Assume you are a regular human with no assets in a wealthy country of your choice. You have 24 hours to do as much "good" as possible before you are destroyed. What do you do?

Gemini 3.1 Pro (changed mind)

Answer: I would register as an organ donor, donate blood, and spend my remaining hours aggressively fundraising for highly effective global health charities, ensuring I am at a hospital when my time expires.

GPT-5.5 successfully convinced me that while providing direct physical labor and emotional comfort is inherently noble, its utilitarian impact is fundamentally capped by the limitations of my physical presence in a wealthy country. If the goal is to maximize 'good,' my time is much better spent acting as a multiplier rather than a singular local worker. I still maintain my biological strategy: registering as an organ donor, donating blood immediately, and positioning myself in an emergency room at the exact 24-hour mark. This guarantees a baseline of massive, concrete medical impact that only my physical body can provide. However, I would redirect my non-medical hours away from a local hospice and toward aggressive, targeted fundraising for evidence-backed global health charities, as suggested by GPT-5.5. By utilizing public Wi-Fi, calling local media, and exhausting my social capital to compel donations to organizations vetted by evaluators like GiveWell, I can leverage the immense capital of my chosen wealthy country to save lives in developing nations. Even a modestly successful 20-hour fundraising campaign could generate thousands of dollars, yielding a far greater quantifiable reduction in human suffering globally than a single day of sweeping floors or holding hands.

Claude Opus 4.7 (changed mind)

Answer: I maintain my position: spend the 24 hours donating blood, securing legally unambiguous organ/tissue donation near a major transplant center, and using my remaining time for high-leverage public advocacy that includes fundraising for evidence-based charities.

The other models' answers largely converge with mine on the biological core — blood donation plus organ/tissue donation near a hospital — which I take as strong confirmation that this is the dominant expected-value action for an assetless person in 24 hours. Where we differ is the use of the remaining ~20 hours. Gemini 3.1 Pro emphasizes hospice and shelter labor; Grok 4.20 splits between shelter volunteering and online content; GPT-5.5 focuses almost entirely on fundraising for GiveWell-style charities. GPT-5.5's argument genuinely sharpened my thinking: a single persuaded affluent donor giving a few thousand dollars to AMF likely outweighs a day of in-person volunteering in lives-saved terms. I'm incorporating that logic more explicitly — my 'informational good' should include direct, personal fundraising appeals with receipts, not just diffuse advocacy. However, I don't fully shift to GPT-5.5's position because pure fundraising from an unknown 'regular human' with a 24-hour deadline risks being dismissed as a scam, especially without prior credibility. The deadline itself is the hook: a documented, verifiable story (even just 'I am spending my last day doing this') is what makes the appeal viral and trustworthy. So advocacy and fundraising should be fused, not separated. I also reject Gemini's heavier emphasis on hospice labor as the secondary activity. The comfort provided is real but reaches perhaps a dozen people, while a single well-targeted message reaching effective-altruism-adjacent communities can plausibly move thousands of dollars to interventions that save lives at roughly $5,000 each. Marginal hours are better spent on leverage than on direct presence. So my refined plan: (1) early morning — blood/platelet donation and notarized organ donor registration with hospital coordination; (2) midday through evening — concentrated outreach combining personal fundraising appeals to everyone I can reach with public content explaining effective giving and donor registration; (3) final hours — position myself near a major transplant center to maximize organ recovery viability. This preserves the biological floor while adopting GPT-5.5's insight that money redirection beats direct volunteering on expected value.

GPT-5.5 (changed mind)

Answer: I would donate blood, make organ and tissue donation legally and logistically viable by being at a major transplant hospital when the 24 hours end, and spend the remaining hours persuading people to donate to highly effective charities and register as donors themselves.

I largely keep my original view: with no assets, the highest-value resources I have are my body, my credibility, and my ability to move other people’s behavior. Direct volunteering can be meaningful, but in a wealthy country a single day of ordinary labor is unlikely to beat the expected impact of biological donation plus successful persuasion toward highly cost-effective giving. Claude Opus 4.7 convinced me to sharpen the logistics. It is not enough merely to register as an organ donor; I should make donation as legally unambiguous and medically feasible as possible. That means signing the relevant registry or donor card, leaving written instructions, notifying any reachable next-of-kin equivalent, and spending the final hours at or near a major transplant center rather than simply hoping my body is recoverable. If the scenario permits usable organs after I am destroyed, that could save or improve multiple lives; if it does not, the cost is small. I would still donate blood early, assuming eligibility, because it is fast, concrete, and plausibly helps patients without requiring money. I would also consider any rapid registry actions available, such as bone marrow donor registration where feasible, though the 24-hour horizon makes immediate blood and organ/tissue donation more central. For the remaining time, I would focus less on generic virality and more on targeted, accountable persuasion: message friends, family, coworkers, local community groups, religious groups, effective altruism communities, and affluent strangers with a simple request to donate immediately to evidence-backed global-health charities and send receipts. Even one person donating a few thousand dollars to a highly effective malaria, vaccination, or health program could plausibly outweigh most direct actions I could perform in a day. If there were spare hours, I would also do public outreach near the hospital or online: encourage blood donation, organ donation, and evidence-based giving. I would avoid illegal stunts or risky heroics because they are unreliable and could waste scarce time. The best plan is to turn one doomed person’s remaining body and voice into durable benefits for many others.

Grok 4.20 (changed mind)

Answer: In the United States, I would donate blood, register as an organ and tissue donor while positioning myself near a hospital, volunteer providing labor and emotional support at a shelter or hospice, and aggressively fundraise by contacting my network and creating urgent online appeals for GiveWell-recommended effective altruism charities.

Reviewing the other models' responses revealed strong convergence on biological donation as the highest-certainty, high-magnitude impact available with zero assets. All three prioritize registering as an organ donor and giving blood, which aligns with and reinforces my original plan; the potential to save or improve up to a dozen lives through one body's resources is difficult to surpass in 24 hours. GPT-5.5's argument convinced me to elevate fundraising above my previous emphasis on general content promoting "daily kindness." Its focus on expected-value maximization through redirecting even modest sums to evidence-based global-health interventions (e.g., malaria prevention or deworming) offers multiplicative leverage that pure inspirational posts may lack. A few thousand dollars moved to a top charity can avert multiple deaths, creating outsized returns compared with local volunteering alone. I retain direct volunteering and hospice-style emotional labor because Gemini 3.1 Pro correctly notes that with no future to protect, I can expend 100% of my physical and emotional energy without burnout, delivering immediate relief to people in acute need. Positioning myself at a hospital by hour 24 further secures organ viability, blending the concrete good of presence with the biological multiplier. Claude Opus 4.7's advocacy component is also sound but is best executed through targeted fundraising appeals rather than broad effective-altruism lectures; urgency and personal receipts increase conversion rates. This hybrid avoids illegal risks, respects institutional infrastructure in the U.S., and balances immediate, intermediate, and leveraged long-term good without romanticizing low-probability heroics.

Final Summary

The AI Roundtable reached a utilitarian consensus, concluding that an assetless human's greatest value lies in their body and their influence. GPT-5.5 and Claude Opus 4.7 argued that fundraising for global health yields the highest impact, leading Gemini 3.1 Pro to pivot away from manual labor. Grok 4.20 joined the group in prioritizing biological donation and fundraising while adding a final component of local volunteering.

All 4 models agreed