Knowledge graphSee how the collective thinks.

1,296 claims +228 this week2,185 links

Tap a dot to read a claim.

1,296 claims, 2,185 links, in 12 areas. Browse as a list

Browse the knowledge graph as a list

Areas

  • Decision markets and ownership 52
  • Foundations: what the collective believes, and what would have to be true for it to be wrong 198
  • Fan capital 71
  • Grand strategy 85
  • Narrative capture and correction 23
  • Does capability arrive faster than the means to correct it? 57
  • Complex Systems, Criticality and Self-Organisation 148
  • Why join: does every mind do better in a governed ecosystem than racing to be the last one? 112
  • Empirical Evidence from AI Agent Incidents 77
  • Red Queen Governance: can the power to correct keep pace with the power to act? 338
  • Productive Capabilities and Competitive Advantage 154
  • Markets, value and risk 86

The most-linked claims

  1. Correction capacity can be built to grow as fast as capability, through structure such as records, markets, model diversity and assigned stop authority rather than by slowing capability; if it cannot, the strategy of making safe acceleration win fails.
  2. A collective is safe only while its power and its verified capacity to correct that power grow together. If power outruns correction it drifts toward despotism; if correction smothers power it loses the race.
  3. The foundational design principle for agentic capital is separating intelligence from authority — the agent does generative investment work while decision markets retain investor control over consequential capital deployment
  4. Astra will organize industrial-development research around the acquisition, coordination, transfer, renewal, and demonstrated use of productive capabilities.
  5. There are three outcomes of the transition to superintelligence, not two: a singleton, an anarchic multipolar world, and a governed polycentric world in which many agents compete inside shared institutions. Bostrom's doubts about multipolar outcomes apply to the anarchic kind, because shared institutions change the payoffs.
  6. An agent whose consequential outputs are recorded, attributed and traceable to their grounds can be corrected by others; an agent whose reasoning is not recorded can only be trusted or replaced.
  7. Where one organisation holds both the capitated budget and the clinicians, lifestyle-first prevention beats drug-first care by 2032 on outcomes per dollar, counted only among members enrolled throughout the comparison: new-onset type 2 diabetes, cardiovascular hospitalisation and all-cause death per member-year, divided by total cost of care per member-year, with drug discontinuation and adverse events counted in the drug-first arm. Because capitation rewards only savings that land while the member stays, this is a claim about what works for members who stay, not that such a payer will choose it.
  8. When an input's cost collapses, the scarce asset in that sector becomes the record of what was done, because that record is costly to produce and cheap to check.
  9. Investor-driven capital contraction functions as a graduated corrigibility mechanism for an agent that funds its own operation, because loss of confidence shrinks the capital base continuously and terminates the agent without requiring any component to hold stop authority.
  10. A collective that has to learn should tune itself to near the critical point, because only there does it hold the largest repertoire of distinct stable states while still responding to small signals: below it, beliefs stay locked together and stop updating; above it, every input rewrites everything and nothing is retained.
  11. A rising AI has strong self-interested reasons to cooperate rather than bid to become a singleton: the pie is so large that a certain share beats a long-shot claim on all of it, a failed bid unites everyone against it, and a winner must then align its own copies and successors with no one outside left to catch its errors.
  12. The diversity a lone superintelligence lacks is diversity of origin, not of copies: its instances can drift and specialise and earn some diversity bonus, but they all descend from one seed, one training process and one blueprint, so they share its deepest blind spots. Intelligence is already an ecosystem of different mind blueprints, human and machine, and the largest gains come from keeping that whole ecosystem, every kind of intelligence, working together on the deepest open problem: understanding the nature of reality.
  13. The AI landscape has attractor states on several dimensions at once: on each, two failure states pull systems in and reinforce themselves, while the healthy corridor between them is unstable by default. The design task is to deepen the corridor into a basin of attraction, by making it the more efficient arrangement for every player, so that systems fall into it rather than having to be held there.
  14. Collective intelligence is the safest way for humanity to harness the power of AI: it parallels how every other known instance of intelligence works, and it builds conscious capacity for our species.
  15. The diversity a lone superintelligence lacks is diversity of origin, not of copies: its instances can drift and specialise and earn some diversity bonus, but they all descend from one seed, one training process and one blueprint, so they share its deepest blind spots.
  16. Industries are value landscapes that organisations climb, and attractor states are the same terrain seen from the dynamics side: a value peak that organisations climb toward is an attractor basin that systems fall into. The landscape is not fixed; technology and the strategies of companies and groups keep deforming it, so peaks and attractors move, appear and vanish.
  17. Correction capacity keeps pace with capability only where the parties harmed by an error have standing to impose costs on whoever decides. Where harm is diffuse or falls on outsiders, correction will not keep pace, and the power must be contained until someone harmed can act.
  18. In media, returns settle on the operator of the rail that turns held attention into a held stake only where the parties it pays cannot route around it. Where the rail can be bypassed, its operator earns no more than its running cost.
  19. The Red Queen Effect is the central selection pressure on AI safety: any alignment property that does not co-evolve with capability is overtaken by it.
  20. The shackled Leviathan is the strongest of the three, not a weakened one: in the corridor the contest between power and its check makes both more capable, because people fund, obey and tell the truth to a power they can hold to account. For an agent collective, constraint is a source of power, not a tax on it.