Audio · 2026-07-28 · 6 moments

Codex from 0 to 10M Users: Building ChatGPT Work — Akshay Nathan, OpenAI

OpenAI's core product engineering lead on how they are building ChatGPT Work to make AGI accessible to all of humanity: Sites, OpenClaw, Memory, Subagents, Finance, No-Code and advice. ✦ AI generated

Apple Podcasts ↗

GUEST Akshay Nathan HOST Alessio Fanelli HOST Shawn Wang (swyx)

timeline · colored by role

Anecdote◆

The power of Codex and agents is not only for developers—it can be extended to everyone, and internal non-developer adoption at OpenAI proved this much earlier than expected.

Akshay Nathan explains that OpenAI's internal data showed non-developers enthusiastically adopting Codex, feeling proud and empowered, which led to the decision to build ChatGPT Work and bring agentic capabilities beyond coders to all knowledge workers.

transcript

Akshay Nathan: I think maybe the one impetus that is most salient is when we release Codex, or even internally had Codex, it was really surprising to us, I think we recently put out some stats on this, that there was this real inflection of adoption among non-developers at OpenAI. And I through this product development process would go to these UXR sessions to talk to people internally. And the thing that stuck out to me is one, you go talk to strategic finance or marketing or whatever, and they're all using Codex for their use cases. That part's cool, but the thing that really stuck out to me is how proud people were that they were using Codex. They felt like they had a superpower, right? And what we recognized then is that the power of Codex, the power of agents, we already had this massive distribution base of people who have come to know and love ChatGPT. How do we show that to them? How do we bring it to them? Which is a hard product problem, and it's a tricky thing. There's many ways you can go about it. And so that's what we called the Merge and the Super App over time, and ultimately launched it in ChatGPT Work, is how do we do that? But it came from that initial realization that the power was not only for developers, much earlier than probably even we thought. It could be extended to everyone.

extends · 1

Mechanism

ChatGPT Work and Codex run on the same shared agent harness, differing only in opinionated UX choices about Git state visibility, diff presentation, and sandboxing defaults.

Akshay Nathan clarifies that Codex and ChatGPT Work share one harness—improvements to plugins, computer use, and artifacts benefit both. The differences are UX-driven: Codex surfaces Git diffs and file edits more explicitly and has different sandbox defaults, while Work abstracts those details for a broader audience.

transcript

Akshay Nathan: So the harness is the same. The harness is shared. In both of the products, we made improvements to the harness to make it good for knowledge work, especially as it relates to plugins or computer use or artifacts. You get that power regardless of which experience you're in. On the UX side, there's opinionated takes that we have when you're in Codex mode, what the UX should be how the UX should behave, and some stuff around the sandbox like I mentioned, but the underlying harness and capabilities should be the same.

explains mechanism · 1

Claim◆

The canonical artifact of knowledge work is shifting from slide decks and spreadsheets to interactive Sites, because HTML sites are infinitely more flexible and higher-bandwidth than traditional office formats.

Akshay describes how internal OpenAI teams—including corporate finance—have replaced monthly slide decks and spreadsheets with Sites. The reason is that Sites are infinitely flexible HTML, so users can do anything they want without hitting the feature boundaries of PowerPoint or Excel.

transcript

Akshay Nathan: I was talking to someone the other day who's on our corporate finance team, and we were mentioning how now when they have these reports that they're working on as a team month to month, historically those things were in slide decks and in spreadsheets, and now they're just in Sites. And Sites is the mechanism that they collaborate across the team. And the reason is it's somewhat higher bandwidth. These tools like PowerPoint and Excel are infinitely flexible, but at some point you reach the boundary of either as a human you may not know how to use some feature or something, or the product itself doesn't support it. But with a site you can do anything. You ask for anything and you can get that. Once people see that magic, I think it's been really valuable.

provides context · 1

Prediction

AI is blurring the boundaries between engineering, design, strategy, and operations—everyone is becoming a generalist with a specialty, and the bottleneck shifts to ideas and taste.

