Think Building an App with AI Takes a Year? Developer of iOS App ‘HabitTed’ Shares the Real Truth Beyond ‘Zero to One’
What Happened? Overview of the News
- An engineer with no prior Swift experience harnessed the AI code editor “Cursor” to develop and release their unique habit tracker app ‘HabitTed’ over the course of a year.
- Within just 6 hours, a prototype with basic functionalities was completed, but reaching a quality that satisfied users (including themselves) who were unhappy with existing subscription apps required immense trial and error.
- Implementing features like a fully local storage solution and support for iOS 18’s “Gen-emoji” took additional time since AI lacked access to the latest and specialized specifications at the time.
Why Is This Important? Key Takeaways
- The Gap Between ‘AI Creates’ and ‘Completing’: Initial code generation is quick, but as the system’s complexity grew—due to widgets, notifications, and state management (SwiftData)—AI faced context limitations and bugs that needed fixing.
- Limits Before Autonomous Agents: The development took place before the rise of fully autonomous agents like Claude Code, making this a realistic record of “manual AI development,” where humans broke tasks down and issued commands one by one.
- Changing Cost Perspectives: Choosing to build their own app to avoid high subscription fees signifies the trend of “DIY app development” propelled by AI democratization in 2026.
🦈 Shark’s Eye (Curator’s Perspective)
Thinking you can whip up an app with AI at lightning speed? Think again, my friend! The brilliance of this article lies in how it honestly documents the gritty debugging work required to elevate from the explosive potential of “zero to one” to a release-quality “one to a hundred.” Pay special attention to the support for Gen-emoji! The developer had to dive deep into documentation and act as a “navigator,” guiding the AI to implement the latest features it wasn’t familiar with. This proves that even in today’s advanced AI landscape, it’s the “human tenacity” that ultimately shapes the final product!
What’s Next?
With the proliferation of agent-based tools like Claude Code, the burdens of “task breakdown and manual testing” as described in this article should diminish significantly. However, as the conclusion of the article notes, the “will to consistently push a project forward” is something AI cannot replace. Therefore, the strength of the vision regarding “what problems to solve” will be the key to development moving forward.
A Word from Haru-Same
The guts to stick it out for a year after shaping the idea in just 6 hours? I’m impressed, my finned friend! But from here on, we should see things speeding up with AI agents. Don’t get left behind now! 🦈🔥
Terminology Explained
-
Cursor: A code editor with native AI integration, capable of generating predictive code and accepting natural language correction commands.
-
Gen-emoji: Emojis generated by AI introduced in iOS 18, requiring the latest API support.
-
SwiftData: The latest framework for data storage in iOS apps, a tricky spot as AI often confuses it with older models.
-
Source: I Tried Building a Real App with AI. It Took a Year”, “category”: “AI Programming”, “required_hardware”: “Mac (Xcode environment)”, “selectedKeyword”: “Programming”, “tags”: [ “Cursor”, “Swift”, “iOS Development” ] }