<a href="https://www.google.com/preferences/source?q=finance.biggo.com” rel=”nofollow noopener” target=”_blank”>Add preferred source
A writer who covers smartphones and IT completed an iPhone calendar app in 13 days with no hands-on programming experience and published it on the App Store. Nearly all of the code was generated by Claude Code, an AI tool from US-based Anthropic. The developer filed a first application with Apple roughly 73 hours after conceiving the idea, but the app was sent back twice during review, and it took 13 days to go live. Development costs stayed within existing flat-rate plans, with the only added expense being the $99 annual Apple Developer Program fee (about ¥16,000). The case is drawing attention as an example of a division of labor in which AI writes the code while humans make the judgments and bear the responsibility.
Key Elements
A writer who covers smartphones and IT completed an iPhone app in 13 days with no hands-on programming experience and published it on the App Store. The case is drawing attention as an example of nearly the entire development process being handed over to Claude Code, an AI coding tool from US-based Anthropic.
The app that went live is “Tsuranari Calendar,” a calendar app optimized for the inner screen of the foldable iPhone “iPhone Duo.” The developer majored in law at university and had only learned the basics of programming while studying for Japan’s Fundamental Information Technology Engineer Examination. The developer had published web apps before but had never built a native app.
Development began on the night of October 23, the day the iPhone Duo went on sale. The developer fleshed out the concept through conversations with Claude and, that same day, produced a document laying out screen designs and requirements. By the following night, an app that ran on an actual device was finished, and three nights later it was submitted to Apple for review. From concept to first submission took about 73 hours. The app finally went live on the 13th day after the device’s launch.
AI writes the code, humans bear the responsibility
What stands out about this development effort is that it adopted a method called “vibe coding,” in which the human does not read the code. Claude Code can read and write files on a computer and even run commands to verify that things work. The developer stuck to the role of describing the desired features and giving instructions for fixes after looking at the screen on an actual device.
That said, the developer did not hand over full authority to the AI. In the instruction document loaded at the start of work, the developer set a rule requiring the AI to report the scope of impact and its plan before making fixes or adding features, and not to begin implementation until the developer approved. Claude Code’s “auto mode” is a mechanism in which a separate AI model halts only operations it deems dangerous, and in practice operations such as rewriting the change history on GitHub were stopped.
The development environment also involved some ingenuity. Building an iPhone app requires Xcode for Mac, but the developer normally uses a Windows machine. Using Claude Code’s remote-control function, the developer connected from Windows to a Mac mini. Combining SSH and Tailscale, the developer set up a system by the second day in which the AI on Windows could assemble the app on the Mac and launch it in the simulator. The developer also had the AI write more than 300 tests for verifying behavior, and ran regression tests with every fix.
Apple’s review sent the app back twice
The development looked smooth, but Apple’s review sent the app back twice. The morning after the first submission, Apple sought additional information on the grounds that the submission came from a developer account with little review history. The developer explained the app’s purpose and how to use it in English, attached a screen recording of operations on an actual device, and resubmitted, but it took nearly five days to get the next response.
The second time, Apple flagged the wording on the pre-permission screen asking for access to the calendar. The developer changed the button label from “Allow Access” to “Continue” and resubmitted, and the app was approved shortly after midnight on the 13th day after launch.
Apple aims to review more than 50% of submitted apps within 24 hours and more than 90% within 48 hours, but reviews take longer when additional confirmation is needed. Dealing with review remains a heavy burden for individual developers, and even when AI writes the code, this step remains human work.
Quality control and safety challenges
Development methods in which the code is not read carry concerns about quality and safety. The developer had Claude Code itself check for safety issues and also had Codex, OpenAI’s coding AI, read the code. The checks revealed that addresses in the “shortcuts://” format, which launch a separate app, could slip into URLs extracted from notification emails, so the developer restricted the formats that could be used. The checks also turned up a problem in which, if the destination calendar was deleted while an event was being edited, the event would be saved to another calendar without permission; the developer changed the specification so that the app does not save and instead asks the user to choose again.
Accessibility verification also emerged as a challenge. The developer confirmed problems in which event displays broke down when text size was set to maximum, and in which date visibility dropped depending on color vision characteristics. Against the contrast ratio of 4.5 to 1 required by WCAG, the web content accessibility guidelines, iOS’s standard red came in at 3.55 to 1 even under normal vision and at 2.93 to 1 for people with deuteranopia, so the developer adjusted the red and blue to exceed the standard.
The developer also found a bug in which, under the Japanese era calendar setting, the year was displayed only as “8,” and standardized the display on the Western calendar.
Development costs and revenue model
The developer did not sign up for any new services for development, instead using as-is the $200-a-month (about ¥32,000) Ultra plan for Claude and the $20-a-month (about ¥3,200) Plus plan for ChatGPT that were already being used for work. The only added expense was the $99 annual Apple Developer Program fee (about ¥16,000).
Tallying Claude Code usage logs, the developer had consumed about 1.76 billion tokens by November 8, including updates after launch. Of that, about 1.73 billion tokens came from cache reads that reuse past conversations, while new input and output together totaled about 23 million tokens. Converted to Claude Opus 5.5 API pricing, that would be equivalent to about ¥86,000 (at ¥150 to the dollar), but it stayed within the flat-rate plans and no additional charges were incurred.
The app offers its basic features free of charge and has neither ads nor its own server. Operating costs are kept close to zero. After launch, the developer began offering “Wa no Iro” (Japanese Colors), a theme that lets users change the screen color scheme, for ¥300.
Lowering the barrier to individual development while challenges remain
This case shows that AI coding tools have reached a level where even individuals with little programming experience can publish a practical app. At the same time, deciding what to build, verifying behavior on actual devices, handling Apple’s review, and taking responsibility for bugs after launch all remain with humans. The developer also points out that AI-to-AI checks are no substitute for expert audits, and that because the AI writes the tests themselves, the AI’s misunderstandings can creep in.
Anthropic offers Claude Code to developers while also pushing out a suite of collaboration tools for corporate users, including meeting support, whiteboards, and spreadsheets. Under a vision of raising the productivity of knowledge workers to the level of software developers, a division-of-labor model in which AI writes the code and humans make decisions and bear responsibility is spreading from individual development to corporate business processes.
Once added, BigGo Finance appears first in Google Search Top Stories, so you get the broadest, most up-to-the-minute, and most comprehensive global financial news first.
