Aditya Agarwal

Mentioned in 1 episode

"If you look at something like Open Evidence — a phenomenal company — you think you go to Open Evidence and there's like one call to an LLM model, but in reality when you hit Open Evidence there's actually 50 different instantiations of calls to various models to run different parts of what they're serving. Right now 80 to 90% of the calls within that one call trace happen to open-source models."

Mentioned in 1 episode

"Goodfire is the leading AI interpretability lab. What that really means is that you can go and introspect what is happening in these models — it no longer has to be a black box. They just released Silicico, which was a huge release just yesterday. You can actually go in and figure out which weights are being activated, why did the model do something, why did it choose to pursue a particular line of thought."

Mentioned in 1 episode

"One of the companies that we incubated at SPC about 7 years ago was BET, which is one of the leading providers of these open-source models. Just seeing the amount of usage these models are getting across most of the enterprises that we are familiar with."