When 16 Nobel Laureates Say We Must Act Now_

On July 13, 2026, something quietly unprecedented happened. Over 200 economists and technology leaders signed the same letter. It was published by Stanford's Digital Economy Lab under the title "We Must Act Now: A Statement on AI's Transformation of the Economy."

The signatories include Eric Schmidt. Reid Hoffman. Ben Bernanke. Joseph Stiglitz. Daron Acemoglu. Yoshua Bengio. Paul Krugman. Gita Gopinath, the First Deputy Managing Director of the IMF.

Sixteen Nobel laureates. The heads of the companies building the technology.

This is not outsiders warning about insiders. This is insiders warning about themselves.

Four sentences

The entire letter is 88 words. Four sentences. I want to quote it in full, because its brevity is the signal:

AI may become radically more powerful over the next 10 years. This could drive an unprecedented transformation of our economy, larger than the Industrial Revolution, but unfolding over a vastly shorter time frame. It could bring risks, including large-scale job displacement, as well as opportunities such as major gains in living standards. Economists, policymakers and technology leaders must act now to understand the economics of transformative AI and to build the incentives, guardrails, and institutions needed to steer AI in a direction that complements humans and benefits society.

The people who have spent their careers studying economic transitions could not find nuance worth adding. Four sentences. When economists agree this tightly, the nuance is gone.

The letter does not prescribe specific policies. It does not say "tax robots" or "ban AI" or "break up the labs." It says: understand what is coming, and build the institutions to steer it. That is the minimum bar, and 226 of the most qualified people on earth are saying we are currently below it.

The curve they are pointing at

The letter references a transformation "larger than the Industrial Revolution, but unfolding over a vastly shorter time frame." This is not a metaphor. There is data behind it.

A research group called METR has been measuring one variable: how large a task an AI can complete autonomously, translated into how long a skilled human would need for the same task. In 2023, the answer was a few minutes. In 2024, just under an hour. In 2025, it passed twelve hours of sustained professional work.

The length doubles roughly every three months.

METR's methodology is contestable — the tasks are drawn from domains where evaluation is verifiable, which means creative and interpersonal work are underrepresented, and the sample is narrow. But the trajectory converges across multiple independent measurement approaches. The direction is what matters, and the direction is unmistakable. That exclusion also means the curve is likely conservative: the domains it cannot measure are precisely where AI capability doublings will matter most for organizational readiness.

The Industrial Revolution took roughly 80 years to restructure an economy. The AI transformation, if the METR curve holds, compresses comparable restructuring into a window measured in months, not decades. Every regulatory cycle, every educational curriculum update, every organizational restructuring process operates on human timescales. The curve operates on its own.

The insiders' paradox

The letter is careful. It says "could" and "may." But read between the lines. Anthropic's founders left OpenAI over safety concerns, then built a competing lab with a safety mandate embedded in its charter. Multiple major AI labs have now published internal capability evaluation frameworks, telling governments: here is how we measure when our own systems become too dangerous. The AI lab CEOs who signed this letter do not benefit from guardrails around their own products. They signed anyway.

This is the insiders' paradox. The people with the strongest incentive to downplay the risk are the ones escalating it. Either they are all wrong, or they are right, and the risk is worse than public discourse reflects. I think they are right. The gap between their private assessment and the public conversation is itself one of the most dangerous things about this moment — institutional preparation follows public discourse, and the discourse is months behind the curve.

What this means

The letter asks us to build guardrails. Guardrails are not abstract. They are decisions made now, before the curve doubles again.

For an enterprise, this means four moves, sequenced:

First 30 days: Governance. Establish decision rights by process tier. Tier 1 covers regulated or high-stakes decisions — human in the loop, no exceptions. Tier 2 covers operational decisions where a human reviews AI output before action. Tier 3 covers low-risk, high-volume decisions where the system runs autonomously with audit logging. Who is accountable when an agent operating at each tier makes a decision that harms someone? That answer must exist before deployment, not after an incident.

Days 30–60: Workforce mapping. Map every role in the organization against the METR curve: fully automatable at today's capability, automatable at the next doubling, or requires judgment the curve does not yet capture. A simple starting point — score each role on two axes: how verifiable is its output (can a system check its own work?), and how much of the work involves novel context that was not present at training time. Roles that score low on both are automatable now. Roles that score high on one or both are the ones you invest in. This is a planning exercise, not a reduction exercise.

Days 30–60: Vendor risk. Run a separate assessment on every AI system already deployed. Most enterprises are locked into one or two hyperscaler stacks through proprietary API surfaces, model format incompatibility, and training-data gravity that makes migration prohibitive once dependency is established. Procurement needs AI-specific contracts with data residency, exit strategies, and model drift terms negotiated before that dependency is established, not after. The lock-in is both technical and contractual; an exit strategy that addresses only one of the two is not an exit strategy.

Ongoing: Board literacy. If the board cannot read the METR curve, it cannot make informed decisions about AI risk. If the board has not raised AI risk unprompted in the last quarter, that is your signal to schedule the session — not wait for it. A board-level AI literacy session covering (1) the METR curve and its implications for your specific industry, (2) your current AI deployment inventory and hyperscaler dependencies, and (3) a proposed cadence for quarterly AI risk review. Run it once, then make the quarterly review the mechanism that keeps it current.

The Industrial Revolution gave us 80 years to build labor unions, public education, and institutions that could redistribute the gains. The AI transformation will give us a fraction of that. The window for building those institutions is now, and it is narrow.

The question is no longer whether. The question is whether we build the institutional capacity to absorb this transformation in time. The 226 people who just told us to act are the ones who know the curve best. They gave us four sentences. We should not waste the time it bought us.