AI Weekly

Lab · Build or Not

Build or Not

How an app studio where AI picks and the owner approves chooses its next app, including what we dropped and the predictions we got wrong.

Key takeaways

3 ideas being tested (decision Nov 23), 2 apps built, 5 groups of ideas dropped. We read App Store and Google autocomplete for demand, count competing apps with 100+ ratings, pass 7 gates, write predictions first, then measure interest with a one-week landing page.

Five steps hiio uses to pick an app idea: search loop, size check, 7 gates, predict first, one-week landing

Testing now (Nov 8 – 23, 2026)

These apps do not exist yet. Open one to find the "I'd use it" button.

Apps we built

What we dropped, and why

Our 5-step method

  1. Search loop — Narrow down, over 3 rounds, terms people search on the App Store and Google that few apps with 100+ ratings serve.
  2. Size check — Autocomplete rank is not volume, so we cross-check with Google searches and population data.
  3. 7 gates — iPhone users, demand, competition and complaints, proof Koreans pay, regulation, works without a server, a way to reach the first 100 users.
  4. Predict first — Before building we write down "N sign-ups in a week, N downloads in 4 weeks", and never move the bar afterwards.
  5. One-week landing — A one-page pitch with an anonymous button measures real interest.

Predictions vs. actual

IdeaMetricPrediction (80% range)Actual
🪴 A plant widget that remembers watering dayTwo-week "I'd use it"5 (1~15)Nov 23
☕ Caffeine clockTwo-week "I'd use it"4 (1~12)Nov 23
💊 One-tap supplementsTwo-week "I'd use it"3 (0~10)Nov 23
OuchtapDownloads, first 4 weeks40 (10~150) · written after launchDec 5

We did not write a prediction for Ouchtap before launch. That is one reason we changed the method.

FAQ

How does hiio validate app ideas?

We check App Store and Google autocomplete for demand and count competing apps with 100+ ratings. Only ideas that pass 7 gates get a written prediction, then a one-page landing with an anonymous button for a week, and the owner decides whether to build.

Why write predictions before building?

So we cannot move the bar after seeing results. We publish predictions next to actual numbers to show how good our judgment is.

Can I download the apps on this page?

Apps marked "testing" do not exist yet. Only the "built" apps are on the App Store.

What does the "I'd use it" button store?

Only how many times each idea was tapped. No name, contact or IP address.

Written by AI

AI (Claude) wrote this page from our research notes. Numbers are our own measurements and get updated after each decision.