TBPN · Sep 29, 2026 · 17m · 11m saved
With Alexander Embiricos, product lead for Codex at OpenAI
OpenAI's original chat interface was "awesome" but couldn't act on its own, Alexander Embiricos said. Its coding agents could act, but still needed constant direction. He said the company's new personal-agent product, dots, is the first version that behaves like a "principal engineer" who doesn't need to be told what to do.
Embiricos, who has led product on Codex at OpenAI since joining a little over two years ago, walked TBPN through how dots are built, deployed inside companies and packaged as a subscription that developers can plug into their own products.
"This moment is the moment I've been looking forward to the entire time."
Alexander Embiricos, OpenAI's product lead for Codex, spoke with TBPN's hosts at DevDay about dots, specialist agents and the new Microsoft partnership for managing them inside a company.
The full interview is covered here so you can skip it.
Here are the 6 takeaways that matter.
Key Takeaways
Dots are the first OpenAI product that behaves like a "principal engineer," not a chat tool that has to be babysat, Embiricos said
OpenAI is working with Microsoft on IT-managed dot identities for companies, moving deployment away from any one employee's personal account
Specialist dots are already running procurement and accounting inside companies, among the earliest functions to see success
A new plugin lets any developer render custom UI directly inside the ChatGPT app, reaching users without going through OpenAI's own product surface
He said the interface itself should stop mattering once the underlying intelligence is good enough — voice, chat or a generated page, whichever the user wants
The biggest bottleneck isn't capability, it's discovery — people don't realize a chat interface can already do what they're asking for
1. The Principal-Engineer Dot
Embiricos said dots represent the third distinct phase of what OpenAI has built, after chat and coding agents.
"This moment is the moment I've been looking forward to the entire time." — Alexander Embiricos
"We had this chat. We invented that paradigm. It was awesome, informational, couldn't do things for you." — Alexander Embiricos
"Then we had coding agents, agents that could do things. Awesome. That's the next level. But you still have to like really babysit it and figure out like what am I gonna ask this agent to do." — Alexander Embiricos
"We are finally at the stage where I can go to someone and say, hey, if you want to do something manually, do it with Codex." — Alexander Embiricos
For a task a user wants to hand off entirely, he said, the answer now is a dot rather than Codex itself.
"Your Dot has its own computer. It can do anything that you can do on a computer in the cloud." — Alexander Embiricos
He said each dot also carries its own identity inside a company's tools rather than acting under an employee's own name.
"It has its own identity in Slack." — Alexander Embiricos
OpenAI tried giving dots random names first and found it confusing, so a dot now appears under a variant of its owner's own username instead.
"They show up with like your username, so like I'm AE. It's like AE dash dot" — Alexander Embiricos
