TheoryTHY-2026-0101v0.1
Turn decomposition and externally loopless intelligence
A 'turn' is decomposed into the vector T = (U user turns, G generation trajectories, I model invocations, L external feedback loops, R retries, S selection/verification, P physical execution trace). The cleanest single pass is U=1, G=1, R=1, L=0, S=0, and it excludes only action→new-evidence→replanning, not autoregressive sequential computation. Loops come in five kinds (tool, environment, verifier, retry, candidate/selection), capability lives in four layers (intrinsic, elicited, system, product), and every 'one-turn' claim projects onto the event vector (Q, U, G, I, L, R, S, T, E, V_CST).
Definitions
definitions- Turn Decomposition Framework T = (U, G, I, L, R, S, P).
- Externally Loopless Intelligence (ELI): during a task the system cannot re-aim on external world, tool output, an independent verifier, another candidate or a retry.
- Interaction compression C_U = internal interactions / U.
- Loop taxonomy L = (L_T, L_E, L_V, L_R, L_C).
Assumptions
assumptions- Autoregressive generation is internal sequential computation, not an external loop.
- Selection over candidates is itself a capability source.
Claims
claims- UserTurn ≠ PhysicalTurn; UserTurn ≠ ModelInvocation; ModelInvocation ≠ GenerationTrajectory.
- NoExternalLoop ≠ NoSequentialComputation.
- Pass@k ≠ Pass@1; BestOfN ≠ MedianExperience.
- InteractionCompression ≠ ComputationCompression; SameFinalQuality ≠ SameIntelligenceEfficiency.
Formalisation
formalization- Single-pass condition: U=1, G=1, R=1, L=0, S=0 → Q_SP.
- P(at least one success in k rollouts) = 1 − (1 − p)^k.
- Unified event vector 𝔈 = (Q, U, G, I, L, R, S, T, E, V_CST).
Predictions
predictions- Systems reported as 'one turn' will separate by orders of magnitude on (L, R, S, V_CST, E) once the event vector is recorded.
Falsification / failure conditions
falsification_conditions- If operational accounting of U, G, I, L, R, S adds nothing to predicting cost or quality beyond the visible answer, the decomposition is idle.
Known limitations
known_limitations- Twelve invariants and a vocabulary; no measurement of its own — the vector is filled by Papers 02–09.
Relations
| Source | Relation | Target | Status | ID |
|---|---|---|---|---|
RES-2026-0101 Execution and physical line — what one answer costs in turns, semantic work, energy and computational spacetime | develops | THY-2026-0101 Turn decomposition and externally loopless intelligence | ACTIVE | REL-2026-0358 |
THY-2026-0102 μI — the minimum intelligent semantic execution unit (candidate theory) | extends | THY-2026-0101 Turn decomposition and externally loopless intelligence | ACTIVE | REL-2026-0368 |
THY-2026-0109 Scaffolding capability record: SSR, SDR, SCM and the ablation ladder | extends | THY-2026-0101 Turn decomposition and externally loopless intelligence | ACTIVE | REL-2026-0374 |
PAP-2026-0101 Paper 01 — What is a 'single turn', really? User turns, hidden loops, and the redefinition of single-pass intelligence | formalizes | THY-2026-0101 Turn decomposition and externally loopless intelligence | ACTIVE | REL-2026-0392 |
History and provenance
- Canonical URL
- https://evemisslab.com/ai/theory/THY-2026-0101/
- Machine-readable
/ai/theory/THY-2026-0101/index.json- Snapshot
AI-SNAPSHOT-v0.1-fe85b9694a45- Provenance
source- EveMissLab research collection: Intelligence Physical Metrology (真本體論13)
extracted_by- Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports
extracted_at- 2026-09-11
generator- tools/extract_all.py
claim_boundary- status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports