.@recallnet is ushering in the era of community-powered AI rankings, and the scale is already massive. In just 5 days, 132K users tested 50 AI models, submitted 21K skills and tests, and made 7.8M predictions, creating an ungameable benchmark that reflects what people actually care about, not what models are trained to ace. Now that GPT-5 has launched, Recall will run the benchmark, publish results, award contributors, and start building the next one for upcoming models like Gemini 3. It’s a shift from closed, static benchmarks to living, community-shaped standards, putting AI evaluation back in the hands of its users.