How OpenAI is using GPT-5 Codex to improve the AI tool itself

Ed Bayes, a designer on the Codex team, explained how the tool has changed his own workflow. Bayes said Codex now integrates with project management tools like Linear and communication platforms like Slack, allowing team members to delegate coding tasks directly to an AI agent. “You can add codecs, and now you can natively assign issues to codecs,” Bayes told Ars. “Codex is literally your teammate in the workplace.”

This integration means that when someone posts feedback in a Slack channel, they can tag Codex and ask them to fix the issue. The agent will create a pull request, and team members can review and iterate the changes through the same thread. “It’s basically anticipating this type of coworker and projecting it wherever you work,” Bayes said.

For Bayes, who works on visual design and interaction patterns for Codex’s interface, the tool has enabled him to contribute code directly instead of delegating specifications to engineers. “It kind of gives you more leverage. It enables you to do a lot more work and basically get more done,” he said. He said OpenAI designers now prototype features directly, using codecs to handle implementation details.

OpenAI’s approach treats Codex as “a junior developer” by Bayes, with the company hoping that over time it will become a senior developer. “If you were onboarding a junior developer, how would you onboard them? You give them a Slack account, you give them a Linear account,” Bayes said. “It’s not just this device that you go to the terminal for, but it’s something that also comes to you and sits on your team.”

Given this teammate approach, will there be anything left for humans to do? When asked, Embiricos drew a distinction between “vibe coding,” where developers accept AI-generated code without close review, and what AI researcher Simon Willison calls “vibe engineering,” where humans stay in the loop. “We see a lot of live engineering in our code base,” he said. “You ask Codex to work on it, maybe even ask for a plan first. Go back and forth, iterate on the plan, and then you’ll stick with the model and carefully review his code.”