Why the AI Revolution Won't Happen Overnight

Matt Shumer sits at his desk, typing what would become a viral manifesto. "Something Big Is Happening," he declares to his followers on X, comparing our AI moment to the early days of Covid. The OthersideAI CEO argues that artificial intelligence has crossed a critical threshold—from helpful assistant to general cognitive substitute. Meanwhile, Mrinank Sharma quietly resigns from Anthropic's safety team, leaving behind a cryptic warning that "the world is in peril."

The tech world is buzzing. Prediction markets are heating up. Some are starting to panic. But here's the thing about revolutions: they rarely happen as fast as the revolutionaries predict.

The Gap Between Demo and Deployment

Yes, AI models are becoming impressively capable. Google DeepMind'sDemis Hassabis acknowledges we might need just one or two more AlphaGo-level breakthroughs to reach artificial general intelligence. But there's a vast canyon between "the technology works" and "everything changes tomorrow."

Consider the humble factory. When electricity arrived in the late 1800s, it didn't instantly transform manufacturing. It took decades for businesses to redesign their operations around this revolutionary power source. The internet didn't reshape retail overnight either—Amazon started as a bookstore in 1994, but e-commerce didn't truly explode until the 2000s.

Today, AI adoption covers fewer than one in five US business establishments. Rolling out AI across large, regulated, risk-averse institutions requires massive complementary investments: data infrastructure, process redesign, compliance frameworks, worker retraining. Economists call this the productivity J-curve—early spending can actually depress measured output before the gains become visible.

When Richer Doesn't Mean Busier

Let's grant the optimists their assumption about rapidly advancing AI capabilities. Even then, economic output doesn't explode overnight. History shows us something counterintuitive: richer societies often choose more leisure, not more work.

Economist Dietrich Vollrath points out that higher productivity doesn't automatically translate into faster growth if people respond by working less. Think shorter workweeks, earlier retirements, longer vacations. Welfare might rise substantially while headline GDP growth stays relatively modest.

This isn't a bug—it's a feature. As societies become wealthier, they tend to value time over additional income. The 40-hour workweek wasn't handed down from economic heaven; it was a choice made by increasingly prosperous societies.

The Baumol Bottleneck

Even if AI makes some services dramatically cheaper, demand doesn't expand infinitely. Spending tends to shift toward sectors that resist automation—healthcare, education, live entertainment, personal services—where output remains tightly tied to human time and presence.

This is the famous "Baumol effect" or "cost disease." As wages rise economy-wide, labor-intensive sectors with weak productivity growth claim larger shares of income. A string quartet still needs four musicians and roughly the same amount of time to play a Mozart piece as it did in 1785. Your doctor still needs to examine you personally. Your child's teacher still needs to engage with students individually.

The result? Even spectacular AI gains in some sectors may yield only moderate growth in overall productivity.

The Narrowest Pipe Problem

Economist Charles Jones explains that in any system built from complementary pieces, the narrowest pipe determines the flow. AI might accelerate coding, research, and content creation dramatically. But if energy infrastructure, physical capital, regulatory approval processes, or human decision-making crawl along at ordinary speeds, those become the binding constraints.

Think about autonomous vehicles. The technology is largely there, but deployment remains slow due to regulatory frameworks, insurance questions, infrastructure needs, and public acceptance. The AI revolution faces similar multi-dimensional constraints.

The Adaptive Economy Reality

This doesn't mean change isn't coming—it absolutely is. Shumer's advice to embrace capable AI tools now and weave them into daily work seems prudent. The question isn't whether AI will transform the economy, but how quickly and smoothly that transformation will unfold.

Economies are adaptive, complex systems that create what economist Cesar Hidalgo calls "crystals of imagination"—physical objects that embody accumulated knowledge. When they change, they adjust through gradual reorganization and reallocation, not sudden collapse or instant takeoff.

The $350 billion valuation that Anthropic is chasing reflects genuine optimism about AI's potential. But markets also price in realistic timelines for deployment and adoption.