Inside OpenAI: What It's Really Like to Work at the World's Most Watched AI Company

July 17 2025 16:07
A former engineer's candid account reveals the culture of secrecy, speed, and relentless ambition driving artificial intelligence's most influential company

Calvin French-Owen walked into OpenAI in May 2024 as a seasoned entrepreneur who had co-founded and sold Segment, a billion-dollar data infrastructure company. He left 14 months later with a story that offers the clearest window yet into one of the most scrutinized organizations in modern technology.

His reflections, shared publicly after his departure in June 2025, paint a picture of a company operating at breakneck speed while carrying the weight of potentially reshaping civilization itself. It's a place where the workforce expanded from 770 employees in November 2023 to 3,531 by September 2024, where everything runs on Slack instead of email, and where a single feature can cost as much as an entire startup's infrastructure.

"OpenAI is perhaps the most frighteningly ambitious organization I've ever seen," French-Owen writes. "You might think that having one of the top consumer apps on the planet might be enough, but there's a desire to compete across dozens of arenas."


When French-Owen joined OpenAI, he found himself in the top 30 percent of employees by tenure within just one year. That staggering turnover rate reflects not just the company's explosive growth, but also the unique pressures of working at the epicenter of the AI revolution.

Unlike traditional tech companies with their endless email chains and bureaucratic processes, OpenAI operates entirely through Slack. "I maybe received around 10 emails in my entire time there," French-Owen recalls. This hyper-connected communication style creates both opportunities and challenges. For the organized, it enables rapid decision-making. For others, it becomes a constant source of distraction.

The company's structure defies conventional corporate wisdom. There's no traditional roadmap, no quarterly planning cycles, and no grand master plan. Instead, OpenAI operates on what French-Owen describes as a "bottoms-up" culture where good ideas can emerge from anywhere. Researchers function like "mini-executives," pursuing their own projects and seeing what bears fruit.

This approach has created what might be the most meritocratic environment in Silicon Valley. "Leaders in the company are promoted primarily based upon their ability to have good ideas and then execute upon them," French-Owen observes. "Many leaders who were incredibly competent weren't very good at things like presenting at all-hands or political maneuvering. That matters less at OpenAI than it might at other companies."


Working at OpenAI means operating under unprecedented scrutiny. French-Owen describes regularly seeing news stories about the company break in the press before they were announced internally. Twitter bots monitor the company's every move, and employees can't discuss their work in detail with anyone outside the organization.

This level of attention creates a "very secretive place" with multiple Slack workspaces operating under different security permissions. Revenue and burn numbers are closely guarded secrets, and the company maintains what French-Owen calls a "more serious" atmosphere than typical Silicon Valley startups.

The stakes feel genuinely high. On one hand, there's the goal of building artificial general intelligence. On the other, there's the responsibility of serving hundreds of millions of users who rely on ChatGPT for everything from medical advice to therapy. Meanwhile, the company competes in "the biggest arena in the world" against Meta, Google, and Anthropic, with major world governments watching every move.



Beneath the public fascination and media coverage lies a company grappling with the practical challenges of scaling revolutionary technology. OpenAI runs on a "giant monorepo" written mostly in Python, creating what French-Owen describes as "strange-looking code" that ranges from enterprise-grade libraries built by Google veterans to throwaway Jupyter notebooks from newly-minted PhDs.

The company's infrastructure choices reflect both its rapid growth and its unique position in the market. Everything runs on Microsoft Azure, though French-Owen notes that only three services feel truly reliable: Azure Kubernetes Service, CosmosDB, and BlobStore. This has led to significant in-house development, with OpenAI reimplementing many core infrastructure components.

The scale of computation required is mind-boggling. A single niche feature built for the Codex product consumed the same GPU resources as Segment's entire infrastructure. "Nearly everything is a rounding error compared to GPU cost," French-Owen explains, highlighting how the economics of AI development differ fundamentally from traditional software.


One of the most interesting insights from French-Owen's account is OpenAI's strong connection to Meta. The company experienced a series of high-profile resignations in 2024, including safety researchers, but has maintained what French-Owen calls a "very significant Meta to OpenAI pipeline" in engineering talent.

This connection makes sense when you consider the similarities between the companies. Both built blockbuster consumer applications, both operate with nascent infrastructure, and both prioritize moving quickly over perfect planning. The Meta influence shows up in OpenAI's infrastructure choices, with in-house implementations of systems that mirror Meta's internal tools.

"In many ways, OpenAI resembles early Meta," French-Owen observes. The comparison extends beyond technology to culture, with both companies maintaining a "strong bias for action" where teams can form quickly around promising ideas without waiting for permission.


