NekozeFix is an iOS app that uses a desk-mounted iPhone or iPad to alert users to slouching with a BGM cue.

What specific issues did the developer face with AI-generated code?

Three main failures: bone-display misalignment due to coordinate system mismatches, unstable judgments from sensor noise, and the need to raise the minimum OS to iOS 17 to simplify code.

What is the development loop the developer recommends?

A four-step loop: implement based on specs, verify on real devices, reflect findings back into the spec, and record decisions in ADRs (Architecture Decision Records).

AI Coding AssistantsZenn AI/ML10h agoNov 12, 2026, 16:00 JST

NekozeFix developer: AI code needs real-device testing loops

Send on LINE

3 Key Points

What happened
@da-shina built the iOS app NekozeFix, which alerts users to slouching with a BGM cue, and found AI-written logic that passed on paper failed on real devices, including bone-display misalignment and jittering judgments.
Why it matters
The failures show that AI-generated implementation can look logically sound but still break due to physical factors like camera orientation and sensor noise, so developers need to verify on actual hardware and feed those findings back into the spec.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Get these free every morning

Context & Analysis

@da-shina, a developer at Apple World, built the iOS app NekozeFix and encountered a gap between AI-generated code and real-device behavior. The core lesson is that AI can produce code that is logically consistent but fails when confronted with physical realities such as camera orientation, sensor noise, and OS deprecations. By adopting a loop of device testing, spec revision, and ADR (Architecture Decision Records) documentation, the developer was able to correct issues like a 90° offset in coordinate systems and reduce deprecated API warnings from 15 to zero. This experience suggests that for individual developers using AI, maintaining a feedback loop between real-world testing and specification updates may be more effective than expecting AI to handle all edge cases upfront.

FAQ

What is NekozeFix?: NekozeFix is an iOS app that uses a desk-mounted iPhone or iPad to alert users to slouching with a BGM cue.
What specific issues did the developer face with AI-generated code?: Three main failures: bone-display misalignment due to coordinate system mismatches, unstable judgments from sensor noise, and the need to raise the minimum OS to iOS 17 to simplify code.
What is the development loop the developer recommends?: A four-step loop: implement based on specs, verify on real devices, reflect findings back into the spec, and record decisions in ADRs (Architecture Decision Records).

Zenn AI/MLRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.