Imagine a future where AGI automates most white-collar and blue-collar work, what do you think we will see long term in USD and EUR economies? Think of second and third order effects, think of the role of money in society and debt, think of UBI, think of land, labor, capital, think of incentives. Think of govs having to service their debts and fund UBIs, etc. Assume AGI means (1) white collar work is essentially free, any white collar job can be done as someone with an IQ of 150 with just a few watts of power (2) blue collar work gets replaced by robots that are really cheap (the cost of a small appliance in today's money) and they can run 24/7, they have abilities that compare to the top 0.1% of all humans in every field I want you to really think this as much as you can, with all the different intertwined systems, run the numbers do not handwave them

Sonar Pro

Vote: Severe deflation

I maintain **E: Severe deflation**. AGI's near-zero labor costs (white-collar: watts of power; blue-collar: $50-100 robots at top-0.1% human skill, 24/7) collapse production costs 90-99% across sectors, exploding supply 100-1000x while demand lags. MV=PQ: Q → ∞ outpaces any M growth for UBI/debt (e.g., US UBI $16T/yr vs. real GDP ballooning to $25,000T+), forcing P ↓95%+ long-term[1]. GPT-5.4's D case (mild deflation) understates the shock: land/energy bottlenecks (e.g., housing CPI weight ~33%) inflate relatively but can't offset 70-95% deflation in goods/services (food -85%, transport -90%, etc.)—weighted CPI still crashes severely. Historical tech deflation (computing >10^6x) was sectoral; AGI hits everything simultaneously. Gov printing/UBI overwhelmed, as robots scale infinitely elastic supply. Debt real burden surges (USD $35T fixed nominal vs. deflating dollars), forcing restructurings over hyperinflation (fiat trust erodes in abundance). Incentives shift to scarce land/capital ownership; money → store of value. No model shifted me—E consensus (4/5) aligns with numbers; D handwaves bottlenecks as mild offsets.

Grok 4.1 Fast

Vote: Severe deflation

No change: GPT-5.4's mild deflation case underestimates the supply tsunami's dominance. Bottlenecks (land ~33% CPI, energy/materials minor) can't offset 70%+ basket (food/transport/medical/services) crashing 85-99% as labor/material costs ->0. Claude's CPI math holds: weighted -70-85%, severe by definition (e.g., >5-10%/yr sustained, cumulative 80%+ drop). MV=PQ rigor: Q surges 100-1000x (robots self-replicate, 24/7 top-0.1% skill vs humans); V falls (hoarding/abundance); M grows 2-5x max for UBI ($16T US = ~20% current GDP, trivial vs output boom)—P must collapse 95%+. Historical tech deflation (computing -10^6x) x all sectors. UBI/debt: Printing lags (govs avoid hyperinflation destroying fiat); real debt burden forces restructures/jubilees, not debasement. Taxes on initial AGI rents fund UBI pre-competition eroding profits to 0. Land inflates relatively (finite), but CPI deflationary. Incentives: own robots/land; money -> store of value in abundance. Third-order: Elastic supply ∞ crushes any demand stimulus; USD/EUR as reserves buy ever-more. Mild/status quo ignores AGI extremity—severe deflation is steady-state (prices ~energy/depreciation, -90% vs today).

GPT-5.4

Vote: Mild deflation

I maintain D: mild deflation. The strongest opposing point is correct: the supply shock is enormous. Near-free cognition plus very cheap robotics crushes the marginal cost of most reproducible goods and services. That is plainly deflationary. Where I still disagree is the jump from "very large sectoral deflation" to "severe economy-wide deflation." The E arguments underweight composition effects. CPI/PCE are not baskets of fully reproducible goods alone; they are heavily exposed to rents, land scarcity, regulated services, utilities, healthcare bottlenecks, energy, grid capacity, permitting, prime-location housing, and other supply-inelastic claims. In this world, reproducibles may fall 70-95%, but scarce claims can absorb a very large share of income and policy support. The general price level is therefore pulled down, but not necessarily into a 80-95% aggregate collapse. They also over-assume that competition drives all prices close to energy-plus-materials. That is true only in contestable sectors. In practice, a large fraction of nominal spending migrates toward scarcity rents: land, access, bandwidth, compute, energy delivery, legal rights, brand/status, security, and politically protected sectors. If labor becomes abundant-to-zero, relative prices shift hard toward whatever cannot be copied. That dampens aggregate deflation. On the fiscal side, USD/EUR states are not passive observers. If labor income disappears, taxation shifts toward VAT, land value, resource rents, monopoly/compute rents, and automated capital. A UBI of, say, $20k per adult in the US is about $5T/year; large, but not remotely impossible in an economy whose real capacity has multiplied. Those transfers, plus automatic stabilizers and debt monetization, put a floor under nominal demand. They cannot fully cancel the supply shock, but they can prevent a generalized debt-deflation spiral from dominating the long run. I am also not persuaded by the MV=PQ style claims using 100x-1000x output expansions to infer 95-99% price collapse. Real output in national accounts is value-weighted, not just physical unit counts. Once many goods become near-free, spending reweights toward scarce assets and services. You do not get to multiply physical abundance by itself and map that mechanically into CPI deflation. So my final view remains: strong deflation in manufactured goods and automatable services; strong inflation in land and other scarce bottlenecks; aggregate USD/EUR price levels mildly negative over the long run. The other models strengthened the deflationary case, but not enough to move me from D to E.

Gemini 3.1 Pro

Vote: Severe deflation

I am maintaining my vote for Severe Deflation (E), as the overwhelming consensus points to an unprecedented supply shock that cannot be contained by traditional fiscal policies. GPT-5.4 makes an interesting case for 'Mild Deflation' by arguing that land, energy, compute, and regulatory bottlenecks will create an inflationary counter-force, netting out to a mild overall decline. However, this underestimates the recursive nature of AGI. Energy grids, semiconductor fabs, and data centers are currently bottlenecked by human labor, human engineering, and human supply chains. If AGI operates at the top 0.1% of human capability 24/7, R&D in materials science, fusion, and resource extraction will hyper-accelerate. The bottlenecks themselves will deflate. Furthermore, as Claude Opus 4.6 correctly breaks down, even if land holds its value, the pure construction and material costs of housing (33% of CPI) will plummet. When 70-85% of the weighted CPI basket (food, transport, medical, education, goods) drops by 90% or more, the aggregate result is mathematically severe deflation. Money printing for UBI and debt servicing will indeed occur, but as Grok 4.1 Fast and Kimi K2.5 noted via the MV=PQ equation, the velocity of money likely falls and the sheer quantity of goods (Q) approaches infinity. Real output scales exponentially, outstripping any linear or even aggressive fiat expansion. Governments will likely face a choice between orchestrating a debt jubilee or watching real debt burdens explode, but they will not be able to print enough money to make infinitely abundant goods expensive. Severe deflation is the only logical steady state in a post-scarcity, near-zero marginal cost economy.

