Fractal LLM Locality Visualizer
A deterministic browser-local research instrument.
54839a93- LOCAL_DENSE
- COARSE_CONTEXT
- EVIDENCE_PINNED
- PRUNED
- FRONTIER_SELECTED
Each control executes a declared browser-local transition. Repeated dual-outcome controls alternate deterministically.
NOT YET EXERCISEDEvidence anchors cannot be pruned by a convenience threshold.
NOT YET EXERCISEDTraversal changes order, not canonical node identity.
NOT YET EXERCISEDDisplayed cache/locality metrics are labeled proxies unless measured on hardware.
NOT YET EXERCISEDQuery selection exposes omitted nodes.
No transition yet. Operate a declared control.
This transparent browser-local instrument teaches the authored state contract. Exercising an assertion demonstrates fixture behavior only; it is not evidence of production performance, compression ratio, model uplift, cognition, safety, or universal correctness.
Purpose
Explain locality-preserving ordering, multiscale neighborhoods, sparse selection, Q-Frontier prioritization, and evidence-preserving context reduction.
Mechanism
Tokens or graph nodes are placed on a synthetic manifold. Users compare row-major, Z-order, and Hilbert traversals, then inspect local dense blocks and coarser long-range summaries.
Deterministic fixture
A 256-node project graph with clustered topics, dependencies, evidence anchors, and a query targeting one active objective.
The fixture must live in content/demos/d09-fractal-llm-locality/fixture.json, validate before the demo renders, and be resettable without a network call. A future live adapter may be added behind a visibly separate mode.
Required controls
- Choose traversal
- Change scale
- Set local radius
- Enable Q-Frontier
- Pin evidence anchors
- Adjust sparsity
- Inspect moved nodes
- Compare cache-line proxy
Visible states
LOCAL_DENSECOARSE_CONTEXTEVIDENCE_PINNEDPRUNEDFRONTIER_SELECTED
Every state must have text, iconography, and a non-color-only distinction. State transitions are logged in an in-memory demo receipt visible in the inspector.
Core assertions
- Evidence anchors cannot be pruned by a convenience threshold.
- Traversal changes order, not canonical node identity.
- Displayed cache/locality metrics are labeled proxies unless measured on hardware.
- Query selection exposes omitted nodes.
Layout
The page contains:
- a concise mechanism explanation;
- the interactive stage;
- a state and evidence inspector;
- a reset control;
- a “what this proves / what it does not prove” panel;
- links to the related paper and benchmark contract.
On narrow screens the inspector becomes a bottom sheet. The stage must remain usable at 360 CSS pixels.
Accessibility
- All operations are keyboard reachable.
- Pointer gestures have button or keyboard equivalents.
- Animated transitions honor
prefers-reduced-motion. - Focus remains visible and returns predictably after sheets close.
- Diagrams expose concise text alternatives.
- Status is announced through a polite live region only when the user initiates the transition.
Instrumentation
Record locally:
demo_id
fixture_version
action
prior_state
result_state
validation_result
timestamp_relative
Do not send telemetry. Provide “Export demo trace” as JSON for debugging and reproducibility.
Automated tests
- fixture schema validation;
- deterministic reset;
- every required control transition;
- every assertion above;
- invalid and stale fixture handling;
- keyboard path;
- 360 px and 1440 px screenshots;
- zero critical console errors;
- no request to a paid API in default mode.
Publication boundary
This is an explanatory and experimental visualizer. It does not claim production inference speedup.
The interface must never convert illustrative values into a benchmark claim. A measured result appears only when linked to a versioned run receipt.