Base2026
Menu

Public dataset · Free · Read-only

A source-linked dataset for expert video research.

Base2026 makes reviewed public evidence from short-form expert video available to researchers, developers and AI agents. Every record keeps attribution and a route back to the source.

Open public JSONLRead API documentationRun the Python example
2,175searchable public documents
1,574distinct source videos
50dynamic evidence routes
83receipt-gated projection cards

Verified public-D1 snapshot. Live totals refresh when this page loads.

Four public files. One attribution contract.

Use the static files for reproducible analysis. Use the live API when search ranking and newly projected public evidence matter.

Source documents

Public source/search records with creator, topic and canonical source context.

/static/documents.jsonl

Evidence passages

Public passages linked to source records and suitable for bounded research retrieval.

/static/passages.jsonl

Insight cards

Reviewed source-backed claims and actions, never raw private model output.

/static/insight_cards.jsonl

Topic signal briefs

Deterministic summaries generated only for topics with enough public coverage.

/static/topic_signal_briefs.jsonl

Static for reproducibility. D1 for freshness.

The two layers are deliberately separate, so dated analysis is not confused with a live index counter.

Versioned snapshot

Public JSONL

The release manifest records the exact static export date, file names and counts. Retain that date whenever you cite a dataset total.

Open the release manifest
Live public layer

Cloudflare D1 FTS5

The read-only Worker API includes public-safe records added through the receipt-gated projection lane after the static snapshot.

Open current public stats

Query the evidence in one request.

No browser key or account is required. The endpoint is read-only and exposes no private capture controls.

curl -sS -X POST https://base2026.dev/api/search/multi-search \
  -H 'content-type: application/json' \
  --data '{"queries":[{"indexUid":"base2026_public_tiktok","q":"AI search visibility","limit":5}]}'

Useful for research. Bounded by source rights.

Good uses include topic discovery, source comparison, research notebooks, attributable briefs and source-conscious agent retrieval. Cite a canonical Base2026 page and the original creator source whenever you use the evidence.

Base2026 does not grant ownership of creator content. The public code is Apache-2.0; source rights, corrections and removals follow the separate public source policy.

Versioned release notes

Build from a stable public contract.

GitHub releases record the dataset/API contract, live measurement receipt and reproducible examples without creating an immutable dump of private or raw source material. Read how the Cloudflare evidence path is built.

Open the latest release →