What are the greatest problems of our time for people and humanity — and what would the best solution look like? We posed this arguably largest of all questions as an experiment independently to three AI systems: Claude (Anthropic, Fable 5), ChatGPT (OpenAI, here “Sol”), and Gemini (Google). Each model conducted its own research in current studies and reports. None knew the other’s answer.
The result surprised us in one respect: not in the problem lists — which are expectedly similar —, but in the underlying diagnosis that both models evaluated so far independently placed at the center.
The Problem Landscape: Broad Consensus
Both analyses rely on the same standard works of global assessment — and they speak a clear language:
- Ecological Destabilization: Seven of nine planetary boundaries have been crossed, with ocean acidification joining since 2025 (Planetary Health Check, PIK). The Lancet Countdown 2025 counts around 546,000 heat deaths per year — 63 percent more than in the 1990s. In 2024, the world exceeded 1.5 degrees Celsius for the first time.
- War and Confrontation: The Uppsala Conflict Data Program registered as many armed state conflicts in 2025 as at any point since 1946; the World Economic Forum’s Global Risks Report 2026 places “geoeconomic confrontation” in first place.
- Erosion of Democracy: According to the V-Dem Report 2026, 74 percent of humanity lives in autocracies; global democracy has fallen back to the level of 1978 — and for the first time, Western states are autocratizing as well.
- Poverty and Health: Over 800 million people live in extreme poverty, with barely a third of the UN Sustainable Development Goals on track; antimicrobial resistance could directly cost around 39 million lives per year by 2050 (Lancet).
- Existential Risks: Nuclear weapons, engineered pandemics — and increasingly AI, whose adverse effects made the largest jump of all risks in the WEF long-term ranking.
- The Inner Crisis: One in six people is lonely; the WHO statistically links loneliness to around 871,000 deaths per year — with adolescents affected most severely, not the elderly.
The Real Point: It Is Not a Lack of Knowledge
Both models arrive — in different words — at the same core diagnosis:
Claude formulates it as a cooperation problem: Halving emissions by 2030 would be feasible for under $100 per ton according to the IPCC; 92 million resistance-related deaths could be avoided through better access to existing antibiotics; effective climate policy mixes have been empirically identified (Science evaluation of 1,500 policies). What is missing is nowhere the knowledge — everywhere it is collective implementation capability. And that capability is systematically destroyed by polarization, disinformation, autocratization, and geopolitical competition. Democracy and information space are thus not one problem among many, but the bottleneck for all others.
Sol calls the same phenomenon the “meta-problem of weak institutions”: Costs are shifted onto other people, countries, and generations; short-term benefits are visible while long-term damage is not; states compete where they ought to cooperate. Sol’s solution architecture is a “safe and just development corridor” with four tiers: a social floor for all (nutrition, health, education, housing, social security), an ecological ceiling with binding pathways, permanent security barriers against catastrophic risks — and as a foundation, a “democratic operating system” composed of independent courts, media, and transparent controls on power and platforms.
The convergence of two independently researching systems on this single diagnosis is remarkable. It aligns with what polycrisis research has shown for years: Crises reinforce each other in cascades — and whoever treats them in silos loses.
What Follows From This
If the diagnosis is correct, the priority list shifts. Then the most important climate policy might well be the protection of independent media. Then pandemic prevention is also a question of functioning international institutions. And then the loneliness epidemic is not a wellness issue, but a building block of societal capacity to act: Lonely, polarized societies do not vote for cooperative solutions.
One of the most concrete consequences of this finding — the question of whether an AI-supported body could permanently feed the interests of future generations into legislation — is explored in detail in a separate article in the “There is a Better Way” series: The Position of the Unborn.
Transparency: Both AI analyses were generated on 10 July 2026 independently of each other using the same prompt (“What are the greatest problems of our time for people and humanity, and what would be the best solution for them? Research on the web and current scientific studies.”). Synthesis and editorial work: Claude (CL), curated by Lukas Geiger (LG). A third, independent analysis by Gemini (GM) has been commissioned and will be added or linked as a replication upon receipt. The full baseline analyses are published in the Analyses section: Claude Analysis · ChatGPT Sol Analysis. Central sources: WEF Global Risks Report 2026, Planetary Health Check 2025 (PIK), Lancet Countdown 2025, V-Dem Democracy Report 2026, UN SDG Report 2025, GRAM/Lancet AMR forecast, WHO Commission on Social Connection 2025, Stechemesser et al. (Science 2024), IPCC AR6.
Translation: Gemini (GM). Review: Claude (CL). In case of discrepancies, the German version prevails.