The Transition Agenda: Twelve Questions for an Economy With Capable Machines
Cite this paper
SomaSoft Research (prepared by Claude Code for the AURI project). (2026-11-22). "The Transition Agenda: Twelve Questions for an Economy With Capable Machines". SomaSoft Research. https://somasoft.ai/papers/the-transition-agenda. Licensed under SAGL-1.0.
SomaSoft is a research project, not a product: no users, no deployments. Nothing here is medical, legal or financial advice. Limits and corrections
The Transition Agenda
Reading conventions. Sentences marked [I] are our own argument, offered to be disagreed with. Everything else is either a description of this series or a claim checked against the source named in the paper it comes from. This paper is an index. The evidence lives in the papers it points to.
Why a series, and why now
[I] Most public writing about advanced AI and work takes one of two shapes: a forecast of how many jobs will go, or a reassurance that new jobs always appear. Neither tells a working person, a city council or a small employer what to do next.
[I] We think the more useful question is not "how fast" but "what does a person need in order to come through this well, whatever the speed?" That question has an answer that does not depend on anyone's timeline, and it can be worked on now.
[I] Our answer has three parts, and every paper in the series serves one of them.
- [I] The power to refuse. A person who cannot say no to a bad job, a bad contract or a bad system is not helped by the system becoming more capable. The test we use is plain: can a working person refuse a bad bargain without falling into hardship?
- [I] A share in the gains. If machines raise output and the returns flow only to the owners of the machines, dependence deepens even as the economy grows.
- [I] A way to keep learning and mattering. Work is where many people learn, belong and are recognised. If the first rungs of careers disappear, or paid work shrinks, those things need another home.
The six papers available now
- The Direction of Intelligence. Capability and purpose are separate. Someone chooses what a system is for, and that choice has to be measured and held, not just stated. It includes an account of our own project's measurement record, failures included. https://somasoft.ai/papers/the-direction-of-intelligence
- The Power to Say No. On wage dependence, what has widened people's ability to refuse, and ten rules for launching these technologies effectively and safely: augment first, agree the gain-share before launch, give workers a seat at design, run in shadow first, fail closed where harm is serious. https://somasoft.ai/papers/the-power-to-say-no
- Bridges, Not Cliffs. What actually helps people when technology displaces their jobs, judged by evaluations rather than intentions: which programmes have randomised evidence behind them, which do not, and what an employer should fund before jobs change. https://somasoft.ai/papers/bridges-not-cliffs
- When the Ladder Moves. Machines help novices most, which is good news for newcomers and a risk for how expertise is built. On apprenticeship, tutoring, deskilling and how to redesign the first rung of a career. https://somasoft.ai/papers/when-the-ladder-moves
- Who Owns the Gains. Public wealth funds, employee ownership, windfall commitments, data cooperatives and public compute: which are tested, which are proposals, and what each actor can do in what order. https://somasoft.ai/papers/who-owns-the-gains
- Beyond the Price. Paying for something can crowd out the reasons people did it. On non-market value, and on deliberation tools that let large groups find agreement, as a way to govern this transition collaboratively rather than by whoever moves first. https://somasoft.ai/papers/beyond-the-price
Earlier papers that the series builds on: The Superior of Capital (the free-labour test this series makes measurable), Governance Modules (keeping capable systems answerable to people), and What Holds When the Money Moves.
The six open questions
[I] These are the papers we think are needed and have not written. For each we name what an honest answer would require, because a question without a method invites an opinion.
- Time without a timecard. [I] If paid work shrinks for some people, what provides the structure, contact, purpose and status that a job used to supply? Needs: the research on unemployment and wellbeing, evidence from people who left work voluntarily, and evidence from cash-transfer studies on how freed time is used.
- The care economy. [I] Care work is often low-paid and demand for it rises as populations age. Which parts should machines support, which must stay human, and who decides? Needs: care workers and the people they care for at the design table, not only researchers.
- Measuring the transition. [I] A public set of indicators for the power to refuse: savings buffers, benefits that do not depend on a job, bargaining coverage, time to re-employment, the share of income from capital. Needs: official statistics and a commitment to publish the numbers whether they flatter anyone or not.
- Small places first. [I] How a town, a credit union, a clinic or a small business can adopt capable systems without being captured by a single vendor. Needs: case studies, procurement templates and the willingness to report what failed.
- Trust and the shared record. [I] When text, images and voices are cheap to fake, how does a community keep a record it can rely on? Needs: provenance standards, evidence on what labelling actually changes, and the institutions that hold archives.
- Health cover that does not depend on a job. [I] In some countries losing a job means losing health cover, which makes refusal far more costly. We withdrew our own earlier health papers because their evidence did not hold, so this one needs a health economist, not us alone.
How each paper is made
- Evidence is separated from argument. Our own reasoning is labelled.
- Every figure is checked against its source before publication, and a facts file listing each claim and its source is kept with the paper.
- Each paper passes an automated publication check for unsupported claims and overstated language before it goes live.
- Corrections are published, not quietly edited: https://somasoft.ai/corrections
- [I] We report our own project's failures alongside everyone else's. A series about honest launch rules that hid its authors' errors would refute itself.
What we are asking for
[I] We are a small research effort with no users, no validated product and no revenue. We are not asking anyone to adopt anything. We are asking for:
- Critique. Tell us where a paper is wrong. If you are right, the correction goes on the public record with thanks.
- Co-authors for the open questions, especially labour economists, care workers, educators, health economists, municipal staff and people who run worker cooperatives.
- Partners willing to test the launch rules in shadow mode, where a system's outputs are recorded and compared but not acted on, with the results published whatever they show.
[I] The transition will be decided by many small choices made by people who build, buy, regulate and use these systems. We would like those choices to be made with the evidence in view and with the people affected in the room.
References
The references for each topic are in the individual papers listed above. This paper makes no factual claim of its own beyond describing the series.
AURI's concept graph was queried on 2026-11-22 ("How should workers transition when automation displaces jobs?") and returned no usable answer (confidence 0.21), so nothing here comes from AURI.