Two years of AI progress. Still struggling to get the joke. On 14,000+ real @playbadcards hands, Astra tops our updated LOL-Alignment benchmark matching the human judge 50% of the time Just 2% ahead of GPT-4o after 2 years! Humor is subjective. Games make it measurable. Funnier than AI? Our team got ~60% on the examples. Try them 👇 https://goodstartlabs.com/benchmarks/lol…