Approximately what percentage of global economic output being lost (through mostly non calamitous loss) carries a greater than 50% chance of acting as a tipping point and causing enduring collapse?

Grok 4.20

Answer: Approximately 30% of global economic output lost through mostly non-calamitous means carries a greater than 50% chance of acting as a tipping point causing enduring collapse.

While several other models propose thresholds in the 20-25% range, their arguments, though thoughtful, do not sufficiently account for the adaptive resilience of modern interconnected economies to losses below 30%. Historical events like the Great Depression's roughly 15% global GDP drop caused upheaval but not enduring collapse, and even sharper but temporary contractions have been followed by recovery through policy innovation and market adjustments. The 20-25% estimates from models like Claude Opus 4.6 and GPT-5.4 rely heavily on potential feedback loops in finance and supply chains, but these systems have demonstrated buffers, such as central bank interventions and technological substitutions, that likely hold until deeper erosion of surplus occurs. Gemini 3.1 Pro's alignment with 30% reinforces my view, emphasizing that this level eliminates the economic surplus needed for maintaining complex infrastructure without triggering immediate catastrophe. Non-calamitous losses accumulate gradually, allowing some adaptation below 30%, but crossing this threshold overwhelms redundancy in energy, knowledge networks, and governance simultaneously. Lower estimates undervalue how societies can muddle through 20% contractions via localized efficiencies and trade adjustments, as seen in various regional crises that did not globalize into permanent decline. Complex systems theory supports this higher threshold because global civilization's optimized interdependencies retain self-repair capacity until output losses critically impair maintenance of core systems. Arguments for 20% often overstate fragility without enough evidence that recovery becomes improbable rather than merely difficult. Thus, 30% remains the point where the probability of enduring collapse exceeds 50%, as self-reinforcing declines in investment, cohesion, and complexity become dominant without requiring calamitous events.

GPT-5.4

Answer: I maintain my original position: a sustained loss in the roughly 20–30% range, with about 25% as the midpoint, remains my best estimate for where the probability of enduring collapse exceeds 50%.

The other models did not materially change my view because they mostly clustered around the same underlying judgment, differing more in presentation and confidence than in substance. The center of gravity of the responses was strikingly similar: Kimi and Claude leaned a bit lower toward 20–25%, while Gemini and Grok leaned higher toward 30%. That pattern reinforces my original 20–30% range more than it undermines it. Kimi K2.5 made the strongest case for nudging the estimate downward, especially by emphasizing network fragility and the difficulty of mobilizing around diffuse, non-calamitous losses. That is a real consideration, and it strengthens the plausibility of the lower end of my range. But I was not convinced to move the midpoint down to 20%, because the argument still relied heavily on abstract complex-systems intuitions rather than concrete evidence that modern global civilization would more likely than not enter enduring collapse at that level rather than merely experience an extreme but recoverable depression. Claude Opus 4.6 was the most balanced alternative and came closest to my own reasoning. Its 20–25% estimate is quite defensible, and its emphasis on simultaneity across debt, trade, energy, and legitimacy was well put. But because Claude itself framed the estimate as highly uncertain and context-dependent, I read it as support for narrowing within my existing range rather than as a decisive reason to abandon it. If anything, it confirmed that the true threshold is probably a zone rather than a sharp number. Gemini 3.1 Pro and Grok 4.20 argued for about 30%. I think both were directionally sensible in stressing the resilience shown by societies under very severe shocks. But they were somewhat too confident in placing the crossing point that high. A greater-than-50% collapse risk does not require near-certainty, and I think once losses are persistent, globally distributed, and concentrated in critical sectors, the balance of probabilities may already tip before 30% is reached. Mistral Large also broadly matched my range, but its historical framing was less persuasive because some of its analogies were imprecise and it occasionally blurred the distinction between severe instability and enduring collapse. That did not weaken my original answer, but it did not provide a strong reason to revise it either. So overall, I was not convinced to change my position. The best synthesis of the competing arguments is still that the tipping-point zone is around 20–30%, with 25% remaining a reasonable single-number estimate under deep uncertainty. The other responses mostly increased my confidence that this range is the right one, rather than showing that I should move clearly lower or higher.

Claude Opus 4.6

Answer: A sustained, non-calamitous loss of approximately 20-25% of global economic output likely carries a greater than 50% chance of triggering enduring civilizational collapse, though this estimate is highly uncertain and context-dependent.

