as easily as one.” The thing that actually changes is quieter and bigger — that “talk to your users” stops being something you plan, and becomes something that’s just running, in the background, waiting for you to look.