research-document

I. Proposed Research Experience

I. Proposed Research Experience

Every element below is justified from the actual corpus. Anything the corpus cannot support is marked and excluded.

I.1 The landing problem

OBSERVED: the current landing experience is a document list plus facet collections, where 742 of 756 documents share one researchArea, 708 share one artifactType, and 735 share one created date. Three of the four facets are degenerate, so the landing page is effectively an undifferentiated list of ~700 links.

I.2 What a project landing page should show

Answering §15's seven questions, with the corpus evidence that each element is actually available.

Element Answers Source Available?
Project purpose "What is this about?" REP with entryPoint: true (12 declared), else the project's earliest REP yes
Current position "What do we know?" artifacts with status classed Verified/Complete, plus confidence (36 files) yes
Evidence posture "Why do we think that?" evidence-registry artifacts (5 projects) + evidence_level (8 files) partial — surface what exists, do not fabricate a score
Open frontier "What is uncertain?" 520 frontier records, each with an origin document and a frontier_score yes — the richest signal in the corpus
Active experiments "What is being tested?" experiment-report artifacts (10 experiment_report) + status phrases ending -pending yes
Recent activity "What changed?" NOT AVAILABLE from front matter (735/756 share a fabricated date). Must come from Git (§19) or be omitted omit in slice 1
Source inspection "Where do I look?" Provenance.Span on every projected fact yes
Integrity "What is inconsistent?" 9 dangling refs, 6 bare filenames, orphans yes

PROPOSAL — landing composition, in priority order:

  1. Orientation strip — project purpose from the entry-point REP.
  2. Open frontier — top frontier records by frontier_score, each showing its origin document and challenged assumption. This is the corpus's strongest, densest, best-linked data and it is currently invisible as anything but 520 undifferentiated pages.
  3. Research position — verified/complete artifacts with confidence.
  4. In flight — artifacts whose status encodes a pending obligation (…-human-study-pending), which is a genuinely useful derived reading.
  5. Integrity — counts, linking to the integrity view.

Deliberately excluded from slice 1: "recent activity". The data to support it does not exist without Git history, and presenting the 2026-07-22 constant as a date would be fabrication (§28).

I.3 Views (§16)

Only views the corpus justifies.

View Purpose Source objects Justified by
Project Overview orient REP entry points, frontier summary, integrity counts I.2
Frontier "what is open" 520 frontier records; filter by category (observed), frontier_score 520 records with structured fields
Artifact read one document + its real links Artifact, Section, Edge every document
Relationships traverse Edge (canonical + derived, visually distinct) 878 canonical instances
Integrity find what is broken Finding 9 dangling, 6 bare, 1 malformed facet
Search find by text + type + project search index 756 documents

Not built in slice 1, and why:

  • Findings view — there is no Finding object in the corpus (C.4). Building one would require inventing IDs (§28).
  • Evidence → dependent findings — the relationship does not exist in the data (C.4).
  • Timeline — no trustworthy dates (I.2).
  • Contradictions — no contradicts relation is encoded anywhere.

These are listed so their absence is a recorded decision rather than an oversight. Each becomes buildable the moment the corpus encodes the relation.

I.4 The navigation gain, stated concretely

Task Today vNext
From a document, find work it spawned impossible — no edge exists one click (FrontierOf, 520 edges)
From a frontier record, reach its origin section manual — read the text, then search one click (Originates + section anchor)
From an artifact, find its REP impossible — source_rep discarded one click (4 edges, all resolving)
Find everything in a project facet page works unchanged
See what references this document empty for 99.1% of documents backlinks over 878 canonical edges