Under the AI competition, the geniuses at OpenAI have also experienced the taste of 996, and worse, "this may become the norm in the future."

An AI researcher who previously worked at xAI and OpenAI recently announced his resignation due to exhaustion—this scene is becoming a microcosm of the Silicon Valley AI arms race. From “996” to extreme working hours of 16 hours per day and 100 hours per week, the culture of overwork is rapidly spreading in Silicon Valley’s AI community.

Researchers and industry insiders warn that this high-intensity work culture has taken root at top AI institutions like OpenAI and Anthropic, and is spreading to startups. Nathan Lambert, Senior Research Scientist at the Allen Institute for AI, bluntly stated, “This environment comes at a ‘human cost’, including losing time with family, health damage, and ultimately inevitable professional burnout.”

At the same time, the background of this overwork wave is far more complex than just personal choices: AI is reshaping the labor market, junior positions are shrinking dramatically, the pressure of layoffs and technical anxiety intermingle, causing practitioners to face the dual dilemma of ‘creating the future while worrying whether there’s still a place for themselves’. Economists warn that this may only be the beginning of a much larger labor transformation.

Starting from a resignation statement

What triggered this discussion was a widely circulated tweet in the AI community.

A researcher, hyhieu226, who had worked at both xAI and OpenAI, recently posted on social platform X, announcing his departure from OpenAI. He wrote that working at both companies was a “once-in-a-lifetime experience.” He met “the best people—not the best in AI, not the best in tech, but the best among humanity.” He helped create “highly intelligent entities that will genuinely improve human life,” and he took pride in this work.

However, he then switched tone: “But this intense work comes at a price. I can’t believe I would ever say this—I am exhausted. The mental health problems I used to scoff at are real, painful, terrifying, and dangerous.”

He said he would take a break from cutting-edge AI labs, return to Vietnam with his family, and seek healing. This restrained yet moving statement put the overwork issue in the AI industry center stage and resonated widely among AI practitioners.

Not just ‘996’—something even worse

Silicon Valley’s overwork culture isn’t new, but the era of AI has pushed it to new extremes.

According to Business Insider, AI research scientist Sebastian Raschka said in a conversation with podcast host Lex Fridman that the work culture he observed in Silicon Valley is not exactly “996” (9am–9pm, six days a week), “but it’s heading in that direction.” “You have to keep delivering, and it’s really hard,” he said.

According to Axios, citing the Wall Street Journal, the situation is even more extreme: top AI researchers now work up to 100 hours per week, and their self-deprecating work schedule is “0-0-2”—from midnight to midnight, with only two hours off on weekends.

According to devby.io, Mythril AI co-founder Sanju Lokuhitige revealed he works seven days a week, about 12 hours a day, “Sometimes I spend the whole day coding, completely lacking work-life balance.”

Another employee described living with the founder in a two-bedroom apartment, working from 9am to late night, surviving only on takeout food and short periods of sleep, and bluntly said: “This isn’t ‘996’ anymore, it’s 16-hour shifts.”

According to Axios, a San Francisco startup called Sonatic even explicitly asks employees to ‘be on duty seven days a week’ in its job postings, offering free housing and dating app subscriptions as compensation.

The company's 21-year-old CEO said, “When everyone is focused on the same mission, the odds of success improve,” and employees needn’t “worry” about housing, food, or socializing.

Driven by both competition and passion

This wave of overwork isn't simply caused by external pressures; internal motivations are also significant.

Raschka admitted that when he did AI research in academia, he didn’t work overtime because he was forced, but because he chose to. “It’s both passion and competition that drive this workaholic mindset,” he said.

Lambert also pointed out that OpenAI and Anthropic employees accept the high-pressure culture because they genuinely want to do the work, especially programmers.

Mintlify engineer Kyle Finken said it’s one of the most creative and productive periods in industry history, and many people work overtime not just out of pressure, but out of passion for the technology itself.

According to Axios, many of these top AI researchers are already millionaires, and some earn seven-figure salaries.

“Some of them are now multimillionaires, but a few said they don’t even have time to spend their wealth.”

However, the sustainability of this high-intensity model is being questioned. Computer science professor Cal Newport defines it as “pseudo-productivity”—where the metric is not work output, but the visible time employees spend at their desks or logged into Slack.

In his book “Slow Productivity,” he cited numerous examples from Newton to Jane Austen to argue “doing less is key to producing good work.”

Double blow: burnout and layoff anxiety

The costs of high-intensity work are manifesting, and a deteriorating external environment is compounding practitioners’ anxiety.

Lambert made clear that there is a “human cost” in this environment: losing time with family, shutting oneself from the outside world, and health problems. “You can only keep going for so long. People really burn out,” he said. Raschka’s cost was neck and back pain from skipping rest.

Meanwhile, layoff pressures are reshaping the power dynamics in the labor market. According to devby.io, global tech companies are expected to lay off about 250,000 people in 2025, with AI as a major contributing factor. Tech leaders like Zuckerberg and Musk have openly discussed the potential for AI to replace junior and mid-level engineers and continue pushing teams to improve “efficiency” and “resilience.”

Former Google, Microsoft and Salesforce executive coach Mike Robbins points out, “The balance of power has shifted: five years ago, if you were a software engineer, you could make demands.” Now, companies are much more powerful, as they’re less worried about losing employees.

According to devby.io, Indeed Hiring Lab data shows that junior tech jobs have declined by a third since 2022, while demand for specialists with five or more years of experience keeps rising. Mintlify engineer Finken admitted, “Many people are asking the same question: Will I still have a job in three years?”

This might just be the beginning of a bigger transformation

Economists’ warnings extend this debate well beyond Silicon Valley.

Devby.io reports a Stanford University study recorded “significant reductions in employment for early-career workers in sectors impacted by AI,” describing these industries as “canaries in the coal mine” for the broader economy.

Anthropic CEO Dario Amodei predicts that AI could eliminate up to half of junior office jobs within five years.

IMF Managing Director Kristalina Georgieva compared AI’s impact on the labor market to a “tsunami,” noting that up to 60% of jobs could be changed or disappear in some developed economies.

But in Silicon Valley, the race continues to accelerate. As Axios says: Overwork is back—at least until these engineers figure out how to replace themselves with robots.

Risk warning and disclaimerThe market is risky and investment requires caution. This article does not constitute personal investment advice, nor does it consider the individual investment goals, financial circumstances, or needs of specific users. Users should consider whether any opinions, viewpoints, or conclusions in this article suit their own situation. Investment based on this article is at your own risk.