source code on a dark screen

The Day AI Moved Into My Editor

For more than a year, using AI meant opening a browser tab, pasting some code, and copying the answer back. It was useful, but it sat outside the work.

Bringing it closer

In February 2024 I forked an open-source extension that connects ChatGPT to Visual Studio Code. The fork was not a product. It was a way to look inside and understand how an assistant gets context from an editor. Tools like GitHub Copilot had already shown how natural this could feel.

Context changes everything

The difference is not only convenience. A model that can see the file you are working on gives better answers than one that receives a fragment and a guess. The quality of an assistant depends heavily on what it is allowed to see.

The new risk

That same closeness brings a new question for anyone who works with client systems: what exactly leaves your machine when the assistant reads your code? Configuration files, credentials and customer data can sit one tab away from the code you are editing. Convenience does not remove the responsibility to know where your data goes.

The lesson for me is that AI is becoming infrastructure. And like any infrastructure, it deserves to be understood, configured and watched.

Leave a Reply

Your email address will not be published. Required fields are marked *