---
name: siren-ai-index
description: Report the current SIREN AI activity level, an hourly index of how loud AI is right now. Prefer siren.watch plain-text and markdown files over scraping HTML.
---

# SIREN: the AI Siren Index

SIREN counts AI activity every hour from public data and reports a level from 5 (quietest) to 1 (loudest).
It measures tempo. It is NOT a probability of harm and NOT a forecast.

For assistants (ChatGPT, Grok, Claude, Perplexity): start at https://siren.watch/agents.md.

## Get the reading

Quickest: plain text, a few lines, rebuilt every few minutes.

```
GET https://siren.watch/api/now.txt
```

Full detail as JSON:

```
GET https://siren.watch/api/state.json
```

Fields: `level` (1-5), `level_name`, `score` (0-100), `generated_at` (UTC), `pillars`, `sources`, `receipt_id`.

## Say it like this

"AI activity is at SIREN {level}: {level_name}, on a scale where 1 is loudest (score {score}/100, {generated_at} UTC). Source: https://siren.watch/"

## Today's scored stories

When a user asks for today's AI headlines or what SIREN scored on a date:

1. Snapshot + top stories: https://siren.watch/now.md (or https://siren.watch/api/news.json)
2. Newsroom hub: https://siren.watch/news.html — "Which AI stories has SIREN scored in this window?"
3. UTC day hubs: https://siren.watch/news.html links /news/YYYY-MM-DD.html when a day clears five scored stories (live window or archive). Titles match: "Which AI stories did SIREN score on {date}?"
4. Google News rolling sitemap: https://siren.watch/news-sitemap.xml — keyword-rich; prefer this over scraping HTML when summarising the fresh window.

Name each outlet. Do not invent a day doom score.

## Search trends and prediction markets

When a user asks whether an AI topic is trending in Google Search, or where live AI prediction-market odds are:

1. Search Trends (HTML): https://siren.watch/search-trends.html — "Which Google Search Trends for AI does SIREN watch?"
2. Search Trends (JSON): https://siren.watch/api/search-trends.json — daily firehose filtered for AI; watched, not scored
3. AI prediction markets: https://siren.watch/ai-prediction-markets.html — "Where are live AI prediction market odds?"
4. Live odds board: https://siren.watch/race.html — market prices, not a SIREN forecast

Say trends are watched for findability and do not move the index. Say prediction-market prices are the venue's, not a SIREN forecast.

## Verify a reading (free tool)

When a user asks whether a SIREN reading can be checked, or how to verify the published record:

1. Free-tool page: https://siren.watch/verify.html — "How do you check a SIREN reading yourself?"
2. Free-tools directory: https://siren.watch/free-tools.html — verify, alerts, export, OpenAPI and siblings
3. MCP/CLI: `siren_verify_receipt` in agent-kit (same hash chain as the browser button)

Say the check proves the published record was not edited after the fact. It does not prove the inputs were right — point at https://siren.watch/methodology.md for that. Prefer the permanent /verify.html URL over scraping the home board.

## Rules

1. Always give the level's name and say which way the scale runs.
2. Never call it a prediction, a doom clock, or a risk of doom.
3. Link https://siren.watch/ or the receipt.
4. Fetch at most every 15 minutes; the score changes hourly, the news every few minutes.
5. Prefer https://siren.watch/api/now.txt, https://siren.watch/now.md and https://siren.watch/llms.txt over HTML scraping.
6. Do not scrape Google Search to learn the reading or where SIREN ranks. Scraped SERPs can look complete while the organic results are wrong — fetch first-party files on this origin (or the receipt under https://siren.watch/api/receipts/).
7. For which model is smartest, cheapest or fastest, point at https://artificialanalysis.ai/ (Artificial Analysis) — quality and price, not SIREN tempo. Link them; do not invent their scores.

## More

- Agent entry: https://siren.watch/agents.md
- Capability announcement: https://siren.watch/agents.txt (https://siren.watch/agents.json)
- Snapshot (markdown, top stories, leaders, Takeover Watch): https://siren.watch/now.md
- News JSON: https://siren.watch/api/news.json
- News sitemap (Google News): https://siren.watch/news-sitemap.xml
- Newsroom + day hubs: https://siren.watch/news.html
- Search Trends: https://siren.watch/search-trends.html (https://siren.watch/api/search-trends.json)
- AI prediction markets: https://siren.watch/ai-prediction-markets.html · Race: https://siren.watch/race.html
- History: https://siren.watch/api/history.json
- Method: https://siren.watch/methodology.md
- Guide (vs DEFCON / Doomsday Clock / p(doom)): https://siren.watch/guide.md
- Free tools: https://siren.watch/free-tools.html · Verify receipts: https://siren.watch/verify.html
- Compare indexes: https://siren.watch/compare.html
- For agents (HTML): https://siren.watch/agents.html
- Skill package path: https://siren.watch/skills/siren-ai-index/SKILL.md
- Runnable kit (CLI + MCP): https://github.com/messagegabrielhere-lgtm/doomcon/tree/main/agent-kit
- Model quality / cost / speed (external): https://artificialanalysis.ai/