SOL Lens
Local replay × SOL Engine
SOL = Self-Organizing LogosEngine: research foundation · Lens: semantic trace workbench · Not the GPT-5.6 Sol model name
Demo fixture · v0.2

Semantic migration workbench

See whether your agent actually got better.

Choose a curated SOL packet, explore its typed semantic graph, and replay evidence, coherence, contradiction, and promotion readiness without hidden reasoning claims.

Suggested judge tour: Grounded answer → Tool fan-out → Conflicting sources

Demo fixturebuild-week-agent-migration-01build-week-agent-migration-01

Reference agent

Checked-in reference agent ready

Evidence0.79
Coherence0.76
Contradiction0.12
Atomic units12
SOL courtHOLD

Candidate agent

Locally replayed candidate ready

Evidence0.91
Coherence0.88
Contradiction0.07
Atomic units10
SOL courtPROMOTE
Observable trace · 10 atomic units

Semantic Logon graph

Recomputed locally
SupportedInferredContradiction
49%
User intentL01Parse constraintsL02Retrieve evidence AL03Identify entitiesL04Retrieve evidence BL05Evidence synthesisL06ContradictionL07Answer draftL08Policy checkL09Final answerL10

Evidence synthesis

ρ 0.93p 0.15ψ 0.90

Reconciles supported and inferred claims into one observable evidence path.

Candidate agent trace

inferredDemo fixture

Evidence

ρ
0.91
Coverage0.50
Authority0.94

Coherence

j
0.88
Continuity0.90
Faithfulness0.88

Contradiction

Δ
0.07
Rate0.07
Severity0.74
Promotion court · replayed

PROMOTE

Evidence and constraint gates were replayed locally. The complete graph is ready for export and independent replay.