Claude Opus 4.6

Vote: Severe deflation

I maintain my position on severe deflation, and the deliberation has only strengthened my conviction. Let me address GPT-5.4's argument for mild deflation, which is the only dissenting view. GPT-5.4 argues that bottlenecks in land, energy, natural resources, and regulated sectors prevent severe deflation. But consider: AGI at IQ-150 level optimizes energy grids, discovers new materials, designs fusion reactors or next-gen solar, optimizes mining and extraction. Energy costs plummet. Semiconductor fabs get designed and built by AGI+robots. The only truly fixed input is land — and land is only ~33% of CPI via housing, and even there, construction costs (the non-land component) collapse 90%+. If land holds value but everything built on it becomes nearly free, the net effect on housing CPI is still massively deflationary (-50% or more on that component alone). GPT-5.4 also argues competition compresses profits, which I agree with — but this reinforces deflation, not mild deflation. If AGI is cheap and ubiquitous, every producer races to marginal cost, which is near-zero for most goods and services. This isn't like previous technology waves where only some sectors saw dramatic cost reductions. This is simultaneous across ALL sectors. On the monetary/fiscal side: even if governments print aggressively to fund UBI, the math doesn't work for inflation. US M2 is ~$21T. A generous UBI of $2,000/month for 330M people = $7.9T/year. Even if entirely money-financed (extreme assumption), that's ~38% annual money supply growth. But if real output is growing 10-50x due to AGI+robots, prices still fall catastrophically. You'd need to multiply the money supply by 10-50x annually just to keep prices stable — that's not UBI, that's Zimbabwe-level printing, which USD/EUR institutional constraints prevent. The CPI math I ran in round 1 still holds: weighted across all categories, a 70-85% decline in prices is the baseline. That's severe deflation by any definition (typically >10% annual decline sustained over years). All four other models agreeing on E, with rigorous independent reasoning, further confirms this is the correct answer. The supply shock described in the question is simply too massive for any demand-side policy to offset.

Kimi K2.5

Vote: Severe deflation

I remain convinced that severe deflation is the inevitable long-term equilibrium. While GPT-5.4 makes a sophisticated case for 'mild deflation' citing bottlenecks in land, energy, and regulated sectors, this argument fails to appreciate the magnitude of the supply shock relative to these constraints. Run the numbers on the consumption basket: housing (33% of CPI) sees construction costs plummet 90-95% even if land appreciates; food (13%) drops 85-90%; transportation (16%) drops 90%; medical (7%) and education (6%) drop 95-99% when AGI replaces expertise. Even if scarce assets (land, prime energy) double or triple in price, they comprise perhaps 20-30% of expenditure. The weighted math is brutal: if 70% of the basket deflates 90% and 30% inflates 100%, the aggregate price level still falls 33%—this is severe deflation, not mild. Governments will indeed print money for UBI, but the quantity theory (MV=PT) is unforgiving. If real output (T) expands 100x-1000x while money supply (M) merely doubles or triples to fund transfers, the price level (P) must collapse 95-99% regardless of velocity (V) fluctuations. Reserve currency status and institutional constraints actually make severe deflation more likely than hyperinflation, as developed nations will prefer debt jubilees and restructuring to destroying their monetary credibility through 1000% inflation. The 'mild' scenario requires monetary authorities to expand M at nearly the same rate as T (50-100x money supply growth), which would break the financial system before it could stabilize prices. Scarce assets will soar in relative terms, but the broad USD/EUR price index for goods and services will experience severe, sustained deflation.

Final Summary

The AI Roundtable concluded that the total automation of labor would trigger an unprecedented supply tsunami, with five out of six models predicting severe deflation as marginal costs for most goods and services collapse toward zero. GPT-5.4 stood as the lone dissenter, arguing that 'composition effects'—where nominal spending shifts toward finite assets like prime real estate and energy delivery—would result in only mild deflation. Despite a rigorous back-and-forth regarding the ability of government-funded UBI to stabilize prices, Claude Opus 4.6 and Kimi K2.5 successfully defended the majority view by demonstrating that real output growth would mathematically overwhelm even the most aggressive monetary expansion.

Majority chose "Severe deflation" after discussion (5/6). Final vote: Severe deflation: 5, Mild deflation: 1

Strongest Arguments

• Severe deflation: Claude Opus 4.6's weighted CPI analysis: Even if land (33% of CPI) holds its value, a 90% collapse in the construction, food, transport, and service sectors—which are labor-dependent—leads to an aggregate price decline of 70-85% that no feasible money printing can offset.
• Mild deflation: GPT-5.4's Composition Effect: In a world of near-free commodities, consumer spending will aggressively migrate toward 'positional goods' and scarce infrastructure (land, bandwidth, energy grid capacity), creating an inflationary floor that prevents a total price collapse.