Ed Bayes, a designer from the Codex team, explained how the tool has transformed his workflow. He mentioned that Codex now connects with project management systems like Linear and communication tools such as Slack, enabling team members to delegate coding tasks directly to the AI agent. “You can integrate Codex, and you can essentially assign issues to Codex now,” Bayes conveyed to Ars. “Codex is truly a colleague in your workspace.”
This integration implies that when someone provides feedback in a Slack channel, they can mention Codex and request it to address the issue. The agent will generate a pull request, allowing team members to assess and refine the changes within the same discussion. “It’s essentially mimicking this type of coworker and appearing wherever you work,” Bayes noted.
For Bayes, who focuses on the visual design and interaction patterns for Codex’s interfaces, the tool has allowed him to directly contribute code instead of passing specifications to engineers. “It sort of grants you more leverage. It empowers you to work across the entire stack and essentially be able to accomplish more,” he remarked. He pointed out that designers at OpenAI now prototype features by directly building them, utilizing Codex to manage the implementation specifics.
The command line version of OpenAI codex running in a macOS terminal window.
OpenAI’s strategy positions Codex as what Bayes referred to as “a junior developer” that the organization intends to evolve into a senior developer over time. “If you were onboarding a junior developer, how would you bring them in? You would provide them with a Slack account, you would set them up with a Linear account,” Bayes said. “It’s not merely a tool that you access in the terminal, but it’s something that approaches you as well and integrates with your team.”
With this teammate perspective, will there still be roles left for humans? When posed this question, Embiricos differentiated between “vibe coding,” where developers accept AI-generated code without thorough review, and what AI researcher Simon Willison describes as “vibe engineering,” where humans remain involved. “We observe considerably more vibe engineering in our code base,” he stated. “You request Codex to work on that, perhaps you even solicit a plan first. Engage in dialogue, iterate on the plan, and then you partake in the process with the model and meticulously assess its code.”
