A Non-Programmer Used Claude Code to Rebuild an English-Learning App With a Server and Offline Mode

One person with no programming experience, one tool (Claude Code) and three requirements that often trip up even beginner developers: a custom server, offline mode and SEO. The author of a Habr post set out to recreate an English-vocabulary app that stopped working properly after an iOS update. By his account, the result of his first vibe-coding attempt was, "to put it mildly, a surprise." The announcement does not say how much time the build took; the answer is hidden behind the cut. But the task itself is telling.

What we know from the post

The author stresses that he has never been a programmer and that his work has nothing to do with IT. He studied English with a paid iOS app. After one of its updates the app broke, he could not fix it, and the "gestalt was left unclosed." He began studying Claude Code out of general curiosity. His knowledge of AI is basic: a general idea of what an LLM is, a few YouTube videos, and experience chatting with the free versions of DeepSeek and ChatGPT.

This profile matters. He is not an enthusiast who spends weeks on specialist forums, but a typical chatbot user. His experience shows where the line currently runs between "talking to a model" and "getting a working product."

• Builder: a beginner with no development experience; IT is unrelated to his profession.
• Tool: Claude Code, an agentic assistant for writing code.
• Task: a clone of an English-vocabulary learning app.
• Stated requirements: a server back end, offline mode, SEO.
• Author's assessment: the result turned out unexpectedly good.
• Not disclosed in the announcement: time spent, tech stack, subscription and token costs, quality of the final code.

Why a server, offline mode and SEO

A simple page of word flashcards is easy to build, and models have been able to do it for a long time. The trio in the headline is different, because each item pulls in a separate layer of engineering work.

A server means storing user data, authentication and syncing between devices. Offline mode requires a local cache and conflict-resolution logic for when a user studied words without a connection and then reconnected. SEO implies server-side rendering or static generation, correct metadata and indexable pages. All of this usually requires an understanding of architecture, not just the ability to write functions.

If the agent really did guide a beginner through all three layers, it signals a change in the human's role. The job is no longer to write code, but to state requirements clearly and check the result. One caveat: so far we see only the author's claim, not the code and not any metrics.

What to keep in mind if you want to repeat this

One successful case does not make a statistic. Authors rarely write about failed attempts, so you cannot estimate the real success rate from posts like this.

A practical sequence for a beginner:

• Start with a short specification: screens, entities (word, translation, progress), offline scenarios.
• Ask the agent to work in stages, and run the result yourself after each one.
• Ask the model up front which stack it chooses and why, and save the answer as project documentation.
• Do not store secrets and keys in the repository. Ask the agent to move them into environment variables.
• Check database backups and access permissions before real users arrive in the app.

For a personal tool, this approach already looks workable. A product with paying customers needs at least one person who can read the code and take responsibility for security. For the build time the headline asks about, check the full text of the original: that is where the author gives his own figures.