Which AI terms are actually hot, measured the boring way: deterministic mention counts across four independent sources this site already tracks (AI news, Hacker News, research papers, GitHub trending repos). A term spiking in one source is noise; a term hot across several at once is signal. Counts last updated 2026-08-24.
51 of 100: cross-source AI chatter is running hotter than last week. This week's top-20 terms average 309.4 weighted mentions vs 305.3 in the prior week (50 = unchanged).
raw = total mentions across sources in trailing 7 days; breadth = number of distinct sources with >=1 mention in those 7 days (1-4); heat = percentile rank of (raw * breadth) among all terms with raw>0, scaled 0-100, rounded.
HYPE INDEX: index = min(100, round(50 * thisWeek/priorWeek)) when prior>0 else null, where thisWeek and priorWeek are the means of the top-20 term raw counts in each 7-day window.
Counting: each source item (article, HN story, paper, repo) counts once per term per day, matched case-insensitively on whole words against a curated vocabulary of 220 terms. Every number on this page is a count of real stored items; nothing is estimated and no language model is involved. News and repo history starts the day tracking began; Hacker News and paper counts are backfilled from stored daily snapshots.