Key Takeaways

  • Chatbots force users to do the initiation work; proactive agents like Dots run continuously on a cloud computer and work while you sleep.
  • The primary leap in products like Dots is a change in form factor rather than model capability, pairing frontier models with persistent environments.
  • Brockman expects the chat text box to disappear entirely, replacing static buttons and menus with interfaces generated on the fly.
  • Enterprise and consumer workflows are collapsing into the same systems as foundational models manage Slack threads, inboxes, and calendar conflicts without separate tool stacks.

The Chatbot Text Box Was Always Temporary

Software companies spent two years slapping chat text boxes into every product. OpenAI President Greg Brockman thinks that design pattern is already obsolete.

“The chatbt the text box, right? That's where it started,” Brockman said. “It was a text box and what we're heading to, there isn't even a text box, right? It's just like that is that is the AI we were promised.”

A text box demands that the user initiate every interaction, invent the prompt, and wait for a response. That is reactive software dressed up as intelligence. If you have to tell an assistant to check your email every forty minutes, you are still doing the administrative labor. The assistant is just a faster keyboard.

Give the Model an Environment Instead of a Prompt

The real shift is moving from conversational turn-taking to independent execution. Brockman points to Dots, an agent designed to manage tasks across inboxes and Slack channels without human prompting.

“So dots is a proactive agent. It's always on always working. You stop working it keeps working,” Brockman noted.

The technical difference between a reactive chatbot and an agent like Dots is not a leap in raw model intelligence. It is the operating environment.

As Brockman explained, “In some ways what do is it's a change to the form factor, not the capability, right? It's like it's an astro level model. It's got its own cloud computer.”

By running on its own cloud machine, the model monitors communication streams, spots calendar clashes, and flags broken commitments across tools on its own schedule. It does not wait for a ping. It executes in background threads and surfaces conclusions only when action is necessary.

Interfaces Will Generate on the Fly and Disappear

When software does need to communicate with you, Brockman argues it will not happen through fixed menus, dropdowns, or navigation sidebars.

“The best UI is no UI and I think in AI that is where we're going,” Brockman said.

When visual information is required, the model will generate the user interface at runtime and discard it the second you are done. Brockman noted: “So now the AI can do generative UI on the fly. And so I think that even though there's no fixed UI, there's no buttons, there's no menus to learn, having an AI that can actually craft the result to be visually exactly what you want.”

Instead of engineering fifteen settings panels and a dashboard your users must learn to navigate, software can render a single custom comparison table or interactive card solely for the specific choice you need to make right now. Once you decide, the screen returns to zero.

What to Do With This

Audit your product roadmap and cut any feature that requires the user to remember to open your app. If you are building an AI assistant, stop building a text box input and instead provision a persistent worker with access to your user's communication channels. Have it monitor state changes, draft resolutions to problems before the user notices them, and present an ephemeral, single-purpose interface only when confirmation is required.