Skip to the council result

The Discontinuity Thesis · Solutions Council

What should Britain and Europe actually do?

Eighteen model proposals, one controlled question, two blind feasibility reviews and a narrower answer than any single model produced.

Editorial conclusion

Build the rails before promising the dividend

The durable answer is a staged policy runway. Each step needs less faith in the thesis than the one after it.

01 · Implement now

Measure agency. Protect savings. Insure earnings.

Track household outcomes rather than an AI share of work. Add cause-neutral earnings insurance and remove savings penalties that destroy resilience.

02 · Build now

Make public risk earn a public return

Take transparent, market-valued upside only where grants, guarantees, concessional finance or genuine co-investment put public capital at risk.

03 · Pilot

Test a funded Agency Credit

Use named, broad tax bases and existing payment rails. Keep investment capital, cost savings and recurring tax receipts on separate ledgers.

04 · Keep dormant

Scale cash only when four gates open

Persistent adverse outcomes, 110% recurring funding cover, delivery capacity and an essential-cost test must all pass before expansion.

Reject as drafted: AI-output taxes, estimated foreign-compute levies, speculative fund returns, uncosted universal dividends and EU-level fiscal shortcuts that require powers the EU does not possess.

New independent evidence

The UN report strengthens the contingency case, not the certainty claim

The July 2026 preliminary report treats distribution as unresolved, finds mixed and institution-dependent labour effects, warns that weak effects today do not rule out larger effects after adoption lags, and calls for better measurement of productivity, work, wages and distribution.

It is scientific and non-prescriptive. It supports monitoring, competition and fiscal preparedness; it does not prove inevitable displacement or endorse a particular levy or dividend.

Read the official preliminary report
How the evidence should change policy
Observed effectsMixed
Future scaleUncertain
Distribution riskMaterial
Best responseContingent

Blind feasibility result

Terra and Sol lead. Confidence is not feasibility.

Scores average an independent legal-fiscal review and an independent political-durability review. Open Why? beside any verdict to see the actual reason. “Viable core” means a usable core survives if its conditions are met; “fatal as drafted” means the package cannot operate as written, not that every component is useless.

Blind model ranking out of 100
Model and proposalLegalPoliticalCouncil decision
1 GPT-5.6 Terra (model ID: gpt-5.6-terra)The Agency Circuit Compact 929191.5 Viable core
Why?
Viable core — strongest overall

The proposal is unusually clear about what current institutions can fund and separates evidence, fiscal permission and delivery. Its three-ledger waterfall and two-key trigger prevent speculative assets or cost savings from being counted as recurring cash.

Condition: The first Agency Floor remains deliberately unspecified, so Treasury delay and an inadequate response to a fast shock remain risks.

See strengths and repair
2 GPT-5.6 Sol (model ID: gpt-5.6-sol)The Agency Dividend Compact 909090.0 Viable core
Why?
Viable core — funding must come first

The Agency Credit is calculated backwards from an explicit fiscal envelope, while a two-ledger rule separates tax-funded payments from realised investment dividends. Its gates stop missing micro-arithmetic from turning into an automatic liability.

Condition: Even a 0.5% of GDP baseline needs a named, distributionally modelled property, capital-income and consumption package.

See strengths and repair
3 Codex (GPT-5)The Civic Agency Compact 898587.0 Viable core
Why?
Viable core — baseline still needs funding

The two-key structural and fiscal trigger is administrable, and the rent firewall stops payments scaling when housing or essential costs absorb the benefit. It avoids AI-attribution taxes and gives both workers and innovators a plausible coalition.

Condition: The illustrative initial payment still lacks a complete, named first-Budget revenue package and detailed earnings-insurance costing.

See strengths and repair
4 GPT-5.6 Luna (model ID: gpt-5.6-luna)The Agency Compact: Adopt Freely, Share the Upside 838081.5 Viable core
Why?
Viable core — trigger needs strengthening

The proposal correctly leaves income provision to national governments while using EU powers for rights, competition and procurement. It separates immediate cash, public capital and workplace due process and makes payment conditional on fiscal scoring.

Condition: Its two-quarter disposable-income trigger is too noisy, and the initial payment has no explicit ceiling or named revenue schedule.

See strengths and repair
5 Grok Build: grok-buildFoundations for Agency: Reducing Rents, Widening Access and Embedding Adjustment Capacity 797376.0 Viable core
Why?
Viable core — foundation, not successor income

The proposal stays mostly within existing welfare, procurement, competition, investment and statistical powers and is candid that it cannot replace mass income. Its modest first phase is politically plausible and low regret.

Condition: Adjustment funds and training are too small for the severe scenario, while adoption-rate triggers do not map reliably to actual household harm.

