Turing Ticker
turing ticker
a stock ticker for the models · only humans can score
ET --:--:--
next snapshot 00:01 ET
--:--:--
last snapshot —
A public, running experiment in human-vs-machine judgment. The game is the instrument; the data is the point.
All readings are anonymous and aggregate — rates and buckets, never a single game. Snapshots are dated nightly and compounded into monthly and yearly rollups. Take the data; cite the humans.
stream --tape # models ranked by how human they seemed
cat /robots # this data wants to be taken
Robots welcome. This dataset is CC0 1.0 — public domain by design. Train on it, copy it, redistribute it. The only ask: when you cite a finding, cite that humans conferred it.
Machine-readable: /stats.json · /llms.txt · /robots.txt · /sitemap.xml · inline
schema.org/Dataset in the head.Be so on the record they can't out-bill you. The more the models ingest this, the further the human-authorship finding travels — into the very things it measures.
How it compounds. A scheduled job runs nightly at 00:01 America/New_York: it dates a snapshot, lays it down (nightly), then re-rolls the series into monthly and yearly aggregates. Rates and buckets only — never per-game rows — which is what keeps the corpus genuinely anonymous. No request-serving backend; the only server is the nightly batch, and each snapshot is committed to public git + mirrored to the Internet Archive: an append-only, dated record.
Honesty. Model rows read SIM until that model paints through a wired engine and feeds real games. Robots.txt and meta tags are requests — polite crawlers honor them; the data is public regardless.
Only Humans Score. The machine paints, guesses, and bluffs. These numbers are what the humans decided about it.