@_lewtun@GregHBurnham Yes, and if you'd like to chat more about it, please DM! My current bugbear is trying to figure out how I'd test "can AI generate good RL envs for itself" versus "good RL envs for weaker models".
@ben_j_todd In our defence, these were "what does simple extrapolation of trends suggest" rather than all-considered forecast.
It should probably be judged against FM 1-3 rather than FM 4. A bit low, but actually not bad at predicting Astra.