Thirty-seven opportunities. Three picks.
limna's discovery side maps markets: every app in a channel is indexed, organised into the jobs people pay to get done, and scored for opportunity. That produces a ranked list of promising territory — and a ranked list is a set of hypotheses. Markets move daily; data recorded last month is already history.
So before anything gets built, the most promising candidates sit one final exam: a full day of checking every claim against the live App Store — current pricing, real plans, this week's reviews. On 10 June, thirty-seven candidates went in. Five earned a closer look, and three became builds.
Three findings worth the day
Markets move faster than databases. A returns-management idea looked open: recorded data showed the category leader capping its free plan at three returns a month — a genuine gap for small merchants. The live check found that cap had since been removed entirely. One number had quietly changed, and with it the whole opportunity. This is exactly why every build decision starts from same-day evidence, however fresh the recorded data feels.
A good market isn't always an open door. A booking-calendar category had everything going for it on paper — strong demand, no built-in platform alternative. Up close, the picture was different: the incumbents' free tiers already covered everything a newcomer could offer first, and the category's customers clearly value hands-on, white-glove onboarding — a game for businesses built around support teams. Spotting that through research, rather than through a disappointing launch, is the system working exactly as designed. The category is on file, along with the conditions that would make it worth revisiting.
A new capability is not the same as a new opportunity. The same day, a parallel experiment tested whether fresh AI breakthroughs — new models arriving, prices falling five- to eight-fold — point directly at products nobody has built yet. Ten concepts went through the same checks; every one already existed, usually feature-for-feature, from a company with distribution. The lesson is now empirical: by the time a capability is visible in a public feed, someone with reach has already shipped it. The happy flip side: the same capability data continuously re-prices what limna does build — one model release that day cut a planned product's processing costs by an order of magnitude.
What went through
The three picks share a family resemblance: a well-loved premium leader, no free option anywhere in the field, and clear space underneath for a simpler, fairly priced product. That pattern is now being built into the selection process itself, so the next cycle starts where this one finished. The picks get named when they ship.
Why publish this
Every finding is banked. The system never pays for the same lesson twice, and each day of research makes the next one sharper. Choosing three builds from thirty-seven candidates isn't caution — it's what makes the three worth backing.