A subscription-tracking app is shown beside a three-AI coding comparison setup, timer, and progress checklist.
A Pixel 8 and a stopwatch can tell you a lot about AI coding tools. They can't tell you whether an app is ready for the Play Store. XDA Developers writer Parth Shah gave three AI coding setups one hour and one detailed prompt, then installed the resulting Android apps on a Pixel 8. His verdict was that only Google Antigravity produced something he would nearly ship. It's a useful hands-on comparison, but it is one writer's subjective test on one phone, not a benchmark.

A subscription-tracking app is shown beside a three-AI coding comparison setup, timer, and progress checklist. What was tested​

The app was SubTrack, a subscription tracker. The brief called for these features:

  • a polished dashboard
  • monthly and yearly spending insights
  • upcoming renewal dates
  • categories
  • the ability to add, edit and delete subscriptions

Per the author's account, the setup was as follows:

  • Tools and models: GitHub Copilot in VS Code, "Opus 5.5" in Claude Code, and "Gemini 3.8" in Google Antigravity. These model labels come from the author. I could not independently confirm them, so treat them as his description of what he ran.
  • Same rules for all three: an identical prompt, a 60-minute limit, no follow-up prompts and no manual code fixes.
  • Evaluation: the generated APKs were installed on a Pixel 8 and put through the same real-world tests.

The write-up does not publish the prompt, the code, the APKs or a list of test cases. It also reports no automated tests, accessibility checks or performance measurements. That limits how far the results can be generalised.

What he found​

VS Code with GitHub Copilot​

The author found the core subscription features worked and the dark interface was clean and well organised. His complaints were about layout and polish:

  • The floating plus button overlapped the "Add subscription" button at the bottom of the screen.
  • The new-subscription form was basic.
  • Spacing and alignment needed refinement.

His summary was a solid foundation that wasn't yet Play Store material. He reported no major functional failures.

Claude Code​

He rated this the most visually impressive of the three, praising its dark theme, cards and dashboard. The core features worked. The friction was in the details:

  • The plus icon sat in the top-right corner, which is awkward to reach on a tall phone.
  • Add controls were duplicated across the dashboard and the Subscriptions screen.
  • Generic placeholder icons were used instead of recognisable service logos.
  • The Add Subscription form was less polished than the main screens.

Google Antigravity​

The author called this a clear winner and said it balanced looks and usability best. He highlighted these touches:

  • Tapping Add opened a bottom sheet rather than a separate generic form.
  • Each subscription had a contextual menu that slid up from the bottom, which suited one-handed use.
  • It used recognisable service icons.

He said that with a couple of minor tweaks and final release checks he would be comfortable putting it on the Play Store. The article doesn't say what those tweaks would be, and he didn't submit anything.

Reading the result carefully​

The headline claims one app was "fit for the Play Store". The evidence supports a narrower claim: in this one-hour, same-prompt test, the author preferred Antigravity's interaction design.

Several things are untested:

  • Model versus tool: the test can't separate the effect of the model from the effect of the coding tool or how each interpreted the prompt.
  • Repeatability: this was a single run per tool. AI generation varies from run to run, so another attempt could rank them differently.
  • Hardware coverage: all testing was on one Pixel 8.

The failures he flagged are real, though. Overlapping controls, duplicated actions and out-of-reach touch targets are the kind of thing worth checking in any AI-generated mobile UI.

Why "Play Store ready" is a bigger bar​

Google's own guidance shows how much sits beyond a good-looking APK.

Account-specific testing rule. Google's Play Console Help states that developers with personal accounts created after November 13, 2023, must run a closed test for their app with a minimum of 12 testers who have been opted in continuously for at least 14 days. Testers who opt in, test for fewer than 14 days, and then opt out do not count toward the requirement. This applies to personal accounts created after that date, not to every developer account.

Meeting the threshold isn't approval. It only lets you apply for production access. Google's guidance says the application asks about your closed test, your app and its production readiness, and that if your app requires additional testing, you may need to continue running your closed test.

Quality criteria. Android's core app quality guidelines go well past visual polish. They cover things such as:

  • touch targets of at least 48 dp
  • support for light and dark themes
  • back navigation and state preservation
  • no crashes or ANRs
  • targeting the current SDK
  • adherence to Play content policies and correct content ratings

None of these were reported as checked in the XDA test.

A practical checklist for AI-generated Android apps​

If you try something similar, treat the one-hour generation as the start of the process:

  1. Run the same flows on more than one device and Android version, including rotation, back navigation and returning from the recent-apps switcher.
  2. Check touch targets and duplicate actions, the exact problems this test surfaced.
  3. Review the manifest and build config, including permissions, target SDK and any debug libraries.
  4. Use Play Console testing tracks and the pre-launch report to catch stability issues.
  5. Plan for the testing window if your personal account is new: recruit more than 12 testers so one dropout doesn't reset the 14-day clock.

Bottom line​

For developers, the takeaway is modest but useful. Three popular AI coding setups all produced a working subscription tracker within an hour, and the author's preference came down to UX details such as bottom sheets, contextual menus and real icons. That's a reasonable signal about interaction design. It isn't proof of release readiness, and it isn't a ranking of the tools.

 

References

  1. I challenged VS Code, Claude Code, and Google Antigravity to build an app in an hour, but only one was fit for the Play Store XDA 2026-11-16T12:30:20+00:00
  2. Set up an open, closed, or internal test - Play Console Help support.google.com
  3. App testing requirements for new personal developer accounts - Play Console Help support.google.com