We need a Browser War for AI chat

We need a “Browser War” fought over this new turf called “chat.”

Or more precisely, we need the conditions that would make a war possible.

There needs to be a common turf worth fighting over. That’s what is missing. The big companies in AI Chat (OpenAI, Google, Anthropic, etc) are all building walled-garden apps. They fight for customers & revenue, sure, but not about dominance over “AI Chat” itself.

The result is that they are building islands, not moats.

Focus on the pain:

How do I get a second opinion from Claude on my long-running conversation with ChatGPT? I ask ChatGPT to print a summary, and then copy-paste it into a new chat session with Claude. Then the brain gets split, and I end up being a courier running copies of copies back and forth between two browser tabs. Or more, if I want a third opinion.

I’m not the only one annoyed by this clunkiness. Many many “orchestrator” or “harness” products are taking a crack at this coordination problem. Problem is: each one becomes a separate island.

More importantly: they all become indistinguishable & fungible. AI Chat will become more and more commodified as the models converge. Orchestrators will also rise and fall, as they each try solving the same problem the same way.

The ur-sin of all this - why we are in this mess - was that history zigged with the launch of ChatGPT in 2022: the AI Chat product coupled the “service” with the “packaging” (ChatGPT was both a model and also an end-user application). We’ve been on that habit-forming path ever since. Now history needs to zag.

The reason there was a Browser War was because there was a fairly-standardized common turf that was rife for conquest. This constellation of things like HTML, CSS, HTTP were all independent tools. And this thing called a Web Browser was a convenient container to provide a human-facing experience. The War was not over services. The Web Browser is not Ebay or Facebook. But the distribution channel for those companies.

Wouldn’t it be nice if we had a “Web Browser” moment for AI Chat. Let the companies buy their datacenters and build their models. But they don’t also own the application experience. My history of chats in Gemini and my history of chats in Claude and my history of chats in ChatGPT should not be isolated in separate mobile apps. A “Chat Browser” could be the vendor-neutral container.

And I doubt it should look anything like a web browser. The web browser was invented to be essentially a television screen you can drive with a mouse and keyboard. It’s visual. Some sections of its content is “clickable”. It has a sense of a “timeline” with its back-and-forward buttons. Things we take for granted today, but were genuine creative inspirations back when people were wrestling with this problem of the “internet” for the first time.

We need to have a “Browser War” moment because that will force us to chart the undiscovered country for the first time. What is this experience of AI Chat at bottom? It’s much wider than the copy-cat chat-app products with their empty prompt box and a side-panel of history. The current state of play for AI Chat feels as half-baked as if we were all surfing the internet by printing out raw HTML on a command line.

ChatGPT circa 2022 was a fun gimmick. But prompting a text-box over and over again, producing a back-and-forth screenplay of text that spans a hundred pages is not the only way to package this new thing called AI Chat.