What This Is
In week three of the fifth installment of a solo developer's serialized "I Built an App with AI" series, mock data gets swapped for live iPhone health data. We see a developer who built a HealthKit plugin on the DCloud platform (HealthKit is Apple's iOS health-data API, available only on physical iPhones and inaccessible to WeChat mini-programs).
Two things stand out. AI initially recommended an off-the-shelf third-party plugin; after research, the developer overruled it and built in-house — because health data is the app's lifeline, and a third-party abandonment would leave them stranded. A task estimated at 5–8 days shipped its first draft in 8 hours once AI had a clearly written brief. Across the whole project so far, three AI suggestions have been overruled at a decision cost of roughly 8–10 hours. The developer's own takeaway: "AI excels at execution. It doesn't excel at decisions." What we take from it is the same — not how fast AI ships, but how unreliable it is at the decision points.
Industry View
The optimists treat this as the AI coding benchmark: one person, two or three hours a day, shipping core modules of a mid-size app. "Write the brief clearly and AI's execution floor far exceeds human intuition" has been widely shared.
But the piece hides a less uplifting detail. During its research, AI never verified the prerequisite "can WeChat mini-programs access HealthKit?" — the conclusion was off by a mile, and a human caught it. Our read: AI is as good at organizing information as it is at believing it; when premises need challenging, its confidence sits at rookie level. That means AI lowered the execution bar while raising the decision bar — and slick output becomes harder to push back on.
Impact on Regular People
For enterprise IT: AI tools compress the hours billed by outsourced and junior developers, but requirements review and solution selection — the call-making steps — still demand humans, and demand sharper judgment than before.
For working professionals: "Writing the brief clearly" is becoming a more valuable hard skill than it used to be — vague prompts fed to AI yield vague output; precise prompts are the new leverage.
For consumer markets: The bar to building health, expense-tracking, and habit-tracking apps keeps dropping, which means more niche, personalized products can emerge. Template-stamped lookalikes will have a tougher time.