Andrej Karpathy

And then, you know, I think on the post-training side, you sort of have all these different methods for, you know, training models from supervised approaches like SFT to, you know, this preference tuning RLHF to RLVR, which we'll talk a bit more about.

There's a really good article actually in the readings, if you guys read Karpathy's write-up on RLVR in the readings, that talks about what happened in 2025.

You know, RLVR really came to prominence in 2025.

We'll get into that in a second.

Research becomes companies, companies become fortunes, fortunes become gifts, and gifts fund more research.

And the next harvest is already ripening with alumni like Jensen Huang, Andrej Karpathy, and Sam Altman.

Yet all this groundbreaking research and infrastructure just maintains the brand.

The defining property of Harvard, Stanford, and Princeton is that their reputation is not attributed to any one field.

Additional investors include KOTU Operator Collective and Hustle Fund.

Angels include Okta co-founder Frederick Karest, Discovery Loop co-founder and former Google chief scientist Jeff Dean, and OpenAI co-founder Andrej Karpathy, among others.

Mintive is joining a crowded field of products that promise to govern, provision, and offer identity management to enterprise AI agents.

Its platform includes an MCP gateway that allows IT professionals to authorize what the cloud services agents may access and what actions they can perform on them.

bringing its total funding to $11 million.

Koachi, Operator Collective, and Hustle Fund also invested, and the angel list includes Google's chief scientist, Jeff Dean, the AI researcher, Andrej Karpathy, and Okta co-founder, Fredrik Kerst.

Minteev's software gives companies a way to provision, govern, and track the AI agents now acting inside their systems, the way identity tools manage human employees.

As businesses hand agents access to email code and payments, the question of who or what is allowed to do what has become one of the hottest problems in enterprise security.

What 2026 actually looks like? The vibes are loud right now, so let's start with numbers.

Speaking of vibes, Andrej Karpathy coined, vibe coding, in February 2025, for the kind of coding where you, fully give in to the vibes.

Stack Overflow asked about it that year.

72% of developers said it isn't part of their professional work, and another 5% said emphatically no.

As one of the key authors of O1, who I talked about a lot last year, Lucas Kaiser said, when you know the right chain of thought, you can compute anything.

And as Andrej Karpathy said two days ago, those thoughts don't even need to be legible to us.

They might piece together reasoning steps that are translated from other languages or are in their own made-up language.

Remember that the objective is now clearly to get the right answer, so the models will optimize however they can to do so.

So before we get to Nvidia's GTC, let's linger for a moment on AGI, its definition and recent updates to the timeline to AGI.

Yesterday, Andrej Karpathy, who was until recently at the very top of OpenAI, said this about the definition of AGI.

He thinks of it like the OpenAI charter as being an autonomous system that surpasses humans in most economically valuable tasks.

And is it me, or does that definition not automatically foreshadow economic strife? In other words, definitionally, AGI won't have arrived until it can do the work of at least half of all humans.

Now, again, that's the author's opinion, right? I'm not saying if that's right or that's wrong, but the point is, you know, the board at OpenAI nearly imploded and it was done by people that a lot of us, we didn't know who they were, what their motivations were necessarily.

They weren't touring the world, talking to all the world leaders like Sam Altman was doing, right? So Sam Altman, Ilya Satskovor, Andre Karpathy, Mira Murady, Greg Rockman, like all those people, they're in interviews, they're talking to the people, they're explaining their viewpoints.

So when they're running OpenAI, I feel like a lot of people are, I mean, it's not like we just trust everybody 100%, but at least we're like, we know them, we know their opinions.

We've watched, you know, three hour long interviews with them.