LLMs

“It's able to do things. That are not explainable by a simple probabilistic model of auto-suggesting next character, next word, next token, next sentence, right? Like it's gone well beyond that.”

“I think that these single purpose robots is going to be The big winners in the next cycle here where like you have the knowledge work, like LLMs, chat GPT type of stuff, but then I think you're going to see these specific robots. I think both of those are goin…”

“The only thing that's probably done more Then LLMs for vibe coding is the actual meme itself of vibe coding.”

“The key here is when you're doing the initial prompt, don't forget to say, ask me any questions before you get started. So we have the right strategy for this. I've noticed that by putting that in there, that small one, you know, one sentence, you're going to …”

“It might be the biggest hack of using LMs, like, well, is ask for the prompt. You will never be able to out prompt the person, the guy, you know, the software that sees all the prompts.”

“I think if you can build proprietary data sets, which is harder than it sounds or you can build really deep interfaces with your customers, which is also harder than it sounds. Those are moats you can build, but really, I think sometimes what you're really, yo…”

“And LLMs are very good at helping you do DOK one, two, and three. And reasoning around that. And then the part that the humans still do really well is DOK four creating new knowledge.”