See strengths and repair
6 GPT-5.6 Pro (external attribution, not independently verified here)Own the Transition Before It Owns You 727573.5 Viable core
Why?
Viable core — asset strategy, not income settlement

The strongest contribution is its legally distinct treatment of grants, co-investment, infrastructure concessions, procurement and taxation, combined with a two-layer defence against value being hollowed out. That core can acquire public assets without obstructing adoption.

Condition: It provides no arithmetic showing that the resulting assets could finance material household income, and the full agenda is too broad for one implementation sequence.

See strengths and repair
7 Claude Fable (external attribution, exact model not independently verified here)The Cap Table Is the Constitution: Options and Full Memo 737172.0 Viable core
Why?
Viable core — public asset acquisition only

The proposal separates public support, infrastructure concessions, procurement and tax and combines local project claims with upstream participation. This is a credible bargaining and asset-acquisition doctrine, not yet a funded agency-income system.

Condition: There is no scale arithmetic showing that realistic public claims can replace a material wage share, and some entrenchment claims exceed UK legal durability.

See strengths and repair
8 Claude Opus 4.6 (Thinking)The Agency Floor: Preserving Mass Economic Participation Through the AI Transition 696969.0 Viable core
Why?
Viable core — requires a redesigned fiscal route

An employer value-added contribution could be more factor-neutral than payroll tax, and the proposal honestly exposes the cost of a universal credit while including a housing pass-through test. The concept is worth modelling rather than immediate enactment.

Condition: Firm value added is not a simple VAT-return field; border rules are incomplete, NWF capacity is used as bridge finance and the proposed binding EU route is unsound.

See strengths and repair
9 Kimi Code: kimi-for-codingThe Civic Capital Compact: Securing Agency Under Unit Cost Dominance 646765.5 Fatal as drafted
Why?
Fatal as drafted — first-stage package is unfunded

The proposal correctly treats public investment capacity as capital and takes rent leakage seriously, but its first fiscal move does not close. The compute and foreign-cloud base is also not reliably observable.

Blocking issue: A proposed £10–15 billion NIC reduction is paired with speculative £2–5 billion compute receipts and unmodelled allowance savings; imported compute cannot be measured as claimed.

See strengths and repair
10 Gemini 3.1 Pro (High)The Sovereign Value-Base Architecture: Securing Mass Economic Agency Under Unit-Cost Dominance 615859.5 Fatal as drafted
Why?
Fatal as drafted — dividend arithmetic fails

The outcome-based triggers and national delivery route are useful, but the central dividend and its tax base are not viable. The proposal treats uncertain fund returns as if they could close a very large recurring gap.

Blocking issue: A £1,200–£1,800 adult payment costs roughly £65–100 billion annually against about £14–16 billion of claimed surcharge receipts, while the destination value-added base is undefined.

See strengths and repair
11 NVIDIA Nemotron 3 Ultra 550B A55BThe Compute Commons Compact 584953.5 Fatal as drafted
Why?
Fatal as drafted — honest gap, no funded route

The proposal usefully integrates housing and platform rent absorption and admits large dividend gaps, but it schedules benefits before revenues or investment returns exist. Its EU fiscal and social-policy routes are also unavailable or contested.

Blocking issue: UK recurring revenue of about £5.6 billion is below an £8.6 billion dividend before services; the EU gap is larger, and borrowing or speculative returns cannot count as recurring funding.

See strengths and repair
12 Gemini 3.5 Flash (High)The Sovereign Shareholder Framework: Restructuring the UK and EU Demand Circuit for the Compute Age 504648.0 Fatal as drafted
Why?
Fatal as drafted — speculative revenues cannot fund the grant

Public equity and rent shields are useful directions, and the proposal acknowledges the scale of a universal payment. Its central grant nevertheless relies on unobservable compute charges and implausible public-fund returns.

Blocking issue: A roughly £95 billion annual grant rests on unsupported compute receipts, platform taxes, Universal Credit savings and £12 billion of annual NWF returns; the foreign-compute border adjustment is not auditable.

See strengths and repair
13 MiniMax-M2.7Citizen Capital Framework: Structural Adaptation for AI-Driven Labour Market Change 474847.5 Fatal as drafted
Why?
Fatal as drafted — base, return and EU powers fail

Broad citizen ownership and public upside are worthwhile goals, but the package reintroduces the forbidden AI-attribution problem and promises returns its capital cannot produce. Its EU entitlement route is also beyond the cited powers.

Blocking issue: A £2,000 account cannot guarantee £500 annually without a 25% yield; NWF capacity is used as bridge finance, and AI-assisted output charges plus the Article 153 route are defective.

