On July 20, 2026, the Brookings Institution published a commentary arguing that the generative AI productivity surge draws on human judgment earned before the tools existed, while the conditions producing that judgment drain quietly in the background.

The piece, a companion to a July 10, 2026 TechTank article titled Borrowed expertise, frames generative AI as analogous to the 1970s and 1980s pocket calculator adoption but structurally harder to manage. The author writes that an AI system "may be confidently wrong, and only the user's domain knowledge will catch the error. The user without that knowledge is worse off than a calculator user."

What does the calculator precedent look like?

When calculators reached classrooms, mathematics teachers adopted sequencing: arithmetic by hand first, calculator-free testing in early grades, and gradual introduction only after students could recognize a wrong answer. The commentary argues AI requires stricter sequencing than calculators because of three structural differences:

• AI outputs can be confidently wrong within its domain, whereas calculators give correct answers
• AI also selects the framing of the task, not just the operation
• The convenience differential is much larger — an entire essay versus seconds of arithmetic

What does this mean for credentialing?

For any credential whose value depends on the holder's underlying capacity, the credentialing institution must certify that the holder developed that capacity unaided, according to the commentary. This means AI-free assessment at credentialing stages in medicine, law, engineering, accounting, and increasingly research.

The author frames the stakes concretely: "A medical student who used AI to write all their clinical reasoning during training, and who is then licensed to practice based on those reports, is a public-health problem regardless of how productive AI made their education."

The commentary notes that no new legislation is required; accreditors and licensing boards need only take and defend a position. The piece flags that such decisions now happen "implicitly, accidentally, and inconsistently right now, by default" across institutions where the public cannot see them.

What is the research-funding question?

National Science Foundation, National Institutes of Health, and international funders should examine whether their instruments reward genuine novelty or AI-assisted recombination. The commentary cites a Science Advances study (doi 10.1126/sciadv.adn5290) showing AI can enhance individual creativity while homogenizing content across offers as an early warning of what an AI-saturated research literature will look like.

Proposed instruments include:

• Pilot review categories that explicitly weigh whether a contribution introduces a new frame versus extending an existing one
• Small-scale funding programs designed for risky, slow, idiosyncratic work

The commentary links this to documented seniority-biased hiring, citing Stanford Digital Economy work on six facts about AI employment effects, NBER working paper w31161, and SSRN paper 5425555 as evidence that firms reduce entry-level hiring.

What would pedagogical AI require?

The market will not produce scaffolding AI at scale because unit economics of frictionless answer delivery beat unit economics of productive struggle, the author argues. Government procurement can specify scaffold-not-answer systems, evaluated on user development rather than short-term satisfaction.

The piece identifies teacher training as the single highest-leverage education-policy investment available right now, and the one current education-policy conversations most reliably fail to take seriously. No specific funding figure or training capacity target is offered.

What is the time horizon?

The author warns that the cost of repair rises sharply once the pipeline has already failed. In 20 years, the reasoning goes, productivity gains could recede for the simple reason that no one remains to direct the tools — and the institutional knowledge required to repair that failure will itself have thinned.

The policy window the author identifies is the current procurement-decision period for the next decade of educational technology. The commentary closes: "We have built a remarkable technology. The question is whether we can use it without quietly disinheriting the generation that should have been its next set of masters."

The piece offers no measured cost figure for repair or quantified threshold for pipeline failure; its argument rests on a structural analogy rather than experimental evidence. Whether the sequencing model the author borrows from calculator-era math instruction survives contact with language-model capabilities remains untested at any disclosed scale.