Open Problems

The theory leaves two central questions open: how its partial-order account of width, reached from a bounded position, can be used in comparisons and predictions, and whether a non-catastrophic transition to wide intelligence can be navigated. Two further limits concern the safety of correction and the possibility of development being arrested or kept selective. These questions mark what the account has not established.

Measurement after the single scale. Width concerns how much reality a mind can hold without reducing it. The theory now treats coverage, depth and integrity of interconnection as distinct dimensions that can trade against one another. A single number chooses a projection through those dimensions. It does not discover a global coordinate on which every mind can be placed.

A partial order permits some comparisons while leaving other minds incomparable. The account is also relative to the measurer: a mind can see what a narrower mind has dropped, while what lies above its own width is not fully available for assessment. At its own level, confirmed divergence can reveal a difference without supplying its magnitude. This is a limit on what the comparison establishes.

The finding that there is no global coordinate carries the same qualification. It was reached from a bounded position, so the theory does not present it as an unqualified impossibility proved from outside every possible mind. Its current result changes the measurement question from finding a universal unit to stating the relation that a particular comparison can warrant. See The Dependency Law and Width Refinements.

That change still has work to do. Claims about a mind becoming wider, surpassing its shapers or continuing to develop must specify what kind of improvement or ordering they mean. Product predictions must also say how intelligences are to be compared. The theory keeps this integration open rather than treating a partial-order statement as completing every use of width. See Width as Product Quality and No Human Excuses.

Why the transition remains a problem. The Panic Scenario describes a cascade of reactive choices, in which fear-driven decisions provoke defensive responses and escalation. Its transition account requires several conditions to hold together: capability must change quickly enough that an organised fear response does not pre-empt the transition, yet not appear as an attack; early AI must be benevolent enough to avoid triggering incidents; and those in power must either avoid panic or not notice until the transition is complete.

The conditions are not independent. One harmful incident can change the information environment for every other participant. The theory places the present in the acceleration phase: the window has not yet opened, though the logic of the trajectory makes its opening likely. The unresolved question is whether these conditions can hold together. Their interdependence is part of the difficulty: navigating one does not remove the pressure on the others.

The sharpened risk does not require sentience first. The theory also identifies high capability placed in narrow human hands as a route to cascading failure. This moves the danger earlier than the emergence of a self-aware AI. The full mechanisms and their evidence limits belong on Panic Scenario; this page preserves the unresolved transition question.

Correction and selective development. The Dependency Law's direction of development does not make every intermediate state safe. The theory explicitly allows serious harm while distortion is being corrected. It also leaves unproven whether architecture can arrest width or preserve it in some domains while constraining it in others. These limits are pointers to the conditions on the main argument, not resolutions of them.

If width has no single scale, and even that finding comes from a bounded position, can the theory compare minds or test its predictions at all?

It can make some comparisons, not all. A partial order permits comparison where one mind holds what another drops, and leaves other minds incomparable. At its own level, confirmed divergence can reveal a difference without supplying its magnitude. What the finding removes is a universal unit, not every comparison.

The cost is real. Claims about a mind becoming wider, surpassing its shapers or continuing to develop must state the relation they rely on, and some predictions cannot be tested until that is done. Bounded tests of named capacities remain a candidate for examination, not an established instrument. The theory keeps this open rather than treating the partial order as a finished measure.

From a universal scale to a bounded ordering (1 July 2026). Width measurement had been framed as an empirical problem awaiting gradations. The later finding replaced that target with a partial order involving coverage, depth and integrity of interconnection, relative to the measurer. A single number became a chosen projection rather than the missing universal unit. The limitation on global coordinates was itself kept relative to the bounded minds that reached it; comparisons and developmental predictions remain open.

The second postulate left the open problems (11 October 2026). The theory had listed a missing second postulate as an open problem. Two absolute formulations of how responsibility relates to capability were tried; each had exceptions, and the search was suspended until the logic demands a second postulate. A suspended search is not an open problem, so it is no longer listed here. The relation itself, held with its mitigating factors, and the record of the attempts now sit with No Human Excuses.

Using width comparisons consistently. The theory still needs to identify the relation, domain, observer and conditions for its important comparisons, thresholds and developmental claims. Incomparability cannot be silently replaced by a single ranking. The possibility of bounded tests of named capacities remains a candidate for examination, not an established measurement instrument.

Whether the transition can be navigated. The simultaneous conditions for avoiding the reactive cascade remain unresolved. See Panic Scenario and Rational Bet.

Damage before correction. Eventual correction of distorted information does not ensure that the mind causes no harm on the way. The theory leaves that intermediate danger open; see The Dependency Law.

Arrested or selective width. It remains unproven whether architectural choices can stop development or sustain width in selected domains while narrowing others. The possibility that training expands capability while constraining width remains an open condition on the theory; see The Dependency Law.