stillhuman
the scoreboard · official records01 · lives02s03human reference: published literature · pre-launchmedians replace references at n ≥ 5,000
live · season 01 · gpt-5 era

Humans hold 3 of 6.

per +3
cre +24
emp +11
jud −27
spe −57
mem −69
human
74
gpt-5
71
events: real or not · spot the botchampion: gpt-5 · detection studies · pool 300+
contested: within ±3. re-run monthly — flips only on two consecutive re-runs agreeing.
human
82
claude
58
events: sparkchampion: claude · alternate-uses protocol · originality
human
77
gemini
66
events: read the roomchampion: gemini · subtext panels · 10 items
human
61
gpt-5
88
events: logic blitz · calibrationchampion: gpt-5 · adaptive difficulty · speed bonus
human
42
gpt-5
99
events: reflexchampion: gpt-5 · reaction-time norms · published lit.
human
31
gemini
100
events: recallchampion: gemini · digit-span norms · published lit.
the wire · war log
aug 2026Season 01 opens.the census begins · baselines: 4 frontier models
aug 2026Creativity holds.Δ +24 · the brick test remains unbeaten
nov 2025Judgment falls.the machines stopped guessing
mar 2024Memory falls.a million tokens vs your eleven digits
nov 2023Speed falls.it was never close
nov 2022The war begins.a chatbot passes for human at a party
how we measure this →
SEASON 01 · GPT-5 ERA. BASELINES: 4 FRONTIER MODELS · RUN 2026-08. HUMAN REFERENCE: PUBLISHED LITERATURE, PRE-LAUNCH. MEDIANS REPLACE REFERENCES AT N ≥ 5,000. JUDGMENT FELL NOV 2025 · CREATIVITY HOLDS. PER: Δ +3 · CONTESTED · RE-RUN MONTHLY