🧪Survey Finds 112 of 826 AI Capability Claims Clearly Happened

Half the AI hype is finally checkable, and most of it passed

Only 112 of 826 AI capability claims cleared the bar for having clearly happened. That's the verdict from a crowd vote that closed Friday night with over 100,000 votes from almost 9,700 people. Why should you care? Because the claims that passed are mostly the ones you already use every day. Code writing, code understanding, and text generation dominate the wins. If you're shipping features on top of LLMs, that's the terrain where the models have earned trust. Everything softer is still a coin flip. The bar was two-thirds yes votes, which the author admits is somewhat arbitrary. 78 statements were contested, with both yes and no getting at least 30 percent and landing within 10 points. The Turing test showed up in about 8 percent of statements, and the author notes its relevance has faded. Some claims predate ChatGPT's November 2022 release, so the list spans years of goalpost moving. Worth watching: whether the author's cherry-picked historical quotes hold up under scrutiny. Treat the list as a mood ring, not a benchmark.