Uhh could we turn codebases into neural networks and then do verifiable interpretability on them?
Handwired bridges the gap between traditional code and a neural net in a funny way. It turns if-else and loops into weights, like so:
— Cheng Lou (@_chenglou) October 15, 2026
Just spitballing here, I haven’t deep-dived either of these yet but am I seeing something or am I goofin
Oh the above diagram from https://already-clean.goodfire.com/ if you haven’t seen yet it’s got some great viz to go with the findings