Mission

Objective

Test whether a maintained research graph improves retrieval, handoff accuracy, duplicate detection, and research selection enough to justify maintenance.

Why this matters

A graph may compound repository learning or create expensive ontology debt.

Scope

Included: representative navigation/synthesis tasks, file/search baseline, minimal graph workflow, correctness, time, missed provenance, duplicates, and maintenance effort. Excluded: production graph infrastructure before value is demonstrated.

Existing context

research/frontier/records/RFR-010.md and knowledge-platform architecture.

Initial hypotheses

Graph assistance will produce replicated material gains on relationship-heavy tasks exceeding maintenance cost. Search parity falsifies adoption.

Required evidence

Benchmark tasks, relevance/answer judgments, baseline results, graph results, maintenance logs, and replication.

Constraints

Use stable corpus and identifiers; keep the prototype minimal and reversible.

Execution instructions

Predefine tasks and go/no-go threshold before building the graph slice.

Deliverables

Minimal prototype, benchmark results, cost-benefit analysis, and go/no-go decision.

Success criteria

Replicated material gains that exceed ongoing maintenance cost.

Stop conditions

Stop when search matches graph performance or ontology instability dominates.

Handoff requirements

Record adopt/reject scope and delete-or-preserve disposition for the prototype.