Missive #240 (Chatbot fatigue).

Friends:

At Google’s big I/O event today, they announced a wave of updates to Gemini, its AI platform. It got me thinking about a question I keep returning to: will the chat interface still be how most people interact with AI five years from now?

The irony, of course, is that for the first few decades of computing, text at the command line was how we all talked to machines. Then in 1984, the graphical user interface arrived and changed everything. Now, in 2026, we find ourselves typing into blank command lines again. Back to the future, baby.

So is that a temporary regression or a permanent feature of the AI era? If the rumors are true, OpenAI and former Apple design chief Jony Ive are building an answer to that question. They are working on an all-new device that is apparently screenless, worn around the neck, aware of your surroundings, and designed to place calls, answer questions, and navigate the world on your behalf. It runs no existing apps.

No word yet on whether it can order tacos.

Sam Altman has called it “the coolest piece of technology the world will have ever seen.” Maybe. But the path to success for a completely new device category is steep. There are 5.5 billion people who own iOS or Android smartphones, and we’re deeply habituated to them. That installed base gives Apple and Google an enormous structural advantage in whatever comes next, because they already live in everyone’s pocket.

What’s clearer is that the AI opportunity is moving up the stack. The underlying LLMs will keep powering everything — but interfacing with a raw chatbot is no longer where the value is. Humans are visual animals. A well-designed graphical interface guides behavior in ways a blank text prompt simply can’t.

When you open PowerPoint or Word, you’re greeted with templates, structure, a sense of possibility. When you hit “New Chat” in ChatGPT, you get a sterile, blinking cursor.

The analyst Benedict Evans has been making a pointed version of this argument. He notes that mobile operators generate over a trillion dollars in global revenue, spend $200 billion annually on capex, and get utility-level returns for their trouble. They built extraordinary infrastructure that changed the world — and yet everything interesting happens on top of them, built by someone else. The LLM companies are in real danger of ending up in the same position: indispensable, but commoditized and low value.

Chatting with AI felt exciting in late 2022 when ChatGPT first appeared. That novelty is gone. What comes next — whether it’s a polished app experience, a wearable with no screen, or something nobody has designed yet — will be defined by the interface, not the model underneath.

Have a good week, all.

-Bret