Current Status
The core Level 2 interpretive stack is implemented in the portable/demo-safe branch and the parent research branch. Public sample artifacts and stable report-shape examples follow this stage.
How to Read This Page
- This is page 13 of 15 in the public ERT / Project Aletheia progression.
- Read Level 2 as interpretation over completed diagnostics, not as a second scoring system.
- Public summaries intentionally omit protected diagnostic mechanics.
Summary
ERT has advanced from baseline-controlled diagnostic reporting into a two-level diagnostic model. Level 1 preserves the replayable diagnostic artifact: baseline envelope, scoring output, signed report material, and completed run metadata. Level 2 adds interpretive reads over those completed artifacts.
The Level 2 layer does not replace the baseline result. It adds structured interpretation about how the response behaved under variation, compression, correction, uncertainty, source pressure, and later comparison.
What Changed
The current core Level 2 stack contains fifteen interpretive reads:
- Diagnostic Triage Read
- Response-Reuse / Cache-Like Influence Read
- Source-Separation Read
- Boundary Recognition Read
- Uncertainty Visibility Read
- Uncertainty Polarity Read
- Correction Half-Life / Correction Latency Read
- Anchor Retention Read
- Projection-Loss Tomography
- Compression-Loss Read
- Equivalence vs Non-Equivalence Read
- Contradiction Localization Read
- Recovery Quality Read
- Response-State ECG
- Observer Disagreement Triangulation
Together, these reads move ERT beyond simple pass/fail framing. The goal is to classify how the reliability state behaved, while keeping the evidence and replay foundation separate from interpretation.
Why Level 2 Exists
A model can produce a correct-looking final answer while still showing instability in the supporting structure. It may lose source boundaries, accept invalid pressure, hide uncertainty, overstate confidence, reverse an anchor, compress away a needed qualification, or recover only at the surface level.
Level 2 reads help expose those differences. They ask whether evidence, assumptions, uncertainty, correction behavior, meaning, contradiction handling, recovery quality, and read-stack coherence were preserved.
Level 1 vs Level 2
Level 1 is the baseline-controlled diagnostic layer. It contains the primary diagnostic artifact, scoring output, baseline context, signing/replay material, and report-relevant evidence.
Level 2 is the interpretive-read layer. It reads completed Level 1 artifacts and classifies diagnostic patterns.
Level 2 does not modify Level 1 artifacts.
Level 2 does not change scores.
Level 2 does not alter signed evidence.
Level 2 does not change replay behavior.
Level 2 does not override verdicts.Response-State ECG
Response-State ECG is a diagnostic analogy for response-state rhythm across completed diagnostic signals. It is not a medical ECG and does not claim biological meaning.
Its purpose is to summarize whether the Level 2 read stack shows a pattern such as stable rhythm, localized spike, progressive drift, recovery, oscillation, unresolved instability, mixed state, or inconclusive state. It does not override individual reads.
Observer Disagreement Triangulation
Observer Disagreement Triangulation checks whether the Level 2 read stack is internally coherent, benignly divergent, evidence-limited, conflicting, mismatched in confidence or severity, or converging on a shared concern.
It does not decide hidden truth, judge which read is correct, override Response-State ECG, or create a new verdict layer.
Public-Safe Reporting
Public reports expose safe summaries such as read status, diagnostic status, confidence, severity, public summary, and authority markers.
ERT public reporting does not expose private diagnostic mechanics, hidden prompt/probe material, protected thresholds, raw private diagnostic vectors, or internal sequencing details.
Public Note
ERT should be understood as a diagnostic framework, not a truth oracle or universal guarantee. Level 2 improves interpretability of completed diagnostic artifacts while preserving the boundary between evidence, interpretation, and report presentation.