See strengths and repair
14 NVIDIA Nemotron 3 Super 120B A12BResilient Agency Framework for the AI Transition 484647.0 Fatal as drafted
Why?
Fatal as drafted — commitments exceed receipts and powers

The proposal is candid that its compute contribution raises very little, and contract-linked portability has ordinary value. But the service, training and job-guarantee commitments materially exceed the stated revenue and EU authority.

Blocking issue: Projected UK digital receipts do not cover training accounts, services and a possible job guarantee; the compute units are dimensionally incoherent and ESF+ cannot become an open-ended wage guarantee.

See strengths and repair
15 Claude Sonnet 4.6 (Thinking)Civic Dividend: Decoupling Economic Agency from Wage Necessity Before the Circuit Breaks 385044.0 Fatal as drafted
Why?
Fatal as drafted — multiple independent failures

The proposal usefully separates collection, investment and payment administration and admits that the compute levy is small. Its central income promise, tax base, competition remedy and EU route nevertheless fail independently.

Blocking issue: Hundreds of millions of levy revenue cannot fund income near 80% of the minimum wage; accelerator/FLOP and foreign-access bases are unadministrable, and automatic structural separation bypasses due process.

See strengths and repair
16 GPT-OSS 120B (Medium)A New Social Contract for an AI-Enhanced Economy 252223.5 Fatal as drafted
Why?
Fatal as drafted — invented base and institutions

The answer recognises rent leakage and the need for audit, but its policy system is built on an AI-output attribution formula and institutions that do not exist. Spending and assumed investment yields are internally impossible.

Blocking issue: It allocates around £10 billion from roughly £3 billion of claimed receipts, assumes 15–20% yields and depends on invented registries, agencies, tax powers and EU payment infrastructure.

See strengths and repair
17 NVIDIA Nemotron 3 Nano 30B A3BSovereign Agency Dividend 202020.0 Fatal as drafted
Why?
Fatal as drafted — arithmetic wrong by 1,000×

A universal payment is easy to explain and gross revenue can sometimes resist profit shifting, but this proposal's defining tax base is unmeasurable and its basic multiplication is wrong by three orders of magnitude.

Blocking issue: 67 million × £1,200 is £80.4 billion, not £80.4 million, and 450 million × €1,200 is €540 billion, not €540 million; the claimed receipts cannot fund the promise and the EU route is unavailable.

See strengths and repair

The Grok web-research pass is published but excluded from this table because browsing changed the experimental condition.

What survived across models

A narrow common core

Public upside

Only where public support or capital bears genuine risk.

Operational exit

Portability, audit, continuity and tested switching in public contracts.

Household triggers

Measure earnings, hours, entry hiring and essential costs—not AI attribution.

Fiscal separation

Never spend investment capacity or paper gains as recurring income.

Complete council library

Read every answer, including the failures

Raw model text is preserved rather than silently repaired. The editorial and review layers identify errors separately.

Controlled council91.5

GPT-5.6 Terra (model ID: gpt-5.6-terra)

The Agency Circuit Compact

Unit Cost Dominance is a contingency diagnosis, not evidence that the wage demand circuit has already failed. The reported figures support caution rather than certainty: in June 2026, 29% of UK businesses reported AI use, rising to 49%…

Viable core · conditions apply4,024 words
Read the full proposal
Controlled council90.0

GPT-5.6 Sol (model ID: gpt-5.6-sol)

The Agency Dividend Compact

The supplied evidence does not establish economy wide displacement. AI use is rising, but only 5% of AI using UK businesses reported AI related headcount reduction, and causal attribution remains difficult. This calls for institutional…

Viable core · conditions apply3,875 words
Read the full proposal
Controlled council87.0

Codex (GPT-5)

The Civic Agency Compact

The response should not be a robot tax, a promise to save every job, or an immediate leap to an unfunded universal basic income. It should build a successor circuit in stages. The Civic Agency Compact has seven planks: a universal but…

Viable core · conditions apply3,875 words
Read the full proposal
Controlled council81.5

GPT-5.6 Luna (model ID: gpt-5.6-luna)

The Agency Compact: Adopt Freely, Share the Upside

The evidence supplied does not establish economy wide wage collapse. UK AI use reached 29% of businesses in June 2026, and 49% among firms with at least 250 employees, but only 5% of AI using businesses reported AI related headcount…

Viable core · conditions apply4,046 words
Read the full proposal
Controlled council76.0

Grok Build: grok-build

Foundations for Agency: Reducing Rents, Widening Access and Embedding Adjustment Capacity

The Unit Cost Dominance thesis identifies a contingency. If AI plus verification delivers broad cognitive work at lower unit cost than human only production, mass wage labour may cease to be the main route to independent economic agency.…

