The inference

The viability distinction supplies the formal possibility; the emergency case shows the three persistences separating; together they give two achievements where the operative-will picture saw one.

Critical questions

  • Is the case coherent? Yes.
  • Does it separate what the conclusion says comes apart? Yes, if the mapping holds.
  • The mapping premise. Applying viability needs a constraint set K, and the paper says K is “always indexed” but gives no non-arbitrary way to choose the maintenance constraints for an agent’s regulatory capacities. Unanswered, and it is the weakest step: “precarious” is relative to that choice.

Note of 2026-10-18: a second instance, in training

A31 of the exchange of 2026-10-17: “acquiring capacities ≠ governing the acquisition of capacities”. A conventionally trained network gains capacities while the governance of its updates stays with the optimizer and the trainer (CLAIM-367). That is this argument’s conclusion, effective action and the capacity to govern among regimes as different achievements, in a case where the second is held outside the system altogether. It is an instance, not a new premise, and the mapping question above applies to it unchanged: which capacities count as the network’s to maintain.