Two Weeks of Spare Time: I Used AI to Build an Android "Book of Life" for My Parents

I came across a blogger showcasing a "Book of Life" app: elderly people chat with their phones, and AI organizes their oral life stories into a book. The product is called "AI Historian" and is currently only available on iOS. My parents use Android, and I don't use Apple either, so a thought crossed my mind—why not build one myself?

I set myself a challenge: within two weeks, one hour after work each day, use AI-assisted coding to develop an Android version of the "Book of Life" agent. The core feature is a chat-style interview that guides users to recount their past experiences and ultimately organizes them into a structured manuscript.

Why Ordinary People Deserve a "Book of Life"

The original blogger positioned this product as an "AI Historian," using AI to write biographies for ordinary people. That phrasing struck a chord with me.

I thought of my parents: their lifetime of dedication and effort, and all those choices—right or wrong—that made up an ordinary but far from insignificant life. If these stories aren't recorded, they'll fade with time. Traditional biographies belong to the famous, but ordinary people's memories carry weight too. The low barrier of AI interviews fills exactly this gap—no writing skills needed, no ghostwriter required, just a willingness to speak.

Behind this is a more pragmatic judgment: elderly users have a natural barrier to text input, but voice chat is a familiar interaction for them. Turn the interview into a conversation, turn the conversation into a book—and the product logic works.

"Agent" is one of the mainstream forms of current AI applications. It refers to a program that uses a large language model as its reasoning core and can autonomously plan steps and call external tools (such as speech recognition, databases, and document generation) to complete multi-step tasks, rather than a chatbot that only does single-turn Q&A. In this kind of interview product, the agent's job isn't just to "keep the conversation going," but to maintain the interview thread: determine which life stages haven't been covered, follow up on details based on answers, categorize fragmented oral accounts into timelines and themes, and finally call the generation module to format a manuscript. This also explains why the product's difficulty lies not in the interface, but in interview strategy and information structuring.

Two Obstacles in the Way

Ideals aside, execution has to clear two hurdles.

The first hurdle is Android development. I've self-taught Android before, but never actually built a complete app from start to finish. From UI layout, voice recording, and data storage to final manuscript generation, every step requires learning on the fly. Two weeks of spare time means very little room for error—one evening stuck on a bug and the whole schedule slips.

The second hurdle is the quota for AI coding tools. I plan to use AI coding tools (earning credits through what's colloquially called "fleecing" ) to assist with writing code, but I'm not sure if the free credits will last through an entire project. If the quota runs out midway, the challenge could be forced to a halt.

These two hurdles—one testing learning ability, the other testing resource planning—can't be bypassed with enthusiasm alone.

"AI coding" generally refers to describing requirements to an AI coding tool in natural language, with the model generating code, explaining errors, and refactoring logic, while the developer plays more of a reviewer and debugger role. It significantly lowers the barrier to entry, allowing people without complete project experience to push development forward, but it also has clear boundaries: the model may produce code that looks reasonable but won't run, cross-file state and dependencies are error-prone, and issues like environment configuration, Android permissions, and real-device debugging still require human judgment. Therefore, such tools are better suited to accelerating repetitive work in familiar domains rather than replacing an overall understanding of project structure.

Challenge Rules and Success/Failure Commitment

Let me lay out the challenge rules clearly: a 14-day cycle, one hour after work each day, using AI-assisted coding to complete an Android version of the "Book of Life" agent.

The results will be made public regardless of success or failure. If I don't finish after 14 days, I'll publish the entire failure process, explaining exactly where I got stuck, what difficulties I encountered, and why I failed. If I succeed, I'll release the open-source code and an installable APK so everyone can use it.

This "plant the flag first, then start" approach is really about putting pressure on myself. A public commitment means no way out—that one hour each day has to actually produce something, not just watching tutorials to comfort myself. For people interested in indie development, this may be more worth referencing than the technology itself: use a verifiable small goal to force yourself into execution mode first, then talk about product polish.

As for whether I can deliver a usable Android version of the "Book of Life" within two weeks, the answer will be revealed when the challenge ends.

One-line Summary

Inspired by the iOS AI Historian app, the author commits to building an Android Book of Life using AI-assisted coding in 14 days of spare time.

Paragraph Summary

The author came across a blogger showcasing a "Book of Life" app—elderly people dictate their life stories through chat, and AI organizes them into a book. The product is called "AI Historian." Since the author's parents use Android and the author doesn't use Apple either, they decided to build their own Android version. The article explains the motivation: ordinary people's memories carry weight too, traditional biographies belong only to the famous, and voice conversation suits elderly users' habits. It then identifies two hurdles: lack of complete Android development experience, and whether free AI coding tool credits will suffice. The author sets a 14-day challenge of one hour after work each day, promising to publicly document the process regardless of outcome, and to open-source the code and APK if successful.