# I kept losing the things AI made for me, so I built an iPhone library for them

> Source: <https://dev.to/spingras/i-kept-losing-the-things-ai-made-for-me-so-i-built-an-iphone-library-for-them-4pe7>
> Published: 2026-07-18 13:13:22+00:00

I use Claude and other AIs all the time.

A single chat will give me a component. Another will give me a small Python script, Mermaid diagram, Markdown file, or utility that addresses the issue I faced that day.

And then it's gone.

The AI doesn't fail. The problem is what happens once it comes up with something useful, as it ends up locked away in a chat history that wasn't designed to serve as an index.

After a week, I only remember creating "that CSV helper" or "that JSX tool to compare options." I'll remember a few keywords used in the prompt, but not the chat, date, or even which AI generated it. Re-discovering it requires digging through old chats, scrolling, and hoping that I recognize the relevant result.

That was why I made AIpine (ai-pine).

A conversation provides context: the prompt, the iterations, the explanation, and the errors committed in the process. This is helpful during the process of creation.

However, the final product needs somewhere else to be stored.

There were some items that I wished to save which did not have the same characteristics:

Treating them as generic attachments loses their essence. Some need to be rendered, and others need to be read as code. Usually, both would be required: a preview for a reminder of its purpose and the original for editing and reuse.

I wanted a simple iPhone workflow of saving an interesting AI-generated file, finding it later, and opening it in the correct format without making it a separate cloud account or a new project.

This gave me some requirements:

AIpine is the iPhone and iPad app I created for this workflow.

It is a local file library for AI-generated files. They can be imported into the app, organised within it, searched afterwards, reviewed for their source, and previewed in case their rendered form is more interesting than text.

The supported formats are JSX, HTML, Mermaid, Markdown, CSV and TSV, JSON, YAML, TOML, SQL, Python, shell, TypeScript, JavaScript, SVG, and plain text.

In case of a JSX editor, it is handy to see the live interface alongside the code. In case of Mermaid, it is nice to see a diagram instead of a syntax. In case of scripts or other files, it is useful to keep an accessible and searchable reference which does not get hidden under a chat name.

The first decision to make about the products we were developing was that preview mode must not replace source mode.

AI-generated content is not a miraculous thing. You have to examine it, edit it, understand the dependencies behind it, and sometimes copy parts of it to other projects. In many cases, the code behind the output is just as important as the output itself.

That is why AIpine places them both in proximity: useful preview if it exists, original code/text otherwise.

This makes our application not only more usable, but also more transparent.

Nowadays, the question isn’t just: “Can AI create this?”

Yes, it can.

A more pressing one is: What becomes of the practical output once the conversation is out of view?

For me, the answer is: a personal library. A space where the tools, notes, schematics, and scripts that I create using AI are collected without ending up being unsearchable.

This is my solution for creating such a library on iPhone: AIpine.

What do you continue to create using AI, and how quickly do you end up forgetting where you saved the file?
