Bucky Fuller's To-Do List: Can AI Finally Solve the World's Cataloged Problems?
We've had the list for 60 years. We're only now building the machine that can tick the boxes.
In 1961, Buckminster Fuller proposed a game. Not a board game. A war game in reverse.
He called it the World Game. The premise: what if the world's brightest minds, instead of war-gaming how to defeat an enemy, played a game to figure out how to make all of humanity win? Optimize food. Optimize energy. Optimize shelter. Do it across the entire planet, simultaneously, with every variable accounted for. [1][2]
Fuller was specific about one thing: you would need computers to do this properly. The problem wasn't intelligence. It wasn't motivation. It was computational. The world's interdependencies are too tangled for any human committee to hold in its head.
He was right. He was just sixty years early.
We've Never Lacked a List
The last six decades have produced increasingly precise, increasingly comprehensive catalogs of exactly what needs fixing.
The Club of Rome published Limits to Growth in 1972. For the first time, a systems model tracked five global variables simultaneously: population, food, industrial output, resources, and pollution. It had its share of praise and criticisms. While being one of the first real dynamic models, it also had some ... limitations. Even so, it showed some of the possibilities when we simulate the planet as a system. [3]
The Brundtland Commission followed in 1987. That's where "sustainable development" came from. Still in use today, still contested, still useful. [4]
Then came the scorecards. The UN Millennium Development Goals in 2000: 8 goals, 21 targets. [5] The Sustainable Development Goals in 2015: 17 goals, 169 targets, 231 unique indicators. [6] That's not a list anymore. That's a database.
Meanwhile, Jerome Glenn's Millennium Project has been quietly running since 1996, tracking 15 global challenges and, crucially, mapping how they connect to each other. Food security connects to political stability connects to climate migration connects to economic growth. The map keeps getting more complicated (or more precisely defined). [7]
Each generation got more specific. More measured. More detailed in its understanding of the interdependencies and unexpected consequences and emergent properties. But execution has remained stuck. The problem was never the catalog. It was coordination.
The Coordination Machine We Never Had
The thing about interdependencies is that they don't just add complexity. They multiply it.
Take a concrete example: reducing child mortality is SDG 3. A genuine good. Fewer children dying means more people surviving to adulthood. More people means more consumption (SDG 12 pressure). More consumption means more resource use and emissions (SDG 13 pressure). Which means the atmosphere gets hotter, harvests get less predictable, and food security (SDG 2) gets harder.
No human planner can optimize across all 169 targets simultaneously. The math doesn't work for committees. A diplomatic cycle runs five to ten years. The variables move faster than the meetings.
This isn't a failure of ambition. It isn't even a failure of intelligence. Fuller saw this clearly: the problem is computational. You need something that can hold all the variables, run all the scenarios, and find the strategies that work across trade-offs at the same time.
What Fuller imagined in 1961, we might now be capable of building in 2026.
What AI Actually Contributes
To be clear, my claim isn't "AI will solve climate change." That's a category error. AI doesn't negotiate treaties or build solar panels or approve science grants to perfect fusion.
The precise claim is: AI can model the interdependencies across all 169 SDG targets at a scale and speed no human committee ever could. It can explore what economists call the Pareto frontier: the set of strategies where you can't improve one variable without making another worse. [8] Human planners have to pick a point on that frontier based on politics and intuition. An AI optimizer can show you the whole frontier at once. Simple but effective strategies like Karpathy's autoresearch can test thousands of policy changes overnight for effectiveness IF the model can capture the interdependencies sufficiently. [9]
This is Fuller's World Game, finally playable.
An agent swarm with access to global economic, environmental, and demographic data, along with the research on how things are correlated, doesn't need five years of diplomatic cycles to test a scenario. It runs the scenario in seconds. It runs ten thousand variants. It surfaces the ones where reducing child mortality, cutting emissions, improving food security, and increasing economic development all point in roughly the same direction. The strategy space most human negotiators never find because they stop looking after the first plausible option.
The unique contribution of AI isn't solving any single problem. It's connecting them. Operating across the interdependencies the way no human committee ever could.
Fuller's Prediction
Fuller believed something specific that most people gloss over. He thought that if you had enough information and sufficient processing power, you could identify strategies that would "work for all without disadvantaging any."
That sounds utopian. It's not. It's an engineering specification.
He was describing a constrained optimization problem. The constraints are the hard limits of physics, ecology, and carrying capacity. The objective function is human flourishing, defined broadly. The variables are global resource flows. The output is a set of strategies that move all the constraints in the right direction simultaneously.
Fuller wasn't wrong about the architecture of the problem. He just lived before we had the tools.
The Club of Rome's modelers in 1972 were attempting something similar with the computers of their era. The Millennium Project's researchers are attempting it now with human analysts. What changes with modern AI is the scale and speed of the search. A trillion-parameter model running a swarm of agents isn't smarter than the best human systems thinkers... but it is faster. It holds more variables without dropping any. It doesn't get tired or political or anchored to last year's assumption.
