OpenAI engineer quits, describing company as complete chaos

Former OpenAI engineer reveals company turmoil

Calvin French-Owen, once a key engineer behind OpenAI’s Codex project, has pulled back the curtain on life inside the company.

In a candid blog post, he describes OpenAI as an organization grappling with explosive growth, shifting directions at breakneck speed, and operating in “complete chaos.”

While not driven by personal conflict, his decision to leave stemmed from what he saw as a startup-like disorder consuming a company now valued in the multibillions.

The company grew threefold in just one year

French-Owen explained that during his single year at OpenAI, headcount surged from just over 1,000 to more than 3,000 employees.

With such rapid scaling, vital systems broke down, and reporting structures, communication, hiring processes, and project management struggled to keep pace.

Leadership teams scrambled to adapt as roles shifted constantly. The sudden growth overwhelmed a relatively small, mission-driven nonprofit just years before.

Chaos worsened by a lack of central planning

Unlike most large corporations, OpenAI doesn’t have a centralized planning committee or strict architecture oversight, according to French-Owen.

Individual teams drive decision-making by simply launching new initiatives. While this “bias for action” fuels innovation, it also breeds duplicate efforts.

He noted that seeing half a dozen separate libraries for queue management alone wasted work that adds to the operational mess.

Projects are often launched without oversight

At OpenAI, engineers were empowered to “just do things.” French-Owen describes how new projects often emerged from tiny teams without asking permission. If an idea showed promise, teams naturally coalesced around it.

While dynamic and empowering, this decentralized system sometimes led to unrelated teams accidentally working on overlapping problems, compounding internal confusion and resource waste.

Codex coding tool built in sleepless seven weeks

In one startling anecdote, French-Owen revealed how his 17-person Codex team, comprised of engineers, researchers, designers, and go-to-market staff, built and shipped the coding assistant in just seven weeks.

The pace was relentless, with extended hours and little sleep. But the results were stunning: simply launching the tool as a sidebar within ChatGPT led to instant adoption by millions of users.

Move-fast mentality echoes early Facebook culture

French-Owen compared OpenAI’s culture to Meta’s infamous early “move fast and break things” mantra. Many staff, including leadership, hail from Meta, reinforcing this startup-style speed.

However, he argues OpenAI still acts like a small startup despite being a massive firm with significant global scrutiny.

The result is a company caught between scrappy beginnings and corporate scale, unsure how to balance innovation with structure.

Slack remains the company’s primary tool

A surprising detail from French-Owen’s account is that nearly everything at OpenAI runs on Slack. He received just 10 emails in his entire year there.

While Slack supports fast communication, it creates coordination headaches for a company that is spread across thousands of employees.

Projects proliferate in endless channels, and vital conversations get lost, adding to the broader organizational disorder.

Azure dependence limits engineering flexibility

French-Owen revealed that OpenAI runs entirely on Microsoft’s Azure platform, which he found frustrating. He highlighted the limitations of Azure’s services compared to AWS tools like DynamoDB and BigQuery.

With Azure’s constraints, OpenAI has a habit of building in-house solutions where better cloud-native options might exist, increasing engineering complexity and consuming resources.

Costs driven by GPUs dwarf other expenses

One candid admission: OpenAI’s finances revolve around one massive line item: GPUs. French-Owen remarked that “everything else is a rounding error” compared to GPU costs.

Given the computational demands of training and running large AI models, hardware remains OpenAI’s primary expense, shaping much of the company’s infrastructure priorities and long-term strategy.

Leadership constantly shifts in rapid growth

French-Owen observed that nearly every leader he worked under did a completely different role two or three years earlier. As the company scaled, management churned, with people learning on the fly.

This constant reshuffling added to the instability, as new leaders scrambled to define processes while guiding critical teams forward.

Twitter feedback loop shapes company decisions

In an eye-opening revelation, French-Owen said OpenAI watches Twitter (now X) closely. If a tweet about OpenAI trends, staff are likely to notice and respond. He joked that a friend called OpenAI “a company that runs on Twitter vibes.”

This suggests that public sentiment on social media can directly influence OpenAI’s strategic thinking, potentially creating a risky feedback loop for a company building world-shaping technology.

Security measures tighten amid rising scrutiny

French-Owen mentioned that OpenAI has transformed into a more “serious” and “secretive” place, implementing heavy security protocols, including fingerprint scanners.

With massive commercial stakes and fierce competitors, the company has tightened internal access. He attributes this shift to increasing external scrutiny from regulators, rivals, and governments monitoring OpenAI’s every move.

Focus stays on user growth over enterprise sales

French-Owen hinted that OpenAI’s strategy leans toward user-driven adoption rather than top-down enterprise sales. Even for Codex, onboarding was designed for individual developers first.

This reflects OpenAI’s belief that viral, user-led growth is the key to AI adoption, contrasting with rivals targeting big enterprise clients directly. This “product-led” focus may explain the constant prioritization of features and releases.

Lack of coordinated planning fuels code duplication

A recurring pain point in French-Owen’s experience was the lack of central coordination. Teams frequently duplicated work without architectural oversight, creating multiple redundant libraries and components.

This slowed development and increased tech debt. Leadership has acknowledged these challenges, but solutions remain a work in progress amid the rapid scale-up.

According to French-Owen, without a clear central planning committee, most decisions fell to whichever team acted first.

OpenAI leadership changes direction frequently

Perhaps most disruptive of all, French-Owen observed how OpenAI frequently changes direction. Teams can pivot overnight as leadership reacts to new priorities, research discoveries, or public sentiment.

While this enables agility, it leaves employees whiplashed and compounds the internal chaos. Strategic consistency, he implied, remains elusive for a company moving at hyperspeed.

Curious where all this turbulence leads? Find out what’s behind ChatGPT’s latest flaw here.

Inside OpenAI feels like an experiment itself

In French-Owen’s parting words, OpenAI remains a company that hasn’t fully realized it’s a giant. It still behaves like a startup, with teams moving fast and breaking things even as billions ride on every decision.

From chaotic code to sleepless launches, it’s an experiment as much as a business, a glimpse into the messy engine powering the AI revolution.

Want to see how OpenAI’s growing pains are affecting real people? Learn more here.

What do you think about ChatGPT’s ex-employee revealing secrets about the workplace? Please share your thoughts and drop a comment.

Read More From This Brand:

• OpenAI says its next AI might be powerful enough to aid bioweapons
• The Pentagon just inked a $200M AI deal with OpenAI
• OpenAI Expands ChatGPT to Act as Your Daily AI Sidekick

Don’t forget to follow us for more exclusive content on MSN.

If you liked this story, you’ll LOVE our FREE emails. Join today and be the first to get stories like this one.

This slideshow was made with AI assistance and human editing.