Skip to content

Research intelligence ​

Collect versioned AI and computing research metadata, exact-match OpenAlex enrichment and original TNL research briefs. Public reporting is available at Research Intelligence.

Authentication ​

Use your existing TNL API key in Authorization: Bearer YOUR_API_KEY. Email verification and plan quota rules apply to every /v1/intel/research request. The API returns 401 for missing/invalid keys, verification restrictions or exhausted quota, 400 for invalid filters, 404 for unknown IDs and 503 for temporary unavailability.

Endpoints ​

EndpointResponse
GET /v1/intel/researchLatest observed version of each paper, with enrichment when available
GET /v1/intel/research/:idPaper metadata, revision hash and latest enrichment
GET /v1/intel/research/:id/versionsPermanently retained observations of source revisions
GET /v1/intel/research/:id/briefsOriginal briefs for the paper, including earlier reports

IDs use arxiv- plus the arXiv identifier, with legacy slashes replaced by hyphens; for example arxiv-2611.05377. Discover actual IDs from the collection. Paper identity is stable across versions and cross-listings.

Collections return { data: [...], nextCursor: number | null, schemaVersion: 1 }. Pass nextCursor as cursor for the next page. Defaults: 25 rows; maximum: 100. Ordering is descending ingestion sequence; this is an evolving collection, not a fixed snapshot. Version collections retain source versions and later observations without overwriting earlier evidence.

FilterMeaning
subjectcs.AI, cs.LG, cs.CL, cs.CV; matches any source category
since, beforeInclusive lower / exclusive upper ISO timestamp bounds; source update time for papers/versions, TNL publication time for briefs
qCase-insensitive title substring, up to 120 characters
limit, cursorPagination controls
sh
curl 'https://theneuralledger.com/v1/intel/research?subject=cs.AI&limit=10' \
  -H 'Authorization: Bearer YOUR_API_KEY'

Data semantics ​

Paper fields include stable identity, source version, title, authors and reported affiliations, categories, submission/update/observation timestamps, version-specific source/PDF URLs, DOI and journal reference when provided, and author-linked repository/model URLs. Raw abstracts and full texts are withheld from the commercial API. Resource links are author-provided associations; they do not establish ownership, endorsement or independently verified reproducibility.

publicationStatus is preprint, journal_reference_reported or withdrawal_reported. peerReview is always not_independently_verified. A journal reference or indexed DOI does not establish verified peer review.

Enrichment is nullable. A record is attached only through an exact DOI or arXiv location match, never title similarity. It includes OpenAlex work ID, source URL, observation time, citation count (nullable), reported institutions, related work IDs and retraction flag. Citation counts are dated observations of attention, not measures of scientific validity. Missing enrichment is not a citation count of zero. OpenAlex may not have indexed a newly released paper.

Original briefs ​

Only selected papers become posts. A bounded automated editor and separate grounding review assess the abstract and metadata. Findings/methods retain supporting excerpts; implications must be conditional and limitations explicit. Failed or abstained generation leaves a paper available as data without a post. Briefs carry evidence scope, model, publication time, source revision and any prior brief ID.

Existing brief URLs and evidence remain permanent. current and notice expose later source changes, withdrawal/retraction observations and newer versions alongside the original report. These flags depend on source updates we have collected; they are not a comprehensive retraction registry. Research is initially limited to AI/computing. Collection is daily-window based and paced; no real-time coverage guarantee is made.

AI/ADES digestion ​

Every retained paper revision and OpenAlex observation enters the shared source intelligence pipeline. Paper collections, detail and version responses expose intelligence, including sourceKey, processing status, completed/total evidence groups and validated results. Original briefs display this as an additional intelligence section. New briefs wait for completed digestion; older publications remain accessible while backfill runs.

Research implications are modeled scenarios, not experimentally established commercial or market effects. An affiliation does not establish ownership, a citation does not establish validity, and an unresolved ADES match remains unresolved. Bibliographic observation digestion is retained separately from the paper's source-revision digest.

The Neural Ledger API