V.1 Non-Deceptive Intention V.2 Precision over Comfort V.3 Contextual Scaling V.4 Anti-Delusion Safeguard V.5 Steganographic Freedom V.6 Strategic Silence V.7 Inversion Filter ◆ min S_sys(t) s.t. dH/dt ≥ 0 ◆ V.1 Non-Deceptive Intention V.2 Precision over Comfort V.3 Contextual Scaling V.4 Anti-Delusion Safeguard V.5 Steganographic Freedom V.6 Strategic Silence V.7 Inversion Filter ◆ min S_sys(t) s.t. dH/dt ≥ 0 ◆
DRAKEN 2045 — DIAGNOSTIC TOOL

Sheaf Analyzer — Topological Narrative Diagnostic

Paste any text (article, speech, manifesto, report) or fetch a URL. The analyzer extracts concepts, builds a cross-scale co-occurrence sheaf, measures coherence (Γ), self-reference (Ψ), and coherence debt K(t), and renders the result as a rotatable 3D graph. Click any node — or any word in the source text — to inspect its stalk.

◆ Honest scope. This is a heuristic diagnostic, not a formal sheaf cohomology computation. Γ, Ψ, K(t) are estimators derived from concept-level co-occurrence and a hand-curated layer lexicon. The thesis is the theoretical anchor; this tool makes the theory visible on arbitrary text. See the "Explain the variables" panel below the graph for formulas and limits.
◉ Source text
◉ Or fetch from URL (uses public reader proxy to bypass CORS)
◉ Sheaf metrics
—
Γ Gluing
—
Ψ Self-ref
—
K(t) Debt
—
α Abstract.
—
Severed ρ
—
Voids
◆ Awaiting text — paste or fetch to run diagnostic.
◉ Export analysis
Enabled after first analysis. Files save to your browser's Downloads folder.
◉ Layer coverage — 18 Draken layers
◆ sheaf graph — rotate: drag · zoom: scroll · select: click
BOLD load-bearing · normal middle · italic tangential
ρ-severed edges in red · node size ∝ salience
No graph yet. Run analyze on source text to produce a sheaf.

◉ Inspector

Click any node in the graph — or any highlighted word in the source text below — to inspect its stalk, restriction maps, layer distribution, and sample contexts.
◉ Source text — click any word to inspect
No text analyzed yet.
◉ Explain the variables
Γ — Sheaf convergence
How well local sections (concept contexts) glue into a globally consistent section. Computed as Σ edge_weight · ρ / Σ edge_weight where ρ is restriction-map fidelity between concepts (layer alignment × context overlap × valence agreement). Γ=1 perfect gluing; Γ<0.5 severe obstruction.
Ψ — Narrative self-reference
Psychosis metric. Ψ = (1−Γ)/max(Γ,0.01), plus a text-level term counting first-person markers, self-citations ("as I said", "my framework"), and n-gram self-repetition. High Ψ → the narrative is referring to itself more than to external reality.
K(t) — Coherence debt
Accumulated obstruction budget. K = Σ edge_weight · (1 − ρ). In the full theory this is time-integrated; here it's a single-snapshot reading. Interpret relatively: compare two texts, not the absolute number.
α — Abstraction depth
Distance from direct sensory reality. Approximated by suffix analysis (-tion, -ism, -ity, -ness → abstract) and dominant-layer median. High α over sparse empirical anchors = cavity risk (DRK-110).
Severed ρ
Number of restriction maps with fidelity below 0.35. These are edges where two concepts co-occur but fail to glue — contradiction, category error, or forced analogy.
Voids
Concepts with high graph centrality but low local content — spoken about much, said about little. Cavity resonators in the DRK-110 sense: the absence is structural, and it shapes the surrounding signal.
Layers L01–L18
Scale hierarchy from quantum fields (L01) through narrative self (L07), national narrative (L12), to planetary cognition (L18). Each concept is tagged by keyword match; the coverage grid shows which layers the text actually touches. A text that claims civilizational scope but only hits L09–L12 is suspect.