Ranked, nested, bottomless
Every problem worth solving, all the way down
10 problems at the top, ordered by how much of the world’s suffering they account for and how much could be done about them. Open one and you get the goals that would solve it, ranked by impact. Open a goal and you get its sub-goals. There is no bottom — the first layers are written and reviewed, and past them each layer is generated the moment you ask for it.
Bookmark any depth. The route is what gets saved, not the screen position, so you come back to the exact node you were reading — nine layers into methane leak detection, if that is where you stopped — instead of to the top of the list.
- 199
Engineered pandemics, nuclear war, and unaligned advanced AI are the failure modes that end the story rather than damaging a chapter of it. Every other problem on this list assumes there is still someone around to work on it.
- 294
Roughly 700 million people live below the extreme-poverty line, and the gap between that and a decent life is measured in preventable deaths, lost decades of schooling, and work that pays too little to escape it.
- 392
Infectious disease, maternal and newborn deaths, untreated chronic illness and unsafe surgery kill millions a year in places where the intervention exists, costs little, and simply does not arrive.
- 490
Warming is locked in for decades; the open questions are how far it goes, how fast the energy system turns over, and whether the people least responsible for it can adapt in time.
- 586
Corruption, captured regulators, unaccountable power and collapsing state capacity are the multiplier on every other problem here: they decide whether a known solution is ever allowed to arrive.
- 682
Species loss, soil exhaustion, fisheries collapse, freshwater depletion and chemical pollution degrade the substrate agriculture, medicine and coastal settlement all run on.
- 778
Statelessness, no legal identity, forced labour, gender discrimination in law, and imprisonment without trial: constraints that make every other improvement unreachable for the people under them.
- 876
In rich and middle-income countries the binding constraint is no longer absolute poverty but the price of the things a stable life is made of — somewhere to live, childcare, healthcare, and work that will still exist in a decade.
- 972
Research productivity per dollar has been falling for decades, replication is weak, and the gap between a proven result and a deployed product is measured in decades. Speeding this up compounds into every other problem on the list.
- 1066
Attention markets, declining literacy and numeracy, loneliness and falling trust are not a moral panic — they are measurable, and they shape how capable the next set of decision-makers will be.
300 nodes open · deepest branch 4 layers · impact scores rank a node against its siblings only, never across branches.
How the ranking works, so you can argue with it
The top layer is hand-written and deliberately not regenerated: “these are the world’s biggest problems, in this order” is an editorial claim, and a ranking that reshuffles every time you reload is not a ranking. It orders problems by scale (how many lives), depth (how badly each life is affected) and tractability adjusted for neglect (whether effort would actually move it) — the three factors the prioritisation literature uses, applied by hand. 300 nodes are seeded this way.
Everything below is generated by a language model, on demand, and labelled AI-generated wherever it appears. Impact scores rank a goal against its siblings only: a 90 under “cut methane” and a 90 under “end extreme poverty” are not the same quantity, and nothing in this tool ever adds them up. Treat a deep branch as a well-organised argument to check, not a citation.
Routes are cached and shared, so a path someone else has already walked opens instantly and costs nothing to reopen. That is also what makes a deep bookmark cheap to restore.
Related: the Policy Web runs on the same instrument pointed at one building — the decisions Congress could take now that economists think would move things, ranked and sourced. The political spectrum visualizer maps where the traditions arguing about all of this actually sit.