How this thermometer works
Who heated up AI this month? That is the question the thermometer answers: it reads AI news and turns media signals into one temperature per player, from -100 to +100, recomputed daily, with clickable evidence.
What it is: a media-temperature index calculated by public, versioned rules over news collected every day. The number comes from deterministic code: each accepted story can add or subtract points in specific dimensions. What it is not: a scientific ranking, vendor-quality measurement, investment recommendation, or editorial opinion disguised as a scoreboard.
What enters the score
Collection starts with a basket of monitored sources. The pipeline removes duplicates, applies the source policy, matches text against the player alias registry, and only accepts AI stories with a signal attributable to a company. Loose mentions, unsubstantial republications, and items outside policy stay out of the board.
The five dimensions
| Dimension | Core question |
|---|---|
| Market traction | Did it gain adoption, distribution, customers, or usage? |
| Product / technology | Was there real progress in model, product, infrastructure, or cost? |
| Capital / execution | Did it become more able to execute through funding, capex, talent, or organization? |
| Trust / risk | Did trust, safety, regulation, or reputation improve or worsen? |
| Competitive narrative | Did it leave the period looking stronger or weaker? |
Each scored story carries a value by player and dimension: polarity × severity. Severity runs from 1 to 3; a strong positive story can be +3, and a strong adverse story can be -3. The same story can affect more than one dimension when the fact supports it.
News sources
News is extracted from RSS feeds from these sites:
| 9to5Mac | ABC News | Ars Technica |
| BBC News | Bloomberg | CNBC |
| CNET | Engadget | Fortune |
| Fox News | Gizmodo | MIT Technology Review |
| NPR | SCMP | TechCrunch |
| The Decoder | The Guardian | The New York Times |
| The Verge | VentureBeat | Wired |
| ZDNET |