Human input to guarded output

Test an idea before the graph is allowed to believe it.

This workbench lets a human enter a research question, source clue, and claim fragment, then turns it into a review note, action path, graph sketch, and exportable JSON while keeping metadata, sampled evidence, and ontology claims visibly separate.

Start the Workbench

Loading public project data...

Input

Human research signal

Research question Source or clue title Claim or observation

Evidence grade Layer

Text status Confidence

Functional role

Output mode

Graph simulation

Inputs become traceable nodes

draft evidence layer guardrail

Draft node selected

Generate an output or load a queue sample to see how a human input connects to evidence, layer, action, and export.

Project queue

Open items stay visible until the evidence improves

Static human workbench. It reads only public/data/*.json and exports browser-generated notes. It does not write to the database, ledger, raw corpus, or filesystem.