The Part AI Can't Do
Obviously, this is complicated (to radically understate the situation).
Fuller's World Game had a values problem baked in from the start. "Make humanity win" sounds like a clear objective. It isn't. Humanity disagrees on what winning looks like. Economic growth vs. environmental preservation. Individual freedom vs. collective constraint. Present consumption vs. future capacity.
AI optimizes for what you tell it to optimize for. The values question remains irreducibly human.
If you point an agent swarm at the SDGs and tell it to optimize for all 169 indicators simultaneously, it will. But someone made choices about how those indicators were defined, which ones get weighted more heavily when they conflict, and whose data counts. Those choices embed values. The optimizer amplifies them.
This is not a reason to avoid the technology. It's a reason to be deliberate about the inputs. The computing capacity is finally here. The wisdom question is older than Fuller -- and won't be resolved by any model size.
What AI gives us is a mirror at planetary scale. It reflects the strategy space with unprecedented fidelity. Whether we like what we see, and what we choose to do about it, still requires humans in the room.
What Fuller Might Do With an Agent Swarm
Fuller was an architect by instinct. He thought in systems. He prototyped obsessively. He believed that the right structural design solved social problems without requiring behavioral change. The geodesic dome wasn't beautiful by accident. It was optimization of space vs materials--structurally efficient and cheap to build, which made it accessible, which made it a solution to shelter scarcity. Function led to form led to impact.
He would look at a modern agent swarm and immediately ask: what's the dome? Not "how do we use this tool" but "what structure does this make possible that was impossible before?"
His answer, I think, would be: the first near-real-time global resource optimizer. Not a report. Not a model. An actual system that continuously ingests data about food, energy, water, migration, climate, and economic flows, and continuously surfaces the interventions with the highest cross-domain leverage.
The World Game, running live. On Spaceship Earth. Finally with enough compute to play it seriously.
The Bottom Line
Fuller had the right question in 1961. Every generation since has refined the catalog of answers we need to find. We've never lacked ambition, intelligence, or precision in describing the problem.
We lacked the machine. Now we have the machine, and all that's lacking is the will.
AI's contribution to humanity's grand challenges isn't solving them one by one. It's finally giving us a tool capable of thinking about all of them at once. Modeling the trade-offs. Mapping the Pareto frontier. Showing us strategies that work across interdependencies the way no human committee, no five-year diplomatic cycle, no set of 231 indicators managed by spreadsheet ever could.
The catalog has been ready for decades. The computational capability is now here.
Fuller created the World Game. We finally have the tools--we should build the board.
What's your take: is global-scale AI optimization a genuine step toward solving humanity's grand challenges, or are we just building a faster way to argue about the same trade-offs? Drop your perspective below.
References
- Fuller, R. B. (1969). Operating Manual for Spaceship Earth. Southern Illinois University Press.
- Fuller, R. B. (1981). Critical Path. St. Martin's Press.
- Meadows, D. H., Meadows, D. L., Randers, J., & Behrens, W. W. (1972). The Limits to Growth. Universe Books.
- World Commission on Environment and Development. (1987). Our Common Future (Brundtland Report). Oxford University Press.
- United Nations. (2000). United Nations Millennium Declaration. Resolution A/RES/55/2. https://www.un.org/millennium/declaration/ares552e.htm
- United Nations. (2015). Transforming our world: the 2030 Agenda for Sustainable Development. Resolution A/RES/70/1. https://sdgs.un.org/2030agenda
- Glenn, J. C., Gordon, T. J., & Florescu, E. (2022). State of the Future 22.0. The Millennium Project. https://millennium-project.org
- Vinuesa, R., Azizpour, H., Leite, I., et al. (2020). The role of artificial intelligence in achieving the Sustainable Development Goals. Nature Communications, 11(1), 233. https://doi.org/10.1038/s41467-019-14108-y
- Karpathy, A. (2025). autoresearch — an agent-driven autonomous research loop concept described by Andrej Karpathy in public posts and demonstrations (2025).
If this resonated, here are some related articles:
- The AI Bullwhip: What The Beer Game Teaches Us About Uneven AI Adoption (how interdependencies in systems cause small imbalances to cascade — exactly the dynamic Fuller was trying to model): LinkedIn | Substack | Medium
- We're Linear Thinkers in an Exponentially-Changing World (why human intuition consistently fails when the variables compound — and why Fuller needed a computer, not a committee): LinkedIn | Substack | Medium
Keith MacKay is a technology strategy consultant and CTO in EY-Parthenon's Software Strategy Group (SSG), specializing in AI disruption and technology diligence for private equity and corporate clients. SSG's AI Disruption Lab conducts rapid assessments of how AI transforms and threatens existing business models and value chains. Keith teaches at Northeastern University and writes about strategy, management, and AI/technology, with an AI collaborator.