Back to the Transition Moonshot
A plan to carry working people across the AI gap, so the people in the middle of the change share in it instead of paying for it.
No American worker displaced by AI goes more than 24 months without equal or better income, and every one of them gets a real path forward: a new seat, a stake, or a company of their own.
Every major technology in history has worked. Railroads worked. The internet worked. AI will work. The damage has never come from the technology failing. It comes from the years between the breakthrough and the payoff, when the gains go to the people who own the change and the costs land on the people living through it. We call that stretch the Gap Years.
“It will be a little rough” is easy to say from the top. On the ground floor, a little rough has meant a lifetime. This paper treats the transition itself as the moonshot.
Start with a conservative assumption. Do not assume AI is better than a person at the job. Assume it is only as productive per hour. The difference is the hours.
| One full-time person | Hours a year |
|---|---|
| 40 hours × 52 weeks, paid | 2,080 |
| Less 10 holidays | −80 |
| Less 2 weeks vacation | −80 |
| Less 5 sick days | −40 |
| Hours actually worked | 1,880 |
| One AI seat: 3 shifts of 8 hours, 365 days | 8,760 |
| AI seat ÷ one person | 4.66× |
Illustrative. Real AI uptime, supervision time and error correction will lower the ratio, and the hour math does not apply the same way to physical work. Even at half this ratio, the surplus is large.
That surplus is the whole plan. An AI seat that is merely as good as a person produces several times the working hours. There is enough value in that surplus to keep the displaced person whole through the transition and still leave the company far ahead. The question is not whether the money exists. It is whether anyone sets aside a share of it on purpose.
When a company replaces a role with AI, the person in that seat keeps full pay for a set transition period, then steps down gradually, with first right to any new role the company creates. The AI seat’s surplus pays for it.
A small share of every AI seat’s labor savings goes into a portable account that follows the worker, not the employer. This covers what Pillar 1 cannot: companies that never hire anyone at all.
One person with good AI can now build what used to take a department. Give that same leverage to every displaced worker: the tools, the training and seed money to start a one-person company.
A public prize, in the moonshot tradition, for the company, city or team that moves the most displaced workers into equal or better income within 24 months, independently verified.
A rule that only protects existing employees has a hole in it. A founder with AI can build an entire business with no staff, so there is no seat to guarantee. If the protection only lives inside companies, the economy simply grows around it.
That is why the plan has two halves. Pillar 2 attaches the dividend to the AI seat instead of the payroll, so it covers companies that never hire. Pillar 3 turns the same fact into the opportunity: if one person can build the whole stack, then every displaced worker is a potential one-person company. The leverage that displaced them can become the leverage that rebuilds them.
Physical work and manufacturing are different. Robots cost more, scale slower and break in the real world. The plan treats them on their own timeline.
Every serious plan has to survive its critics. These are the hardest objections we know of, with our current answers.
Law already has a model. The WARN Act of 1988 requires employers with 100 or more workers to give 60 days’ notice before a plant closing or mass layoff. A Transition Act could build on that idea:
A discussion draft written to start a conversation, not legal advice and not a finished bill. It needs labor attorneys and economists before it goes anywhere near a legislature.
The change is coming either way. How much damage it does is a choice.