Wake me up when AI can...

A Hacker News community project tallying roughly 800 “wake me up when AI can…” predictions posted between 2016 and 2026 drew more than 100,000 votes from about 9,700 participants after hitting the site’s front page on October 1, 2026. The informal audit asked readers to judge whether specific past benchmarks—ranging from chatbots passing the Turing test to AI writing human-indistinguishable articles—had actually been met. Results showed the bar shifts as capabilities arrive: plain Turing test claims drew a clear “yes,” but pass rates fell once qualifiers such as an actively probing tester or fooling any human over a long period were added. The exercise matters because it quantifies how “moving the goalposts” shapes public perception of AI progress at a time when frontier models and coding agents are already handling overnight troubleshooting tasks. What to watch next is whether similar crowd-judged benchmark archives become a recurring way to hold AI claims accountable as new models ship.