As AI startups race for dominance, companies like Rilla and Browser-Use champion 70+ hour work weeks. We examine the origins of '996' culture, its migration to Western tech, and evidence challenging its productivity and health claims.
The recruitment page for New York-based Rilla bursts with images of energetic teams and slogans like "insane speed" and "customer obsession." Beneath promises of free meals and gym memberships lies a stark requirement: "working ~70 hrs/week in person." This explicit demand highlights a growing trend in AI startups—the normalization of extreme work hours reminiscent of China's controversial "996" culture (9am-9pm, six days weekly). As venture capital floods AI, founders argue such schedules are essential for competitive advantage. But productivity research and health data reveal significant drawbacks to this approach.
The 996 Blueprint: From China to Silicon Valley
Jack Ma, former chairman of Alibaba, called the 966 trend a "blessing"
The 996 model gained prominence in China's tech sector during the 2010s. Jack Ma famously endorsed it, stating in a 2019 blog post: "Being able to work 996 is a huge blessing... most successful people work 996 or more." JD.com founder Richard Liu reinforced this with his "Slackers are not my brothers!" email. However, backlash mounted as employees reported labor law violations and health issues. By 2021, Chinese authorities initiated a legal crackdown, silencing most proponents—though Baidu executive Qu Jing’s 2024 defense of relentless work hours ("I only care about results") briefly reignited the debate before costing her job.
Today, this culture thrives in U.S. AI startups. Rilla's Head of Growth, Will Gao, frames 70-hour weeks as a choice for the driven: "We look for people who are like Olympian athletes—obsessed, infinitely ambitious." He describes flexible but intense rhythms: "If I have a super idea, I'll work until 3am and roll in at noon." Similarly, Browser-Use founder Magnus Müller, operating from a Silicon Valley "hacker house," sees extended hours as unavoidable in AI development: "Giving AI new capabilities is super hard and competitive. Returns come when you immerse yourself deeply."
Will Gao says Rilla seeks candidates "like Olympian athletes"
The Productivity Paradox and Health Toll
Critics challenge the core assumption that prolonged hours yield proportional output. Deedy Das, partner at Menlo Ventures, argues: "Young founders wrongly equate hours worked with productivity. Forcing hustle alienates experienced workers and causes burnout." Data supports this:
- A WHO/ILO study found working >55 hours/week increases stroke risk by 35% and heart disease mortality by 17%, contributing to 745,000 global deaths in 2016.
- Productivity peaks around 40 hours/week, according to multiple studies. Beyond this, fatigue degrades output—Michigan State University research notes negligible differences between 50 and 70-hour weeks.
- Diminishing returns intensify with time: Output per hour drops sharply after sustained long shifts due to cognitive exhaustion.
Despite this, founders like Müller defend the model by comparing tech work to farming: "Farmers work 12+ hours daily, seven days a week. What we do is kindergarten by comparison."
Structural Pressures and Alternatives
AI's breakneck pace fuels this culture. Adrian Kinnersley, a tech recruiter, observes: "VC-funded AI startups race to market. They believe longer hours win." Yet alternatives exist. The UK's 2022 four-day week trial involving 61 companies maintained productivity while reducing stress and improving retention. Ben Wilmott of CIPD advocates "working smarter": "Improve management and adopt AI tools—don’t just extend hours."
Legally, the UK’s 48-hour workweek cap allows opt-outs, enabling 996-style schedules. However, Das notes this excludes diverse talent: "It alienates caregivers and seasoned professionals who deliver more efficiently."
Magnus Müller's team works in a shared "hacker house" environment
The Sustainability Question
For now, the 996 model persists where founders equate personal sacrifice with success—Das admits he’d "be shocked if a founder wasn’t working 70-80 hours." But health risks and productivity cliffs suggest it's a flawed strategy for scaling teams. As Kinnersley concludes: "Competing with a relaxed 35-hour week in AI is hard, but 80 hours is debatable." The AI gold rush may demand speed, but evidence indicates sustainable pacing could prevent a crash.
Sources:
Comments
Please log in or register to join the discussion