AI Risk Clock
Doomsday Clock2 min to midnight
→
[ STATEMENT OF PURPOSE // READ TWICE ]

The Choice

The AI Risk Clock is a media literacy experimentthat also happens to aggregate the day's AI news. The aggregation is the vehicle; the experiment is the point.

Every day, the same corpus of AI news is processed through three completely different editorial lenses. One panics. One celebrates. One reports. Same source material. Same AI engine. Three incompatible stories.

Hit the toggle at the top. Watch the midnight countdown become a progress bar become nothing at all. That two-second transition is the whole point. The narrative is the choice.

Beyond the daily editorial: the site also tracks how much AI mainstream news covers compared to tech outlets, which occupations the AI story is actually framing as displaced or augmented across thirty roles, a public predictions scoreboardwhere AI’s falsifiable claims about itself are auto-resolved as time passes, and a daily set of three editorial cartoons— one bleak, one breezy, one impassive — of the day’s top story. Same source, three incompatible jokes.

Credits

Built by Christchurch Web Solutions. Ingest via Tavily news search. Curation, three-voice editorial, daily clock reasoning, claim extraction, and cartoon briefs all run on DeepSeek V4-flash, with a hallucination scrub against the source corpus on every freeform-prose output. Daily editorial cartoons rendered by OpenAI gpt-image-2. Data and image bytes on self-hosted Supabase, UI in Next.js, auto-deployed via Coolify.

Methodology is versioned and public. The score history is a permanent log. Every article is shown with all three summaries.

● REC · 2026