etn.

GM. Friday. @a16z x @etnshow. Here's the line up:

- @illscience (@a16z)
- @jamdac (@a16z)
- @astrange (@a16z)
- @jamesyflo (Ethos)
- @wlovedaypowell (809)

Join us live.

JUST IN: NVIDIA have themselves opened an investigation into Jensen Huang, after he admitted to the board that he had been intimate with Chinese spy, per NYT.

Andreessen Horowitz General Partner Angela Strange (@astrange) says for founders in legacy industries, finding a client who's afraid of falling behind on AI adoption matters more than domain expertise:

"The best founders generally have some experience in the space, although, to be honest, that's changing. If you're an amazing technical mind and you can attract amazing technical talent, I'm now seeing that you can hire the financial services expertise or you can hire the insurance expertise."

"What you do need to be able to do is find generally a CEO of a company that has realized that they are going to be behind. It's not AI that's your competition; it's your competition adopting AI faster than you."

"[These CEOs are] willing to take a bet to be the first customer of this new AI-native platform. And that's very much happening."

"Example: PennyMac Mortgage Company, right? They were the first customer of Vesta at scale, and they now originate mortgages 25% to 40% cheaper than all of their competitors. Looks pretty smart, right?"

Andreessen Horowitz Partner James da Costa (@jamdac) says every company needs a high-agency, high-output "weapon":

"In the South [of England], a weapon is like, you're on the football pitch, this person is an absolute weapon... Because he gets the ball, he drives up the pitch, he makes tackles, he scores goals... he's extremely dangerous on the field and he turns the course of games."

"In companies, for me, these are people... who have unimaginable throughput. So they're sort of autonomous missiles. They set the target themselves and then they go and execute on those things, and you don't have to worry about the thing getting done. It's already done."

"They create preferential attachment. Because they are setting the bar for what kind of high throughput in the company means, it means that other people raise themselves to that bar in the company."

"Other good people want to work with good people. And so then you start to get this network of really, really good talent that works within the company."

General Partner at Andreessen Horowitz Anish Acharya (@illscience) says using AI to edit the weights of other models is a bigger deal than people realise and could unlock continuous learning:

"There's actually two huge breakthroughs that have happened in the last few months that I think are under discussed... [One is] the paper that was released on Monday about having models edit the weights of other models, which gives you the sort of continuous learning opportunity."

"It was released and there was enthusiasm and then it was criticised, but I think it's a bigger deal than people appreciate."

"The smartest people I've been talking to have been able to not just replicate, but extend it."

"Now look, there may be some constraints, like there may be opposing ideas and models can't hold both of them in their head. You can't be highly conscientious and highly open, right? Because that sort of implies that the same weights are pointed in different directions."

"But you may be able to do things like get models that are able to have high entropy in non-verifiable domains, like beautiful art and prose, but then high precision in verifiable domains where they can also write great code and be mathematical."

"And we've never been able to reconcile those trade-offs before."