Great software always took shape in conversation, not the commit. With agents, the conversation that generates the code is becoming the true source of our software. And Git can't keep up.
So we built something that can. Meet DeltaDB: https://t.co/x0UHntKy9m
i know it's selection bias but several of my friends today are complaining about the Fable restrictions for using it to develop competitive capabilities. It's apparently a sharper restriction than I anticipated.
Training a model to generate RL tasks not too hard, not too easy costs many solver runs per task.
PROPEL predicts difficulty via a probe on its activations instead, amortizing cost and speeding up generator optimization.
New open-ended RL research from @Vmax + @GoodfireAI.
A recommender asks what you are, and shows you more of it.
It doesn’t care about what you are becoming.
Taste is a trajectory that is playing out. I wrote about what it would mean for software to track the shape of one as it moves, instead of pinning you to who you were.
A recommender asks what you are, and shows you more of it.
It doesn’t care about what you are becoming.
Taste is a trajectory that is playing out. I wrote about what it would mean for software to track the shape of one as it moves, instead of pinning you to who you were.