[sharkjacobs]

The problem with Rabbit R1 and Humane is that we're still a massive gulf away from a competent human personal assistant who I can rely on to understand what I'm saying and use common sense and reasoning to respond and perform tasks reliably.

This feels like the same thing. If I ask my assistant about "music festivals in Boone North Carolina in august" and they give me 5 results, 0 of which actually match what I asked for I'm throwing my hands up and never asking them for help again.

I use LLMs all the time, regularly throughout every single day, at work and home, and sometimes I really struggle to articulate what they're good at, but this doesn't feel like it. LLMs are really good at talking, so they're easy for casual users to interact with, and we keep seeing products which try to use them to "take away the friction" of various tasks, but I don't think that's it. I think they're at their best when used deliberately and critically.

I think that GPTs were a good product, but not a popular one because most people haven't thought of a need which GPTs would fill. Search is a problem that everyone knows is a need and since "google sux" now its one that is ripe to be filled by a competitor. But I don't see how this is a real improvement, the problem with search is that the results are worse, not that they need to be summarized in a more friendly and accessible voice.

[Ancapistani]

> sometimes I really struggle to articulate what they're [LLMs] good at

I think of them like a person with entry level experience and an IQ of about 80.

That doesn't seem super useful, only... that "entry level experience" isn't in a single field. It's in literally everything.

It takes some time to figure out how to interact with them in a way that reliably gives the results you expect, but once you do, you can get a lot done. Taken to an extreme, a mid-career professional can essentially become a team lead that manages LLM-driven processes instead of a team of employees.