# Signal — AI news radar

> An editorial AI news radar covering models, GitHub, research, publications, labs, cloud, Reddit and Hacker News. Each story has an explainable 0–100 editorial score. Updated 2026-10-18T04:27:44Z; scoring version editorial-v3.

The page includes static HTML and JavaScript enhancements. The views below use the same published dataset and ordering as the dashboard.

## Data

- [Markdown digest](https://signal.ai-shiz.com/signals.md): overall ranking and top five per category. The API defaults to 48 hours; use `sort=trending` for the dashboard default, ordered by momentum. Parameters: `sort` = trending | day | week | month | year | now; `category` = all | models | repos | papers | media | enterprise | cloud; `limit` = 1..100.
- [Signals JSON](https://signal.ai-shiz.com/api/signals): same parameters; includes id, title, url, source, UTC ISO 8601 dates, scores, components, reasons, metrics and summary.
- [RSS](https://signal.ai-shiz.com/api/rss?sort=day): `sort` = now | day | week | month | year.
- [Full dataset](https://signal.ai-shiz.com/api/news): archived stories and score breakdowns; can be several megabytes.

## Categories

models — 123, github — 182, research papers — 30, publications — 205, enterprise — 154, cloud — 40.

## Reading scores

- Relevance: 30; freshness: 25; evidence: 15; independent platform coverage: 15; primary source: 5; platform popularity: 10.
- Missing metrics remain `null`, not zero. Platform units are kept separate.
- Scores help prioritize news; they do not verify claims. Check the original `url`.
- Periods use the story date. Hype model dates represent updates, not original model creation.