Artificial intelligence is the massive disruption of our generation — not a candidate for the title, the holder of it. Steam mechanized muscle; electricity distributed it; computing automated arithmetic and records. AI is the first technology that performs cognition — reads, writes, diagnoses, designs, negotiates — while improving continuously and reproducing at almost no cost. Whatever else is uncertain, this is not: the relationship between work, income, and national prosperity is being renegotiated faster than the institutions that depend on it were designed to move.
The renegotiation has begun where careers begin. Since late 2022, employment in the most AI-exposed occupations has fallen roughly six percent for young workers while rising six to nine percent for older workers in the same roles — the ladder is losing its bottom rungs while the upper rungs hold. The 2026 International AI Safety Report finds the same concentration: aggregate employment steady, junior hiring bending. On freelance platforms, writing work and earnings fell within months of ChatGPT's release while demand for machine-learning skills jumped by a quarter. Employer surveys still split — many firms report AI increasing entry-level hiring where it augments rather than replaces — and that split is the honest headline: compression and creation at once, arriving faster than any prior transition.
History counsels against both panics on offer. Agriculture went from employing four in ten Americans to fewer than one in fifty, and the country got richer; the typing pool vanished into software engineering, an occupation no one in 1950 could name. New work will come — history is nearly unanimous. What history does not promise is that the fiscal architecture survives the transit. Most federal revenue rides on wages. Health insurance rides on employment. Retirement rides on the employer. Training happens once, before the first career, in a world that will demand several. Those assumptions built the American century; they are precisely what abundant intelligence erodes.
The Accord's wager is that this question, not any jobs forecast, is the one that decides the era: how does a republic finance itself when labor is no longer the dominant source of national income?
The New American Accord answers with design rather than prediction — and, unusually for a policy document, it shows its work. The Accord's revenue stands on three registers: broad contribution from all compensation, honest pricing of harms, and settlement of the obligations that accumulate inside great fortunes. When AI shifts income from payroll to capital — to software, models, and the machines' owners — the third register catches at settlement what the first loses in withholding. The system was drawn so that the money is taxed where the economy actually puts it.
That claim would be rhetoric if it weren't testable, so we tested it, in public. The Accord's fiscal model is published, and its stress laboratory lets anyone run the experiment: erode labor compensation year after year and watch the fifty-year debt-retirement path respond. The results are exact. The statutory path absorbs, automatically and without a single act of Congress, labor erosion approaching half a percentage point per year sustained for decades — four to five times the entire average labor-share decline the United States has recorded since 1970 — before debt-retirement slips even ten years. Push far beyond that and the system still refuses to spiral: at the outer edge the automatic correctors reach their legal limits, the debt plateaus, and the decision returns to Congress — a system whose worst case is a scheduled public choice rather than a crisis. And one dependency is now settled by model output rather than assumption: if the job market collapses, zeroing immigration entirely — workers, growth, and the associated revenue — moves the debt-retirement date by nothing. The Accord's solvency asks nothing of anyone's border politics.
Beneath the revenue question sits the family question, and here the Accord's seemingly unrelated programs resolve into a single design. Healthcare that ignores your employer, so a career change is not a medical event. A public account that moves money instantly, so a transition month is not a payday-loan month. A savings account opened at birth, so the generation entering the turbulence holds capital, not just résumés. A skills account that accrues across life — a blunt and deliberately slow instrument, candidly, meant to make trade school, certification, and community college ordinary purchases rather than crises — because the occupations of 2050 cannot be named in 2030, and flexibility outlasts any forecast. Call these what an engineer would: shock absorbers. A car carries them because roads carry potholes, and the point is that a bump costs the passengers a jolt instead of the vehicle. A family that can absorb a career interruption without losing the house, the coverage, or a decade of savings is a family that can take the next job, in the next industry, in the next town — which is exactly the adaptive capacity a fast transition pays for.
A word to the people who own the machines — offered as to co-authors, because that is what the Accord's terms make of them. The great fortunes of the AI era will be built on the most expensive inheritance ever assembled: the research universities, the courts that enforce a contract, the grid, the schooled customers, the stable currency. Henry Ford grasped the domestic version of the bargain — the workers had to be able to buy the cars. The abundant-intelligence version is larger: machines can make nearly everything except customers capable of buying what machines make, and citizens invested enough to keep the system legitimate. The Accord asks the winners of this transition to carry a settlement proportionate to what the country made possible, and offers in return the only asset no fortune can purchase privately — a nation that remains worth being consequential in. The most farsighted holders of wealth have said as much themselves for years. The Accord is the first architecture that takes them at their word.
Two weeks ago the United States turned two hundred fifty. The founders were the original long-horizon stewards — they capitalized a set of institutions on a wager that ordinary people, given a floor of rights and a frame of law, would out-create every alternative arrangement of society. Two and a half centuries of compounding have paid that wager beyond any founder's arithmetic. Abundant intelligence is the second quarter-millennium's opening test of the same proposition: whether the returns of an extraordinary technology flow through the whole population that made it possible, or pool at the top of it. The Accord's purpose is to make the founders' wager pay again — to ensure that when intelligence becomes abundant, opportunity becomes abundant with it.
Sources to link: Stanford/ADP-line study on AI-exposed employment by age (−6% young vs +6–9% older, late 2022–mid 2025); International AI Safety Report 2026 (junior-worker concentration; freelance writing −2% listings / −5.2% earnings; ML demand +24%); Strada Institute 2026 employer survey (augment-vs-automate split); BLS labor-share series (post-1970 decline calibration); the Accord's Stress lab at /scoring/stress (v3 thresholds, which include the ruled surplus-symmetry relief: retirement slips +10 years at 0.46%/yr erosion, +25 years at 0.53%/yr, plateau at the corridor cap past 0.64%/yr; World A coupling shown at the ruled 0.5 with 0.3/0.7 sensitivity).