| --- |
| title: README |
| emoji: 📉 |
| colorFrom: blue |
| colorTo: green |
| sdk: static |
| pinned: false |
| --- |
| |
| # MTS |
|
|
| **Livestreamed news. Open-source receipts.** |
|
|
| MTS is a daily livestreamed news/talk show on X — [@mtslive](https://x.com/mtslive) · [mts.now](https://mts.now). We cover the stories that actually matter: courtrooms, capital markets, geopolitics, internet culture. And we publish the underlying material as machine-readable datasets so anyone can fact-check, remix, or train on the exact same source we used to make the show. |
|
|
| Hugging Face is where our editorial work lives in its rawest, most reusable form. |
|
|
| ## Featured release |
|
|
| ### [`musk-v-altman-exhibits`](https://huggingface.co/datasets/mtslive/musk-v-altman-exhibits) |
|
|
| Every PX and DX trial exhibit from *Elon Musk, et al. v. Samuel Altman, et al.* — emails, Slack threads, text-message extractions, board memos, term sheets, IRS filings — with transcribed text, structured metadata, and our editorial commentary. **CC-BY-4.0.** |
|
|
| ```python |
| from datasets import load_dataset |
| ds = load_dataset("mtslive/musk-v-altman-exhibits", split="train") |
| ``` |
|
|
| ## Why this org exists |
|
|
| The work that goes into a single segment — pulling exhibits off PACER, transcribing 300 pages of filings, normalizing dates, cross-referencing names — is usually thrown away the moment the broadcast ends. We think that's backwards. The receipts should outlive the segment. So everything we use, we ship here. |
|
|
| Every dataset on this org is built from a public-interest editorial project, and every one ships with: |
|
|
| - **Structured metadata** — filter, group, query without parsing PDFs yourself |
| - **Transcribed / normalized text** — no re-OCRing the same source |
| - **Direct provenance** — links back to the primary documents |
| - **A real dataset card** — license, schema, known limitations, citation |
|
|
| ## In the pipeline |
|
|
| - **`ufo`** — declassified UAP filings, structured |
|
|
| ## Editorial principles |
|
|
| The same standard whether the source is a 600-page court filing or a 30-second viral clip: |
|
|
| 1. **Cite the primary document.** Always. Datasets without a `source_url` or equivalent don't ship. |
| 2. **Show the transcription work.** Don't paraphrase what you can quote. |
| 3. **Make it redistributable.** If we used it to make the show, you should be able to use it too. |
|
|
| ## License |
|
|
| Default is **CC-BY-4.0** unless a dataset card says otherwise. Use it commercially. Train models on it. Redistribute it. We just ask for credit (*"MTS — mts.now"*) and a link back. Underlying primary documents (court records, government disclosures, etc.) carry their own legal status, which we annotate per dataset. |
|
|
| ## Get in touch |
|
|
| - **Site:** [mts.now](https://mts.now) |
| - **Show:** [@mtslive on X](https://x.com/mtslive) |
| - **Community:** [discord.gg/mtslive](https://discord.gg/mtslive) |
| - **Trial coverage wiki:** [trial.mts.now](https://trial.mts.now) |
|
|
| If you build something interesting on top of one of these datasets — a model, a tool, a piece of journalism — tell us. We'll boost it on the show. |
|
|
