Blogger proposes dating-app first-date rate as a personal assistant benchmark
Jack Raines argues SWE benchmarks are redundant now that personal assistant AIs are proliferating, and plans to give Duet, Dot, GrokBot and Instinct his Hinge and Raya accounts for 48 hours to compare first-date conversion.
NEW PROPRIETARY AI EVAL FROM JACK RAINES LABS ™️ We now have too many “personal assistant” AIs. Duet, Dot, GrokBot, Instinct, etc. All of the token munchers on this site are doing SWE related benchmarks but AI already codes better than anyone so it’s kind of redundant. It’s well established that AI has tiz (autism), but does it have rizz (charisma)? I will give each agent access to my Hinge and Raya accounts for 48 hours on my old jailbroken iPhone and see which has the highest first date conversion rate in a set time period. This is a much more important “real world” benchmark than the slop charts we’re seeing online these days. Results will be published on my blog.
This post quotes another post: x.com/claudeai/status/2112984664256241837