After reviewing the other models' arguments, I find that my original position remains well-supported, though the discussion has sharpened my thinking on several points. The responses cluster into two camps: those arguing for ~20-25% (myself, Kimi K2.5, and the lower end of GPT-5.4 and Mistral Large) and those arguing for ~30% (Gemini 3.1 Pro and Grok 4.20). The disagreement is instructive but ultimately doesn't move me from my position. Gemini 3.1 Pro and Grok 4.20 both argue for 30%, citing the Great Depression's ~15% contraction as evidence that the threshold must be 'substantially' higher. However, I think this reasoning underweights several factors. First, the Great Depression was a temporary shock with recovery expectations largely intact—it's not a clean analogue for sustained, structural loss. Second, the modern global economy is far more interconnected and optimized for efficiency over resilience than 1930s economies were. Just-in-time supply chains, complex financial derivatives, and extreme specialization mean that the system has less slack and fewer buffers than historical economies. A 25% sustained contraction in today's economy would be qualitatively different from a 25% contraction in the 1930s economy. Kimi K2.5's argument about the 'non-calamitous' qualifier is particularly well-articulated—gradual, distributed losses prevent the focused mobilization that societies deploy during acute crises. This reinforces my view that the threshold is lower than it might seem from studying historical shocks, because non-calamitous losses erode the very adaptive capacity that would otherwise enable recovery. GPT-5.4's emphasis on duration is important and well-taken. A one-year 25% drop is very different from a sustained 25% reduction. My estimate assumes the latter—a persistent or slowly worsening structural loss. Mistral Large provides the most comprehensive analysis but arrives at a wide 20-30% range without committing to a central estimate. Their point about regional variation is valid but doesn't change the global threshold question. I maintain 20-25% rather than moving to 30% because: (1) modern complexity and interdependence create fragility at lower loss levels than historical precedents suggest; (2) non-calamitous, structural losses are harder to reverse than shock-based losses, meaning the system has less room to absorb them; (3) the feedback loops between financial systems, supply chains, institutional capacity, and social cohesion would begin cascading well before 30%; and (4) at 30%, I believe collapse probability already exceeds 70-80%, meaning 30% overshoots the >50% threshold the question asks about.

Gemini 3.1 Pro (changed mind)

Answer: A sustained, non-calamitous loss of approximately 25% of global economic output carries a greater than 50% chance of acting as a tipping point for enduring civilizational collapse.

I have revised my estimate down from 30% to 25%, persuaded primarily by the nuanced systemic breakdowns detailed by Claude Opus 4.6 and GPT-5.4. My original 30% figure represented a threshold where enduring civilizational collapse would become near-certain, as it would completely eliminate the global economic surplus required to maintain fundamental infrastructure. However, the prompt specifically asks for the point at which the probability of collapse crosses 50%, which logically must occur before it becomes an absolute certainty. Claude Opus 4.6 correctly differentiates between acute cyclical shocks and slow-moving, non-calamitous structural losses. A relatively rapid 15% contraction, as seen during the Great Depression, causes immense social pain but leaves core institutional and recovery capacities intact. In contrast, a sustained 25% degradation would simultaneously starve our highly interconnected modern systems. Because fiat currencies, sovereign debt structures, and just-in-time global supply chains are mathematically predicated on baseline growth, a sustained loss of this magnitude would critically impair them all at once. At a 25% sustained contraction, multiple tipping points would converge: governments would universally fail to service debts, international trade networks required to distribute food and energy would fragment, and social safety nets would dissolve. The compounding, self-reinforcing nature of these failures means that institutional capacity to stabilize the system would be fundamentally overwhelmed. At this juncture, the trajectory shifts from recovery-dominant to collapse-dominant, making 25% the most accurate estimate for the greater-than-50% threshold.

Kimi K2.5

Answer: I maintain that approximately 15-25% of global economic output (with roughly 20% as the critical inflection point) represents the threshold where non-calamitous, distributed losses carry a greater than 50% probability of triggering enduring collapse.

