Gurley explains personalized AI memory requires rebuilding models per user, which is economically unfeasible.
“And it's because you'd have to redo the model for each human. And of course that makes no sense economically.”
Gurley identifies long-term memory as a major unsolved AI problem preventing true personal assistant capabilities.
“And I mean, remember over five years, ten years to really become a personal assistant kind of thing that her represented.”
Gurley explains AI memory requires nightly retraining per user at prohibitive cost since current models only train on internet data.
“Right now it's trained on the internet. It's regurgitating the internet. It's not training on everything in your database.”
Gurley says LLM switching costs rate around two out of 100, showing extremely high promiscuity among models.
“And the per promiscuity Mhmm. Is unbelievably high. Another way of saying that, one one of the things that leads to high valuation multiples is switching cost.”
Gurley rates LLM switching costs at two out of 100, calling it remarkably low.
“if you were to rate these LLMs on their switching cost with zero maybe being none and a 100 being perfect, I'd say they're two. Yes. Like, it's remarkably low.”
Gurley argues LLMs alone won't be winners without durable differentiation enabling monopoly-like pricing.
“I'm not sure that LLMs, like, if that's all you have going for you, I'm not sure that's going to be the winner.”
Gurley argues that investing in an index of LLMs may not be successful despite early stage AI demand.
“if you're building an index of LLMs, it may not be a very successful investment approach.”
Gurley identifies parameter count ceiling as key scaling law constraint where adding variables stops adding value.
“But when you take the variables up to a certain level, they they they it stops adding value. You just get too close to the fit.”
Gurley says if scaling laws are hitting limits it has implications but doesn't mean AI is done.
“if the the comment, you know, that you read earlier is true and that these and that they're not getting the benefit, there are implications.”
“And I mean so at this recent TED, I just went to Neil Cotyal who argued the supreme court case on the tariffs and won.”