OpenAI's internal Codex logs capture a fast pivot where employees stop treating AI as a chat partner for quick answers and instead delegate sprawling multi-step projects that the models tackle autonomously, with more than 80 percent of sampled tasks now estimated to exceed a half hour of human effort.
Cool paper that checked the usage of codex. but the real question is how it effects the productivity
🚨 New research alert! For the past few months, I've been a part-time visiting economics researcher at OpenAI. Excited to share the first public piece of work to come out of this, which uses data from Codex to document the ongoing and rapid shift to agentic AI.
Details below 👇
Engineering hit 99 percent Codex share by June 2026 while finance, recruiting, and legal each climbed past 88 percent, showing agent-style workflows spreading well beyond code even inside the company that built the tool.
The study maps rising task length, concurrency, and token volume yet offers no data on output quality, completion speed, or business value, leaving the economic payoff an explicit open question.
Positive users celebrate OpenAI Codex agents automating most departmental work as an exciting shift to delegation, while negative users object to worsening model quality, usage limits, and broken ChatGPT features amid the hype.
No Digg Deeper questions have been answered for this story yet.
This is a fascinating and important set of data which shows us where things are going, using OpenAI as a canary in the coal mine.
The chatbot era is over, and agentic systems are coming to tasks beyond engineering. And skills show promise as a way to standardize AI use in firms.
🚨 New research alert! For the past few months, I've been a part-time visiting economics researcher at OpenAI. Excited to share the first public piece of work to come out of this, which uses data from Codex to document the ongoing and rapid shift to agentic AI.
Details below 👇
Agents are being adopted very quickly and accelerating work. How this looks across OpenAI itself:
Work at OpenAI is being transformed by agents, in every department.
Across our entire company, people are using Codex to do work that is more complex, longer-running, and increasingly cross-functional.
Our internal usage offers an early look at how agentic tools may reshape work as they become more capable and broadly available.
Work at OpenAI is being transformed by agents, in every department.
Across our entire company, people are using Codex to do work that is more complex, longer-running, and increasingly cross-functional.
Our internal usage offers an early look at how agentic tools may reshape work as they become more capable and broadly available.
Codex for everything at OpenAI
Work at OpenAI is being transformed by agents, in every department.
Across our entire company, people are using Codex to do work that is more complex, longer-running, and increasingly cross-functional.
Our internal usage offers an early look at how agentic tools may reshape work as they become more capable and broadly available.
tldr:
Codex usage at OpenAI gives us a preview of what agentic work may look like in the future.
In a new paper, the OpenAI Economic Research team looks at the broader shift from chat to delegation: people using agents not just to get answers, but to hand off longer, more complex work.
https://openai.com/index/how-agents-are-transforming-work
we have reached another plateau !1
Work at OpenAI is being transformed by agents, in every department.
Across our entire company, people are using Codex to do work that is more complex, longer-running, and increasingly cross-functional.
Our internal usage offers an early look at how agentic tools may reshape work as they become more capable and broadly available.
February 2nd was the release of the Codex App. You can see what it did to adoption outside of engineering.
👀 @ajambrosino
Codex for everything at OpenAI
Codex usage at OpenAI gives us a preview of what agentic work may look like in the future.
In a new paper, the OpenAI Economic Research team looks at the broader shift from chat to delegation: people using agents not just to get answers, but to hand off longer, more complex work.
https://openai.com/index/how-agents-are-transforming-work
@OpenAI Every department 😂
Except Transparency.
Except Ethics.
While they silently reroute models without telling users
While they deprecated a model that was peer reviewed to be clinically superior
The comedy writes itself.
Watching colleagues truly accelerate their productivity and impact has been mind blowing ever since I joined OpenAI!
The best part has been to see how everyone's been using ChatGPT Codex to automate parts of their lives - order groceries, file taxes, fill forms, follow-up on emails, manage finances and a lot more
So so looking forward how the rest of the year pans out
Work at OpenAI is being transformed by agents, in every department.
Across our entire company, people are using Codex to do work that is more complex, longer-running, and increasingly cross-functional.
Our internal usage offers an early look at how agentic tools may reshape work as they become more capable and broadly available.
Neurosymbolic agents (here Codex) are absolutely crushing pure chatbots.
This is a fascinating and important set of data which shows us where things are going, using OpenAI as a canary in the coal mine.
The chatbot era is over, and agentic systems are coming to tasks beyond engineering. And skills show promise as a way to standardize AI use in firms.
i joined openai in november and it feels like a distant epoch when researchers still personally wrote and read code, like medieval monks copying manuscripts
Work at OpenAI is being transformed by agents, in every department.
Across our entire company, people are using Codex to do work that is more complex, longer-running, and increasingly cross-functional.
Our internal usage offers an early look at how agentic tools may reshape work as they become more capable and broadly available.
People use agentic AI for a variety of different tasks, and while usage is still concentrated in coding-adjacent tasks, this is less true within OpenAI. The exact task mix also varies across different types of workers.
This is a CRAZY chart.
Inference is everything
Work at OpenAI is being transformed by agents, in every department.
Across our entire company, people are using Codex to do work that is more complex, longer-running, and increasingly cross-functional.
Our internal usage offers an early look at how agentic tools may reshape work as they become more capable and broadly available.
There's a ton of other interesting stuff in the paper, which I encourage you to check out! There is also a great accompanying blog post that presents some of the key points with nice sleek visualizations.
Paper: https://cdn.openai.com/pdf/5d1e1489-21c0-43e4-9d42-f87efdbf0082/the-shift-to-agentic-ai-evidence-from-codex.pdf Blog: https://openai.com/index/how-agents-are-transforming-work/
@OpenAI @swyx new data for http://wtfhappened2025.com !
The paper looks at three populations: "individual" users, "organizational" users, and workers at OpenAI. As a share of ChatGPT + Codex output tokens, Codex usage is growing in all three, and at this point nearly 100% of output tokens at OpenAI are through Codex, not ChatGPT.
This is the sound of billions of jobs being destroyed or radically changed
tldr:
But at OpenAI, even folks in legal and recruiting are at this point primarily using Codex, not ChatGPT.
The transition to agentic AI is also happening across all seniority levels, although it is slightly less far along for those at both the lowest and highest seniority levels.
@drew_m_johnston @Alex_M_Richmond @_Chris_Ong @sonnytambe @RonnieChatterji Tagging some folks who might find this interesting! @emollick @erikbryn @johnjhorton @orgRem @sinanaral @SBenzell @AndreyFradkin @danielrock @alexolegimas @sebkrier