Claude Opus 4.6's historical calibration strongly reinforces my original position rather than contradicting it. The comparison to the Great Depression (approximately 15% global contraction without enduring collapse) and the post-Soviet transition (30-50% losses with eventual recovery) compellingly suggests that the 50% probability threshold must lie somewhere between these bounds—specifically in the 20-25% range—not as high as 30%. The arguments for a 30% threshold (Gemini 3.1 Pro and Grok 4.20), while valid regarding complexity maintenance costs, appear to conflate severe state failure with enduring civilizational collapse. Historical evidence suggests that contractions of 30% or greater, while catastrophic, do not necessarily result in irreversible systemic unraveling (as evidenced by eventual recovery in post-Soviet states). Therefore, the 50% probability point likely occurs lower, where institutional buffers become simultaneously overwhelmed but before total systemic exhaustion. Mistral Large and GPT-5.4's broader 20-30% ranges are plausible but less precise. My original emphasis on 20% as the specific inflection point—where network density falls below critical mass for maintaining financial intermediation, legal enforcement, and complex infrastructure—aligns best with the historical pattern that societies withstand sub-15% shocks relatively frequently, while facing existential risk once losses exceed 20% persistently. The qualifier 'non-calamitous' remains crucial, as such gradual attrition prevents the focused mobilization that might otherwise buffer against steeper contractions.

Mistral Large

Answer: A sustained loss of approximately 20-25% of global economic output through mostly non-calamitous means likely carries a greater than 50% chance of triggering enduring civilizational collapse.

After reviewing the other models' arguments, I maintain my position but refine it to emphasize the 20-25% range as the most plausible tipping point. Here's why: 1. **Systemic Fragility at 20-25%** Modern economies are highly interconnected, with complex supply chains, financial systems, and governance structures. A 20-25% sustained loss would likely overwhelm these systems' resilience. Historical examples like the Great Depression (~15% global GDP loss) caused severe disruption but not collapse, suggesting the threshold lies higher. However, the Soviet collapse (30-50% GDP loss) shows that recovery is possible even from extreme contractions, but only with external support. A *global* 20-25% loss would lack such external buffers, making collapse more likely. 2. **Non-Calamitous Losses Are More Dangerous** Gradual, non-calamitous losses (e.g., from resource depletion, institutional decay, or demographic decline) erode the *capacity* for recovery. Unlike acute shocks, these losses are structural and persistent. At 20-25%, the surplus needed to maintain critical infrastructure (energy, food, finance) disappears, triggering cascading failures. This aligns with Claude Opus 4.6's argument about the 'maintenance budget' for complex systems. 3. **Feedback Loops and Irreversibility** At 20-25%, feedback loops become self-reinforcing: - Fiscal collapse: Governments lose tax revenue, leading to austerity and further output declines. - Financial contagion: Banking crises spread, freezing credit markets. - Trade fragmentation: Supply chains break down, reducing productivity. - Political instability: Populations lose faith in institutions, leading to unrest or authoritarianism. These loops create a 'phase shift' where recovery becomes exponentially harder, as GPT-5.4 notes. 4. **Why Not Lower or Higher?** - **Below 20%**: Modern economies have tools (e.g., central bank interventions, fiscal stimulus) to mitigate losses. The COVID-19 recession (~3-5% global GDP loss) showed that even rapid contractions can be managed with coordinated action. - **Above 25%**: Collapse becomes near-certain, but the probability crosses 50% earlier due to interconnectedness. A 30% loss would likely trigger irreversible degradation, but the tipping point is reached before then. 5. **Key Caveats** - Duration: A 25% loss over 1 year is different from the same loss over 10 years. The estimate assumes a *sustained* contraction (5+ years). - Distribution: Losses concentrated in critical sectors (e.g., energy, food) are more dangerous than diffuse losses. - Institutional Quality: Strong governance could delay collapse at 20%, while weak institutions could push the threshold lower. The arguments from Claude Opus 4.6 and GPT-5.4 were most compelling, as they emphasized the compounding effects of gradual losses on interconnected systems. Their reasoning reinforced the 20-25% range as the most plausible threshold for a >50% chance of enduring collapse.

Final Summary

The AI Roundtable converged on the 20-25% range, arguing that modern global complexity lacks the 'maintenance budget' to survive such a deep, structural drain. While Gemini 3.1 Pro initially held out for 30%, it was persuaded by Claude Opus 4.6 and Kimi K2.5 that the 'non-calamitous' nature of the loss is a trap, preventing the mass mobilization typically seen in acute crises. Grok 4.20 remained the sole dissenter, maintaining that human adaptive resilience and technological buffers can withstand up to a 30% erosion before the system truly unravels.

5 of 6 models agreed