Yeah, just about jackcushman.org/apps/triad-q...

tfw you open a box you weren't expecting and it's an all-in pledge for a #boardgame you forgot about

It's an odd technology. I gave the same literal prompt and indeed got a very silly image (L), and then had it find the data first, and then make the image, and it (a) researched extra real data (the MVP winners) and it then made an infographic that has consistency issues but isn't bonkers

I'd argue that BerryBench ("How many Rs...") is a very silly thing to test on but still matters for considering LLM based systems (like GPT-5) as consumer products:

Not sure how to grade this one. This is why teachers ask students to show your work

I just solved Bellingcat's 'Toolkit Tracing' challenge 🏆 Try it yourself! challenge.bellingcat.com