Perhaps the most revealing part of French-Owen's account is his description of launching Codex, OpenAI's AI coding agent. The tool, which handles multiple software engineering tasks simultaneously, was built from start to finish in just seven weeks.

The sprint exemplifies everything unique about OpenAI's culture. It required "the hardest I've worked in nearly a decade," with most nights extending until 11 PM or midnight, early morning starts at 7 AM, and weekend work. The team included eight engineers, four researchers, two designers, two go-to-market specialists, and one product manager. Everyone was senior, everyone was self-directed, and everyone understood the urgency.

The scope was staggering for such a compressed timeline. The team built a container runtime, optimized repository downloading, fine-tuned a custom model for code editing, handled complex git operations, introduced an entirely new user interface, enabled internet access, and created what French-Owen describes as "a product that was generally a delight to use."

The night before launch, five team members stayed up until 4 AM deploying the main monolith, then returned to the office at 8 AM for the launch announcement. The results speak for themselves: Codex allows developers to simultaneously deploy multiple agents to independently handle coding tasks, and in its first 53 days, it generated an estimated 630,000 pull requests.


One of the most surprising aspects of OpenAI's culture is its relationship with social media. French-Owen describes a company that "pays a lot of attention to Twitter," where viral tweets about the company regularly influence internal discussions. A friend joked that "this company runs on Twitter vibes," and French-Owen suggests that assessment isn't entirely wrong.

For a consumer-focused company, this social media sensitivity makes strategic sense. While OpenAI certainly tracks traditional metrics like user growth and retention, the cultural zeitgeist around AI development plays an equally important role in shaping product decisions. In an industry where public perception can shift overnight, staying tuned to social media conversations becomes a competitive necessity.

This attention to external perception extends to the company's approach to safety and research transparency. French-Owen notes that safety work receives more focus than critics might expect, though much of it remains unpublished. The company faces constant pressure to balance openness with competitive advantage, transparency with security, and rapid progress with responsible development.


French-Owen's experience at OpenAI offers insights that extend far beyond one company's culture. His description of Codex as treating AI agents "like a coworker" where users "send messages to the agent, it gets some time to do its work, and then it comes back with a PR" points toward a fundamental shift in how work gets done.

The betting assumption behind Codex is that we're moving toward a world where human developers delegate tasks to AI agents rather than writing code directly. This represents what some observers call "a platform shift" where "software builds itself, strategies are drafted by algorithms, and value is created by those who orchestrate rather than operate."

French-Owen believes this transformation is inevitable: "Over the long arc of time, I do believe most programming will look more like Codex." The question isn't whether AI will change how we work, but how quickly and in what ways.


French-Owen's OpenAI experience offers several lessons for anyone trying to understand the future of technology and work:

• Speed remains the ultimate competitive advantage. In a rapidly evolving field, the ability to build and ship quickly matters more than perfect planning or extensive documentation.
• Cultural cohesion becomes harder at scale. OpenAI's growth from 1,000 to 3,000 employees in one year created challenges that even the most innovative companies struggle to solve.
• Meritocracy can coexist with chaos. Despite the lack of traditional structure, good ideas and strong execution still rise to the top.
• The right team can accomplish extraordinary things. The Codex launch demonstrates what's possible when senior people work together with clear goals and unlimited resources.
• Public scrutiny changes everything. Operating under constant media attention requires different approaches to communication, decision-making, and talent management.

French-Owen concludes his reflections with a prediction: "The path to AGI is a three-horse race right now: OpenAI, Anthropic, and Google. Each of these organizations are going to take a different path to get there based upon their DNA (consumer vs business vs rock-solid-infra + data)."

This assessment reflects the unique position these companies occupy in the AI landscape. OpenAI's consumer focus, Anthropic's business orientation, and Google's infrastructure advantages represent different approaches to the same ultimate goal. Understanding these differences helps explain not just competitive dynamics, but also the varying workplace cultures and strategic priorities of each organization.

For French-Owen, the experience was transformative. Despite the challenges of giving up entrepreneurial freedom to become "a much smaller piece of a much larger machine," he calls joining OpenAI "one of the best moves I've ever made." His advice to struggling founders is direct: either find new ways to take more shots on goal, or "go join one of the big labs."

The window he's provided into OpenAI reveals a company that embodies both the promise and the peril of our AI-driven future. It's a place where the next breakthrough might come from anywhere, where the stakes couldn't be higher, and where the old rules of corporate America simply don't apply. Whether that's sustainable remains to be seen, but for now, it's working.

As French-Owen puts it: "Right now is an incredible time to build. But it's also an incredible time to peer into where the future is headed."

The question for the rest of us is whether we're ready for what's coming next.