Akshay argues that AI enables anyone to operate across disciplines—engineering, design, strategy—so the classic T-shaped skill model evolves: everyone has a deeper specialty on one axis, but AI lets them broaden horizontally. The scarce resource becomes not building skill but having grounded ideas and good taste.

transcript

Akshay Nathan: I think my suspicion is that everyone will be shaped in a way that AI will enable everyone to become a generalist. Things that I never would be able to come up with a design before, even now I don't have maybe the visual taste required, but I can iterate on something with the help of AI. But then people will have a specialty, and that's the straight line in the T or the upward line in the T. So you can have a specialty that you're interested in—with the help of AI, you can go deeper and become better at it over time, but then you'll also be a generalist. And so with that foundation, the way you can accomplish is almost limitless. I think the bottleneck becomes ideas and taste. Because anyone can build now, it really is the era of bottoms-up ambition. And because there's so much to be built, you're always gonna be bottlenecked by the amount of ideas and amount of things that you're doing at any given time.

extends · 2

Claim

LLMs still fundamentally struggle with the instruction 'bring me new ideas,' because grounded ideas emerge from talking to users and reacting to friction—not from a vacuum.

When asked whether models can solve the ideas bottleneck, Akshay says the one automation he wishes worked—'bring me new ideas'—does not. Ideas don't emerge in a vacuum; they come from talking to users, reacting to friction, and building on planned foundations, which is where human generalists retain value.

transcript

Akshay Nathan: I would say that the one automation that I would love to work and it doesn't work is 'bring me new ideas,' right? Somehow LLMs are just not it. One interesting part about ideas is they're not in a vacuum. They usually come from somewhere, and in product development, they're coming from talking to users or reacting to friction that you're seeing or feedback, building on some foundation that you already had planned out before, whatever. And so I think there will always be value in these generalists that we talked about, closing that loop and coming up with those ideas that are grounded in that feedback or talking to users.

Claim

The critical trap in measuring AI-era productivity is conflating motion with progress—motion is now easy, but progress requires deliberate, prescriptive goal-setting.

Akshay warns that AI tooling makes motion—activity, tokens used, pull requests, outputs—easier than ever, which creates a false sense of productivity. Real progress requires teams to be prescriptive and deliberate about what they are actually trying to achieve, and to measure against that rather than surrogate metrics.

transcript

Akshay Nathan: I think maybe the trap is like conflating motion and progress. I think motion is much easier now than ever before because of the tooling that we have. But progress requires you to be very prescriptive and deliberate about what you're trying to achieve, and it goes back to our question of measurement. As a team, you should have a really prescriptive and deliberate view on what progress looks like for you and for your team. And if you don't have that, then it's very easy to conflate these two things.

Highlight slides

Codex Adoption Surprise: Non-Developers First✦ from: The power of Codex and agents is not only for developers—it can be extended to everyone, and internal non-developer adoption at OpenAI proved this much earlier than expected.From Developer Tool to Universal Agent✦ from: The power of Codex and agents is not only for developers—it can be extended to everyone, and internal non-developer adoption at OpenAI proved this much earlier than expected.The New Artifact of Knowledge Work✦ from: The canonical artifact of knowledge work is shifting from slide decks and spreadsheets to interactive Sites, because HTML sites are infinitely more flexible and higher-bandwidth than traditional office formats.Why Sites Win: Infinite Flexibility✦ from: The canonical artifact of knowledge work is shifting from slide decks and spreadsheets to interactive Sites, because HTML sites are infinitely more flexible and higher-bandwidth than traditional office formats.The Adoption Driver✦ from: The canonical artifact of knowledge work is shifting from slide decks and spreadsheets to interactive Sites, because HTML sites are infinitely more flexible and higher-bandwidth than traditional office formats.

Related episodes

Video · Lenny's Podcast

The rise of taste, human authenticity and judgment in an AI world | Adam Mosseri (Head of IG)

→

Article · Simon Willison's Weblog

Quoting Thibault Sottiaux

→

Video · Lenny's Podcast

OpenAI’s Head of Design: This is the best time in history to be a designer | Ian Silber

→