AI data labeling and infrastructure company providing training data and evaluation for AI models.scale.com ↗
2 takes · first discussed Jul 11, 2025
Private company — no public price to score. We track what they said; valuation-mark tracking is on the roadmap.
Chamath argues that Meta's $15B bet on Scale AI is a wager on human-labeled data, which the Bitter Lesson and Grok 4's results show is the losing architectural approach — implying Scale AI's business model faces severe structural headwinds.
Keith (Guest) argues there is a very short half-life on human-labeled data businesses like Scale AI, as machines are already matching or surpassing human labeling quality in autonomous vehicle and other domains, meaning investors focused on revenue traction missed a fundamental obsolescence risk.
Both Chamath and guest Keith Rabois hold high-conviction bearish views on Scale AI, agreeing that human-labeled data has a very short half-life as a sustainable business. Chamath points specifically to Meta's $15B investment for a 49% stake in Scale AI as a bet now called into question by results like Grok 4, which suggest general computation can outperform human-labeling-heavy approaches. Rabois reinforces this, arguing that machines are already labeling as well as or better than humans in key domains like autonomous vehicles, and that investors were drawn in by near-term revenue traction without fully appreciating the durability risk. There is no disagreement between the two hosts — both see Scale AI's core business model as facing an imminent and potentially rapid decline.
Chamath argues that Meta's $15B bet on Scale AI is a wager on human-labeled data, which the Bitter Lesson and Grok 4's results show is the losing architectural approach — implying Scale AI's business model faces severe structural headwinds.
What has Llama been doing? They just spent $15 billion to buy 49% of Scale AI. That's exactly a bet on human knowledge.”23:20
Keith (Guest) argues there is a very short half-life on human-labeled data businesses like Scale AI, as machines are already matching or surpassing human labeling quality in autonomous vehicle and other domains, meaning investors focused on revenue traction missed a fundamental obsolescence risk.
I think there's a very short half-life on human-labeled data. And so everybody who's investing in these companies, they're just looking at revenue traction, really didn't understand that there may be a year, 2 years, 3 years max when”30:18