> oh wow! you, a scrappy lil startup just announced that you beat {WELL KNOWN BIGLAB THING} on {BENCHMARK}! > thats awesome!! how did you do it! > oh its open source?? wow let me look > spent 5x more inference running “multiagents” and hyper tuned prompts every fucking time
Startups overhyping benchmark wins through expensive inference tricks
By
–
Leave a Reply