# How does SIREN compare to DEFCON, the Doomsday Clock and p(doom)?

> HTML: https://siren.watch/guide.html

Five numbers people reach for when they want to know how worried to be. They measure different things, and only one of them is a count you can check. Right now it reads **SIREN 4, ROUTINE**.

## The short version

| Number | What it measures | Who sets it | Can you recompute it? |
|---|---|---|---|
| **SIREN** | How much is happening in AI: tempo, not danger | A published formula over public data, hourly | Yes, from the receipt |
| **DEFCON** | The readiness posture of US armed forces, 5 (lowest) to 1 (highest) | US military command | No. The current level is not routinely made public |
| **Doomsday Clock** | A symbolic judgement of how close humanity is to catastrophe | The Bulletin of the Atomic Scientists, usually once a year | No. It is a judgement, by design |
| **Pentagon Pizza Index** | How busy pizza shops near the Pentagon are, read as a hint of military activity | A website, from shop-busyness data | Not from anything it publishes that we could find; its own disclaimer says not to rely on it |
| **p(doom)** | One person's stated probability that AI ends very badly | Whoever is asked | No. It is an opinion expressed as a number |

## How to read the SIREN dial

The scale runs from 5 to 1 and **1 is the loudest**. The level is the band the 0–100 composite score falls in:

| Level | Name | Score band | In a few words |
|---|---|---|---|
| **5** | DORMANT | 0–34 | Needle at rest |
| **4** | ROUTINE | 35–54 | Needle breathing |
| **3** | ELEVATED | 55–69 | Off the rest stop |
| **2** | ACCELERATED | 70–84 | Pinned high |
| **1** | UNPRECEDENTED | 85–100 | Off the top of the chart |

The score is built from five pillars: **Capability** (what shipped, and what got written down), **Compute & Capital** (silicon, spend, and the market’s opinion of both), **Attention** (how loudly the world is talking), **Governance** (paperwork with the force of law), **Markets** (where money is posted against an opinion). Each is scored against the index's own frozen record, so a reading says how today compares with what the index has seen before.

## What a loud reading does and does not tell you

A loud reading tells you a lot shipped, a lot was spent, a lot was written or a lot was regulated, and the pillar breakdown says which. It does not tell you whether any of it was good or bad, and it is not a forecast of what happens next. That is why the home page is stamped "not a prediction".

## Why borrow DEFCON's grammar at all

Because nobody needs the scale explained twice. "Five is calm, one is not" is already in most people's heads. The borrowing stops at the grammar: SIREN is not a readiness level and is not issued by any government body.

## Where to go next

- [AI activity index today · SIREN](https://siren.watch/)
- [How the SIREN AI activity index is computed · methodology](https://siren.watch/methodology.html) · [methodology.md](https://siren.watch/methodology.md)
- [Which AI doom indexes can you compare?](https://siren.watch/compare.html) · [Which AI indexes does SIREN contrast?](https://siren.watch/vs.html) · [What are AI doom index alternatives?](https://siren.watch/alternatives.html)
- [SIREN vs PizzINT: which should you use?](https://siren.watch/vs/pizzint.html)
- [What is the history of AI risk?](https://siren.watch/history.html)
- [What are the AI explainers?](https://siren.watch/guides.html)