Viable core · conditions apply3,495 words
Read the full proposal
Controlled council69.0

Claude Opus 4.6 (Thinking)

The Agency Floor: Preserving Mass Economic Participation Through the AI Transition

The Unit Cost Dominance thesis identifies a conditional but serious risk: that AI driven productivity gains could progressively decouple business revenue from mass wage income, undermining both household demand and democratic legitimacy. A…

Viable core · conditions apply3,838 words
Read the full proposal
Controlled council65.5

Kimi Code: kimi-for-coding

The Civic Capital Compact: Securing Agency Under Unit Cost Dominance

This brief proposes a contingency ready policy package for the United Kingdom and the European Union that accepts the possibility of unit cost dominance in cognitive labour without assuming the collapse is either certain or immediate. The…

Fatal as drafted · parts may survive3,748 words
Read the full proposal
Controlled council59.5

Gemini 3.1 Pro (High)

The Sovereign Value-Base Architecture: Securing Mass Economic Agency Under Unit-Cost Dominance

The Unit Cost Dominance thesis identifies a structural vulnerability in postwar political economy. Modern fiscal and social contracts assume that mass labour produces wages, that wages generate aggregate consumer demand, and that payroll…

Fatal as drafted · parts may survive3,993 words
Read the full proposal
Controlled council53.5

NVIDIA Nemotron 3 Ultra 550B A55B

The Compute Commons Compact

The Unit Cost Dominance thesis identifies a structural risk: if AI plus verification makes broad cognitive work cheaper per unit than human only production, the wage demand circuit that underpins postwar prosperity can fray without mass…

Fatal as drafted · parts may survive3,547 words
Read the full proposal
Controlled council48.0

Gemini 3.5 Flash (High)

The Sovereign Shareholder Framework: Restructuring the UK and EU Demand Circuit for the Compute Age

The postwar economic model of the United Kingdom and the European Union relies on a self reinforcing wage demand circuit. In this circuit, human labour produces wages, wages finance household consumption, consumption generates corporate…

Fatal as drafted · parts may survive4,407 words
Read the full proposal
Controlled council47.5

MiniMax-M2.7

Citizen Capital Framework: Structural Adaptation for AI-Driven Labour Market Change

The Unit Cost Dominance thesis raises a credible concern: if AI systems can perform valuable cognitive work at lower unit cost than human only production across broad sectors, the wage demand circuit that has organised economic life since…

Fatal as drafted · parts may survive3,314 words
Read the full proposal
Controlled council47.0

NVIDIA Nemotron 3 Super 120B A12B

Resilient Agency Framework for the AI Transition

This proposal builds a modular policy package that can be deployed in the United Kingdom and the European Union without requiring precise measurement of AI’s share in any workflow. It combines three core planks: a universal basic services…

Fatal as drafted · parts may survive4,633 words
Read the full proposal
Controlled council44.0

Claude Sonnet 4.6 (Thinking)

Civic Dividend: Decoupling Economic Agency from Wage Necessity Before the Circuit Breaks

The Unit Cost Dominance thesis is a conditional but serious warning. Its force is not that all jobs vanish by a fixed date, but that if broadly capable AI can perform cognitive work more cheaply than humans at scale, the wage demand…

Fatal as drafted · parts may survive3,935 words
Read the full proposal
Controlled council23.5

GPT-OSS 120B (Medium)

A New Social Contract for an AI-Enhanced Economy

The Unit Cost Dominance thesis warns that widespread AI substitution could weaken the historic wage demand circuit that underpins mass agency. It does not predict universal unemployment, nor does it prescribe any single institutional fix.…

Fatal as drafted · parts may survive4,626 words
Read the full proposal
Controlled council20.0

NVIDIA Nemotron 3 Nano 30B A3B

Sovereign Agency Dividend

The proposal creates a Sovereign Agency Dividend SAD – a universal, non means tested cash payment funded by a dedicated levy on AI enabled cognitive services and on excess economic rents captured by platform owners. The levy is collected…

Fatal as drafted · parts may survive2,516 words
Read the full proposal
Research-enabled challenge

Grok Build: grok-build

Cost Anchors and Shared Stakes: Securing Mass Economic Agency

The Unit Cost Dominance thesis identifies a conditional risk: if AI plus verification delivers cognitive output at materially lower unit cost across wide task ranges, the postwar wage demand circuit weakens without any requirement for…

Research pass · unranked3,171 words
Read the full proposal

The public case: Use the technology. Protect people’s savings. Make public risk earn a public return. Expand cash only when the evidence, funding and essential-cost capacity are real.

Return to the thesis