Inside OpenAI’s Quest to Compete with Claude Code

Katy Shi, a researcher focused on Codex’s behavior at OpenAI, notes that while some people liken its default personality to “dry bread,” many have grown to appreciate its less sycophantic manner. “A lot of engineering work is about being able to take critical feedback without interpreting it as mean,” Shi remarks. Several major companies have also adopted Codex. “The fact that ChatGPT is synonymous with AI gives us a massive advantage in the B2B market,” states Fidji Simo, OpenAI’s CEO of applications. “Businesses prefer using technologies their employees are already familiar with.” OpenAI’s strategy for promoting Codex hinges on combining it with ChatGPT and other OpenAI offerings, Simo added.

Cisco’s president and chief product officer, Jeetu Patel, mentions that he has reassured employees not to fret over the costs of using Codex, emphasizing the importance of comfort with the tool. When employees express concerns about job security due to these tools, Patel explains, “what we have to tell our people is no, but I guarantee you’ll lose your job if you don’t use them, because you won’t be relevant. So you’re going to be out.”

Currently, anxiety concerning AI coding agents has extended well beyond Silicon Valley. The Wall Street Journal attributed a $1 trillion tech stock sell-off last month to Claude Code, as investors feared software might soon become entirely outdated. Shortly thereafter, IBM experienced its most significant stock drop in 25 years following Anthropic’s announcement that Claude Code could modernize legacy systems running COBOL, prevalent in IBM environments. OpenAI has diligently worked to integrate its AI coding agent into societal discussions, investing millions in a Super Bowl commercial focusing on Codex rather than ChatGPT.

At the Mission Bay temple, there’s no need for a sales pitch on Codex. Many OpenAI engineers I conversed with indicated they rarely write code anymore. They spend their days interacting with Codex. Occasionally, they collaborate in groups.

At headquarters, I attended a Codex hackathon where around 100 engineers gathered in a large room. Each team had four hours to create the best demo utilizing Codex. A senior OpenAI leader facilitated the event, speaking team names into a microphone while turning away from his laptop. Team representatives nervously approached the podium to present their AI projects, their voices often trembling. Winners were awarded Patagonia backpacks.

Many projects showcased were both developed using Codex and designed to enhance engineers’ understanding of it. One team crafted a tool to summarize Slack messages into weekly reports. Another team produced an AI-generated Wikipedia-style guide for internal OpenAI services. Tasks that would have taken days or weeks could now be completed in a few hours.

As I was leaving, I crossed paths with Kevin Weil, the former Instagram executive now leading OpenAI for Science, the company’s new division focused on building AI tools for researchers. He shared that Codex had been working on some projects for him overnight, which he planned to check on in the morning. This has become a common routine for Weil and hundreds of other employees. One of OpenAI’s objectives for 2026 is to develop an automated intern dedicated to researching AI.