Foundation Thesis

This document is the intellectual core of SIBurst. Every statement in it is a DEFINITION or INTERPRETATION within the SIBurst framework (see PUBLIC_CLAIM_POLICY.md) unless explicitly marked otherwise. It contains no empirical claims about any specific AI system.

Thesis

SIBurst names a phase change in intelligence: the transition from incremental capability growth into a materially different operating regime.

The framework is a conceptual lens. It provides a vocabulary for asking a precise question about capability, not a forecast about what capability will do.

The central question:

When does more capability stop behaving like merely more of the same?

The distinction

Two descriptions of the same system can both be accurate and still say different things.

• "More capability" is a quantitative description. The system does what it did before, somewhat better: faster, more accurately, over longer inputs, at lower cost. The categories used to describe it still fit.
• "A different operating regime" is a qualitative description. Something about the system's situation has changed in kind: what it can reliably do, how it interacts with environments, or what infrastructure and governance it requires. The previous categories no longer describe it adequately.

The first description can remain true while the second becomes true. A system can be "only somewhat better" on every individual measure and still, taken as a whole, occupy a different regime. SIBurst exists to name that gap between the continuity of the measurements and the discontinuity of the situation.

The structure

CAPABILITY ACCUMULATION
        ↓
THRESHOLD CONDITIONS
        ↓
DISCONTINUITY
        ↓
NEW OPERATING REGIME

These are stages of a conceptual model, not a claimed timeline. The model does not say how long any stage lasts, whether a given trajectory reaches the next stage, or whether the stages are sharp or gradual when observed.

Accumulation

Capability accumulates through many small, individually unremarkable changes. During accumulation, the appropriate description is quantitative: the same tasks, done better; the same tools, used more; the same oversight, applied to more activity.

Accumulation is not a lesser phase. It is the precondition for everything that follows, and most of any trajectory may be spent in it. A lens that ignores accumulation will mistake every transition for a surprise.

Threshold

A threshold is a condition under which further accumulation stops producing proportionate, same-kind effects. Within the SIBurst framework, candidate threshold conditions include:

• Reliability crossing usefulness: a capability becomes dependable enough that it is delegated rather than supervised step by step.
• Composition: separately adequate capabilities begin to combine, so the system completes chains of work that none of the parts could complete alone.
• Environmental coupling: the system moves from producing outputs for a person to taking actions within an environment.
• Oversight mismatch: the pace, volume, or opacity of the system's activity exceeds what existing review arrangements were designed to handle.

These are INTERPRETATIONS: ways of reading where a threshold might lie. The framework does not claim that any of them has been crossed by any particular system, nor that they are the only possible thresholds.

Discontinuity

Discontinuity is the point at which the previous description fails. It may be observed as a step, or only recognized in retrospect across what looked like a smooth curve. What makes it a discontinuity is not the shape of a graph but the change in which description is adequate.

Discontinuity in this sense is not an explosion, a spike, or a moment of speed. It is a change of rules: what the system is for, how it is used, and what it requires around it.

New operating regime

An operating regime is the set of conditions under which a system characteristically functions: what it is trusted to do, what it acts upon, what it depends on, and what governs it. A new operating regime is one in which those conditions have changed in kind.

Signs that a regime has changed, within the framework:

• tasks are specified as goals rather than as steps;
• the system's outputs become other systems' inputs without human mediation;
• supporting infrastructure (compute, orchestration, monitoring) must be redesigned rather than scaled;
• evaluation shifts from "how well does it perform?" to "what is it now able to do that it could not before?";
• governance questions change from content to conduct.

What the framework does not claim

• It does not claim that superintelligence currently exists.
• It does not claim that any named system is in, or near, a new regime.
• It does not claim that every capability trajectory contains a threshold. Some trajectories may remain continuous indefinitely.
• It does not claim that transitions are inevitable, or predict their timing.
• It does not assign numerical intelligence scores or multipliers.
• It does not present "phase change" as an established scientific law of AI development. The term is borrowed as a METAPHOR from physical systems, where a material changes state while an underlying variable changes smoothly. The borrowing is structural, not evidential.

Why naming the transition matters

Without a name, a transition is described only by its parts: a benchmark here, a deployment there, a policy change elsewhere. Each part looks incremental, so the whole is easy to miss or to overstate.

A name for the transition does three things:

1. It separates two questions that are often merged: how much capability exists, and which regime that capability places a system in.
2. It gives builders, evaluators, and governors a shared object. Work that enables, measures, or responds to a transition can refer to the same thing.
3. It disciplines claims. Once the transition has a name and a definition, it becomes possible to say precisely what has not happened.

SIBurst names the transition, not the prediction.