OpenAI’s annual developer conference on Monday was a spectacle of ambitious AI product launches, a app store for ChatGPT to a superb Video generation API who brought creative concepts to life. But for the companies and tech executives watching closely, the biggest announcement was silence general availability of the Codexthe company’s AI software engineer. This release marks a profound shift in the way software – and by extension, modern businesses – are built.
While other announcements have captured the public imagination, the release of Manuscriptsupercharged by a new specialized model and a suite of enterprise-grade toolsis the driving force behind the entire OpenAI vision. It’s the tool that builds the tools, the proven agent in a world teeming with agent potential, and the clearest articulation of the company’s strategy for winning the business.
THE general availability of the Codex moves it by one "research overview" to a fully supported product, complete with a new software development kit (SDK)A Slack integrationand administrative controls for security and surveillance. This transition declares that Codex is ready for critical work within the world’s largest companies.
"We think this is the best time in history to be a builder; it has never been faster to go from idea to product," said Sam Altman, CEO of OpenAI, during the opening speech presentation. "Building software took months or even years. You’ve seen that building with AI can now take a few minutes."
This acceleration is not theoretical. This is a reality born from internal use of OpenAI – a huge "dog food" effort that serves as the ultimate case study for enterprise clients.
Inside GPT-5-Codex: The AI Model That Codes Autonomously for Hours and Drives 70% Productivity Gains
At the heart of the Codex upgrade is GPT-5-Codexa version of OpenAI’s latest flagship model which has been "specially trained in Codex and agent coding." The new model is designed to work as a standalone teammate, going well beyond simple code completion.
"Personally, I like to think of it a bit like a human teammate," explained Tibo Sottiaux, OpenAI engineer, during a technical session on Codex. "You can associate a program on your computer with it, delegate to it, or, as you will see, give it a task without an explicit prompt."
This new model allows "adaptive thinking," allowing it to dynamically adjust the time and computational effort devoted to a task based on its complexity. For simple queries it is fast and efficient, but for complex refactoring projects it can run for hours.
An engineer during the technical session noted: "I’ve seen the GPT-5-Codex model run productively for over seven hours…in a marathon session." This ability to handle complex, long-running tasks represents a significant advancement beyond the simple, one-off interactions that define most AI coding assistants.
The results within OpenAI have been spectacular. The company reported that 92% of its technical staff now use Codex daily and that these engineers make 70% more pull requests (a measure of code contribution) each week. Consumption has increased tenfold since August.
"When we as a team see the stats it feels good," Sottiaux shared. "But what’s even better is being at lunch with someone who then says, “Hey, I use the Codex all the time.” Here’s a cool thing I do with it. Do you want to hear about it?"
How OpenAI uses Codex to create its own AI products and detect hundreds of bugs daily
Perhaps the most compelling argument for the importance of the Codex is that it is the foundational layer upon which OpenAI’s other flashy announcements have been built. During the DevDay Eventthe company showcased bespoke arcade games and a dynamic AI-powered website for the conference itself, all developed using Manuscript.
During a session, the engineers demonstrated how they built "Storyboard," a tailor-made creation tool for the film industry, in just 48 hours during an internal hackathon. "We decided to test Codex, our coding agent… we were sending tasks to Codex between meetings. We very easily reviewed and merged PRs in production, which Codex even allowed us to do from our phones," said Allison August, head of solutions engineering at OpenAI.
This reveals a critical insight: the rapid innovation showcased at DevDay is a direct result of the productivity flywheel created by Manuscript. AI is an essential part of the manufacturing process for all other AI products.
A key feature aimed at businesses is the new, more robust code review functionality. OpenAI said it "GPT-5-Codex specially trained to be excellent at ultra-thorough code review," allowing it to explore dependencies and validate a programmer’s intent against the actual implementation to find high-quality bugs. Internally, almost all OpenAI pull requests are now reviewed by Codex, catching hundreds of issues daily before they reach a human reviewer.
"It saves you time, you ship with more confidence," said Sottiaux. "There’s nothing worse than finding a bug after you’ve shipped the feature."
Why enterprise software teams choose Codex over GitHub Copilot for critical development
The maturation of Manuscript is central to OpenAI’s broader strategy to capture the enterprise market, a move critical to justifying its massive valuation and unprecedented compute spending. During a press conference, CEO Sam Altman confirmed the strategic change.
"The models are here now, and you should expect us to focus heavily on the truly winning companies with amazing products, from here," » Altman said during a private press conference.
OpenAI President and Co-Founder Greg Brockman immediately added: "And you can already see that with Codex, which I think has been incredibly successful and growing very quickly."
For technical decision-makers, the message is clear. While consumer-facing agents who book dinner reservations are still finding their footing, Manuscript is a proven corporate agent delivering a substantial return on investment today. Companies like Cisco have already deployed Codex in their engineering departments, reducing code review times by 50% and project deadlines from weeks to days.
With the new Codex SDKEnterprises can now integrate this agentic power directly into their own custom workflows, such as automating patches in a CI/CD pipeline or even building self-scaling applications. In a live demo, an engineer demonstrated a mobile app that updated its own user interface in real time based on a natural language prompt, all powered by the built-in Codex SDK.
While the launch of a application ecosystem in ChatGPT and the breathtaking visuals of the API Sister 2 legitimately generated titles, the general availability of the Codex marks a more fundamental and immediate transformation. It is the quiet yet powerful engine that will drive the next era of software development, transforming the abstract promise of AI-driven productivity into a tangible, deployable reality for today’s businesses.
