🚨Amazon's AI chief declares benchmarks are meaningless and real-world utility is what matters—a direct challenge to the leaderboard-obsessed AI industry that's spent years optimizing for synthetic tests rather than actual user value.

We're all going through collective withdrawal, and nobody's talking about it.

There's an article that's been living rent-free in my head for months. It's called "The Death of a Knowledge System" by Nicholas Mikaelson. Dense, long, uncomfortable reading. But it nails something everyone's feeling but can't articulate.

The premise: Our entire economy and institutions are built on propositional knowledge — you're valuable based on what you know. Experts, analysts, strategists. All of it undermined the moment AI became a query away from knowing anything.

But here's the part that keeps coming back: the addiction to certainty.

The withdrawal nobody's naming

Everything in this culture rewards the performance of certainty. It's how we manage fear, avoid vulnerability, earn social value. And AI just cut us off from the supply.

We used to know things. That was our value. Now the AI knows it faster, better, and cheaper. The knowledge we spent years accumulating? Commoditized overnight.

The anxiety you're feeling isn't about AI taking jobs. It's about losing the thing that made you feel certain about your place in the world.

Why learning is the only answer

Stanislas Dehaene defines learning as "the ability to adapt to unpredictable conditions as quickly as possible."

Read that again.

The answer to uncertainty isn't more knowledge. It's faster learning. The ability to adapt when the ground shifts under you — which is happening constantly now.

Learning how to learn is the new superpower. Not what you know. How quickly you can learn what you don't.

The three intelligences nobody's developing

Propositional knowledge (knowing facts) is dying. But there are three other types of intelligence that AI can't touch:

Procedural intelligence: Knowing how to do something. Not just theory — actual execution. The person who can ship beats the person who can explain.

Perspectival intelligence: Seeing the same situation from multiple angles. Understanding context, nuance, subtext. AI regresses to the mean. Humans see the outlier.

Participatory intelligence: Being present, engaged, and adaptive in real-time. This is the intelligence of live conversations, negotiations, and relationships. You can't automate presence.

These are the intelligences that matter now. And most people are still optimizing for the one that's already obsolete.

What this means

If you're anxious about AI, you're not anxious about the technology. You're anxious about losing certainty. The fix isn't resisting AI. It's building the capacity to learn faster than the technology changes.

If you're hiring, stop looking for people who know things. Start looking for people who learn things. The knowledge half-life is collapsing. The learning speed is what compounds.

If you're building with AI, the competitive advantage isn't the model you use. It's how fast you adapt when the model changes. Speed of learning beats depth of knowledge.

What's next

Expect a massive shift in how we value skills. Credentials and degrees are lagging indicators of past knowledge. Learning velocity is the leading indicator of future performance.

The people who win in the AI era won't be the ones who know the most. They'll be the ones who adapt the fastest.

BTW: The article mentions that we're collectively going through withdrawal from certainty. The people who accept that and build learning systems will have a 10-year head start on everyone still clinging to what they used to know.

The Recap: IBM CEO publicly declares that AI data center spending won't pay off, delivering a stunning rebuke from a major tech player while competitors like Microsoft, Meta, and Amazon double down on hundreds of billions in capex. The statement represents the first major crack in the industry consensus that massive infrastructure investment is the only path to AI dominance.

Unpacked:

Bottom line: IBM's public declaration forces the industry to confront an uncomfortable question—what if the infrastructure arms race is a massive misallocation of capital? The statement gives cover to other executives harboring private doubts about whether data center spending will generate returns that justify the investment, potentially reshaping AI strategy across the industry.

Anthropic acquires Bun, signaling AI companies are moving beyond models to own critical developer infrastructure and tooling.

Amazon launches on-premises 'AI Factories' with Nvidia, taking cloud AI infrastructure behind corporate firewalls to compete for enterprise control.

OpenAI declares 'code red' as Google closes the gap, revealing the AI race is far more competitive than the narrative suggests.

Zig quits GitHub over Microsoft's AI obsession, highlighting growing developer backlash against forced AI integration in core infrastructure.

Kalshi doubles valuation to $11B in under two months, showing prediction markets are emerging as the next fintech goldmine post-regulatory win.

Amazon previews AI coding agent that works autonomously for days, pushing toward the future where software writes itself without human supervision.

Microreactor startup Antares raises $96M for land, sea, and space nuclear power—AI's insatiable energy demands are driving a nuclear renaissance.

Google experimentally replaces real headlines with AI-generated clickbait, revealing how AI is degrading information quality in the search giant's core product.

Best of AI™ Team

Was this email forwarded to you? Sign up here.