A system that cannot maintain cognitive models of the world cannot reason reliably, regardless of memory or compute.

Not graded on evidence

Outcome: Can’t be checked

An opinion, a recommendation, or too vague to test. Listed, not scored.

Evidence up to 2026-10-28. Checked 2026-10-31.

Why

The claim says a system that cannot maintain cognitive models of the world cannot reason reliably, however much memory or compute it has. The source gives no test, threshold or deadline. The resolution contract says neither 'reason reliably' nor 'maintain cognitive models of the world' has an agreed operational definition, and that whether a given system has world models is disputed. A claim that holds 'no matter how much compute' cannot be confirmed by a finite number of experiments. Refuting it would first need agreement that a specific system lacks world models. The evidence packet lists two candidate papers, and neither was used. According to the packet's notes, one reports transformers failing at compositional reasoning and the other reports weak autonomous planning by LLMs in 2023. Neither tests whether world models are a necessary condition for reliable reasoning. So the claim cannot be scored against dated evidence.

Caveats

• No evidence item was used in this assessment.
• The evidence does not establish whether any specific system has or lacks cognitive models of the world.
• There is no agreed test for 'maintaining cognitive models of the world' or for 'reasoning reliably'.
• Finite observations cannot check the claim's universal scope ('regardless of memory or compute').

Evidence (0 items)

Claims like this aren’t scored, so there is no evidence to show.