The World Can't Keep Up With AI Labs

A new AI psychosis has taken hold, this time centered on coding agents. Unlike previous hype cycles, the numbers back it up: Anthropic's revenue has tripled since the start of 2026, outpacing growth rates of giants like Zoom during the pandemic and Google at IPO. OpenAI is posting similarly staggering figures. These aren't just vanity metrics—cash flow from agent subscriptions reveals genuine demand. Claude Code commits on GitHub doubled from 2% to 4% in January alone, and analyst Dylan Patel forecasts 20%+ by year-end. A $100 monthly subscription can deliver 10-30x ROI if it automates just 10% of a developer's routine work, making it a no-brainer for many teams.

Skeptics rightly point out weaknesses: the labs remain unprofitable, early adopters are risk-tolerant enthusiasts, and agents still randomly screw up. Reliability concerns, the need for junior engineers, and bottlenecks like requirements gathering and compliance prevent full replacement of humans. Yet the trajectory is clear. As Ilya Sutskever's internal meme "Feel the AGI" suggests, deep learning keeps conquering tasks once deemed impossible—from reasoning to image generation to Go. The evidence may not be conclusive, but the direction is undeniable: agentic systems are the first AI product people pay for at volume and regularly, and compute demand is growing faster than infrastructure can scale.