I recently heard rumors about OpenAI developing a custom browser. At first, I was puzzled by the move. Then I noticed a use case from my own ChatGPT use.

While using Gmail, I kept copy-pasting previous email history into ChatGPT so it could help me draft new emails. That made the limitation feel very concrete: ChatGPT still mostly exists as a standalone app or website, so users have to feed it context by hand. It is tedious, and it keeps the whole experience inside a chat box.

But what if OpenAI controlled the browser? Since we typically use browsers for most computer activities, a custom browser could provide context to the chatbot. Imagine using Gmail within an OpenAI browser. Instead of manually pasting information, you could ask the browser to draft an email based on your current context.

How the interface could work

That made me wonder what the interface would actually look like. Would there be a persistent chat box? A side panel? Something you summon only when needed?

Currently, I create new chat sessions in ChatGPT for different tasks or workflows. If the browser maintained a history of all my interactions, it could guide me through tasks more fluidly. For example, I might use ChatGPT to draft an email to a professor.

Later, I might start a new chat to research that professor. With a context-aware browser, those tasks could connect. After researching the professor in the browser, I could ask it to draft an email using what I had just found.

The browser is where the context lives

This made me think about the browser more generally. When you use a computer, a lot of the important context is already there. You read, search, write, shop, watch, and work through the web. macOS and Windows still matter, of course, but the browser is where much of the day-to-day activity happens.

So the browser might be the best place to capture personal context. If an LLM could use that context well, maybe through a very large context window or a good Retrieval-Augmented Generation (RAG) system, it could become a much stronger assistant.

Instead of repeatedly pasting context into a chat interface and refining the answer, the assistant could start from what you are already doing. That feels like a real reason for OpenAI to care about the browser. Food for thought.