Data & access
Two years of every English-language crypto headline, each scored and typed at ingestion, and a live news index per asset. Available as a data licence, as an API, or as a collaboration. Pricing and scope are discussed personally; a sample is sent on request.
What the data is
One row per headline. Collected around the clock since 4 October 2024, scored within minutes of publication, never re-scored afterwards, so the history is exactly what was known at the time. A full description of the dataset, volume, sources, timing, score distribution, latency and known defects, is research note 4.
| Field | Meaning |
|---|---|
| published_at | publication time, UTC, to the second |
| ingested_at | time the record was written, UTC: the point-in-time anchor, the headline was known from then on |
| scored_at | time the score and the type were written, usually within minutes of ingestion; never changed afterwards |
| assets | assets the headline is about, one or several |
| score | expected effect on price, −1 to 1, assigned at ingestion |
| type | proactive (something not yet happened) or reactive (something that already has) |
| source · domain | outlet and domain |
| title · url | title and canonical URL |
What the type adds to the score, and what it does not: note 5.
The index
The per-asset series built from those rows: 1-hour to 24-hour buckets, weights by score sign and type, with the full history and a live endpoint. The construction and what it predicts are in note 1 and note 2; the weights in production are not published.
How it is offered, and what it costs
| Access | What is included | Price |
|---|---|---|
| Trial | A 30-day slice, all assets, CSV or Parquet with a one-page data dictionary, under the trial agreement. Open three-month slice with prices and a notebook on Hugging Face → | free, on request |
| Developer | Live feed and index, the last 30 days only, one key, no history, no redistribution. For an individual or a team under five. | $500 / month |
| History | Two years of scored headlines from 4 October 2024 as a single CSV or Parquet file, internal use, up to five users. | from $15,000 |
| Data licence | Everything: the history, daily files, the live API feed and the index, up to five users, internal use. Redistribution to your own clients is a separate term or a revenue share. | from $3,000 / month$36,000 / year |
| Research | For a public paper that cites the data: usually free, with the full history. | free, on request |
Prices are the starting point for a standard licence; final terms depend on scope (assets, delivery, redistribution rights) and are set in a short licence agreement. For teams that want to evaluate the live feed before committing, a three-month paid pilot ($5,000) includes current data and API access and is credited in full against a subsequent licence. Invoices in USD or EUR.
Founding customers: the first three licences are $30,000 a year, fixed for two years. Investment or partnership: see the note on the creator page. →
Policies
The documents a data buyer's due diligence asks for. Each is a short PDF, in English and in Russian.
Request access
Say what you are after and where to answer. A reply comes from the creator, usually within a day, with a sample if you asked for one.