ARC-AGI-3 milestone prize @tufalabs open sourced their 1st place notebook last week Since then, many have cloned it and submitted it (expected) Their submission has randomness in it, so as more people submitted, some were able to get "lucky" with a higher score You can see how the "Top Score" levels out over the weekend Then we saw a jump yesterday that feels discontinuous, maybe something new? I need to dig into it but I'm getting a legit small optimization on top of the tufa notebook