The Best Interface Is No Interface

When GPT-3 launched in June 2020, it barely made a ripple outside of hardcore tech circles. But when ChatGPT arrived just over two years later, it became one of the fastest-growing products ever.

Why the difference? ChatGPT wasn’t just about a stronger model (though GPT-3.5 was an upgrade, especially with its new RLHF tuning). The real breakthrough was the chat interface. Suddenly, people could talk to AI in the same way they messaged friends or coworkers. That familiarity made a research project feel like a practical tool.

Beyond the Chat Window

After the wave of “chat-style” copycats (Gemini, Claude, Grok, Copilot) the next big question emerged

• Will chat always be the main way we interact with AI?
• Will every app bolt on its own AI copilot?
• Or should we consolidate into one mega-app that knows all our context?

To me, ChatGPT’s success wasn’t about chat being the interface. It was about chat being familiar. It made a powerful technology approachable.

Which means the next interface for AI might not be some futuristic, auto-generated UI at all. It might just be the ones we already use every day—Slack threads, email chains, text messages, tickets in Linear.

Why Should AI Be Different?

When you hire a human, they don’t come with their own special platform. You plug them into your existing tools and workflows. Why should AI require anything else?

Yes, AI may do things humans can’t—like calling APIs or spinning up MCPs. But often, it can just use a browser, a computer, or the same communication channels we already rely on. Sometimes reinventing the wheel makes sense. Sometimes it doesn’t.

Lumetric’s Approach

At Lumetric, we believe the best interface is no new interface. We make it easy to create powerful agents you can interact with the same way you talk to coworkers - via text or email. Connect them to your apps, give them instructions in plain language, and let them handle the work you don’t want to do.