Claude Fable 5.2 tested: one prompt turns 5GB of federal data into a US water-safety map
alliekmiller · x · 2026-11-10
alliekmiller reports that Anthropic released Claude Fable 5.2 and tested it with a single research question: which big US cities are best positioned for water in the coming decades.
- From one prompt, the model scoured 5GB+ of federal data across EPA, USGS, NOAA, FEMA and the Census Bureau, including 15.5M EPA violation records and 2M PFAS test results
- A dozen other AI agents cross-checked its work and caught errors, such as Alaska's North Slope scoring high due to missing temperature data
- The final map rates 3,100 counties and 900+ city areas on 17 measures: Indianapolis ranks #1 on the aggregate of supply, climate outlook, quality, testing/oversight and backup, with Cleveland 0.1 point behind, then Cincinnati, St. Louis and Nashville
- The model also produced a trailer video. Her takeaway: a research question in your head can now yield this level of insight within hours
More from Fun
- Claude's model picker screenshot: Opus 5.5, Sonnet 5.5, Fable 5.2, Haiku 5.5 — BLCNYY · 2026-11-10
- The joke that captures analyst bias: curing cancer is 'bearish for AI search' — gaganghotra_ · 2026-11-10
- Stanford shows Grok fails face identification, misidentifying Prof. Tatsunori Hashimoto — stanfordnlp · 2026-11-10
- EA critics argue the rationalist community's overconfidence in its own epistemics makes it easy prey — fkasummer · 2026-11-10
- Handing your whole inbox to OpenAI: one quip captures the privacy unease — justalexoki · 2026-11-10
- Friend told of ban 4 hours early, replied 'I'll have Claude look into it'—got banned anyway — NathanpmYoung · 2026-11-10