About
A studio of one that shows its work.
neverboringnow is a one-person product studio in Korea, run by Chanjoong Kim. It builds small, single-purpose software and then does the part that usually gets skipped: it checks whether the thing worked, against real outcomes, and publishes the result either way.
How we decide what survives
Numbers must be checkable.
Every figure on this site is generated from committed measurement data, not typed into the page. Where a number is not verified yet, the page says so instead of showing one.
Consistency is not accuracy.
An instrument that returns the same answer twice has only proven that it is steady. We test the harder thing, whether the steady answer is right. Where the answer does not exist yet we lock the prediction behind a published hash first, so the grade cannot be moved once it does.
Retired work stays published.
Closing a product does not take its pages down. It adds the measurement that closed it, so the record stays useful to the next person tempted by the same idea.
The record so far
The first product was an audience-research tool that asked AI personas a question and aggregated the answers. Scored against real outcomes in June 2026, it was precise, a rerun standard deviation of just 4.8pp, and wrong in a fixed direction, 7.7pp from the truth. We closed paid plans in August 2026, before anyone had ever paid, so that nobody would become the first customer of a number we could not defend. The case study has the full measurement.
That first scoring was retrospective, which is the weak kind. So 44 predictions had been sealed behind a published hash in June 2026, before their answers existed. The first 7 resolved in July, all of them from one football tournament. Nothing in them contradicts the retrospective finding, and 7 is far too few to confirm it either. The case study has those numbers and the reasons they are not a verdict yet. The remaining 37 are still sealed and we do not get to choose when they land.
What runs today is a portfolio of mini-apps on a Korean super-app platform, built one at a time where the audience already was, the opposite of how the first product was launched. Both of the summed counts the platform reports are on the home page, with what they do and do not mean. What they are called, what each one does and how any single one performs are not published, and that is a decision rather than a gap: the aggregate is the part that says something about the studio, and the list is the part that is only useful to somebody copying it.
The five-phase plan this studio was founded on is kept exactly as it was written in April 2026, with the result stamped on top of it, on the archived vision page.
One person, so the list of things we can take on is short. If your problem is narrow and the interesting part is finding out whether it worked, that is the kind we say yes to.