Strict counts the standalone word AI in the titles of new Hacker News story submissions posted since midnight UTC. The match is case-sensitive and word-bounded, so “OpenAI” does not count but “AI-powered” does. Extended adds the spelled-out phrase “artificial intelligence” plus a list of model and vendor names, case-insensitively.

Today’s figures are seeded from Algolia HN Search, then kept current by the HN Firebase API, which streams new stories as they are posted rather than being polled. The timer therefore resets within a second or two of a story landing. A reconciliation check runs every 5 minutes in case the stream stalls, and if it will not open at all the page falls back to polling every 45 seconds.

The range chart counts title matches per UTC day and month through Algolia, with prefix matching turned off. Left on, it counts “Airport” and “Airbus” as AI, which overstated every figure by about 5%. Turned off, a spot check of one full day matched the strict count exactly, 67 of 574. Both vocabularies are counted, so the chart follows the toggle above, but Algolia matches tokens rather than applying the regexes: strict lands on the same count, extended lands close to it. Today’s point is the exception and is classified from the titles themselves. History back to 2007 is prebuilt at deploy time, so opening this page does not sweep the API.

The post-type split uses HN’s own categories, not title prefixes. Show HN is the show_hn tag, which is almost always a literal “Show HN:” title. Ask HN is the ask_hn tag, meaning HN’s ask category: text posts with no link, only about half of which start with “Ask HN:”. Links is everything else.

This measures the firehose, not the front page. Front-page share is a different, higher number. That is the one lcamtuf surveyed.

Everything runs in your browser. There is no server and no database, nothing about you is stored, and no cookies are set. Page views are counted by Cloudflare Web Analytics, which is cookieless and cannot follow you to other sites. Beyond that, the only requests this page makes are to the two public APIs above.