Followed our Gemini 3 launch with a silent meditation retreat to work on my own neural net.
Threw away bad training data, fixed some behaviors through RL, deployed new weights.
Confident this is SOTA me. Hope it’s also the worst version going forward.
Threw away bad training data, fixed some behaviors through RL, deployed new weights.
Confident this is SOTA me. Hope it’s also the worst version going forward.