Into Latent Space: AI Financial Advances artwork

Into Latent Space: AI Financial Advances

Latent Space AI

May 4, 2026

In this episode, we venture into latent space to examine the financial advances of OpenAI and Anthropic. Learn about the interconnectedness of funding and innovation.
**SPEAKER_1** (0:00)
OpenAI and Anthropic, the two biggest AI companies in the entire world, they both signed a multi-billion dollar private equity deals, and they did it on the exact same day. The structure of how those deals happened is different, though, so I think that's going to be really interesting when it comes to what's going on with AI and enterprise. Also, Greg Brockman says that AI went from writing 20% to 80% of all OpenAI's code in one month. Saudi Arabia's Humane just launched on AWS, and it is calling it the first enterprise-grade agentic AI operating system. And Morgan Stanley just bumped the 2026 hyperscaler forecast to $805 billion.
The first thing I want to break down, though, is Greg Brockman's. He went on a podcast recently, and he had all these wild quotes, including talking about AI. He said, quote, even over the course of December, we went from these agentic coding tools writing 20% of your code to writing 80% of your code. This is a 4x jump in one single month. And he also said that OpenAI is 70% to 80% of the way to AGI. He also agrees with Sam and Demis over at Google DeepMind, and says that we're maybe two breakthroughs away. Now, of course, this wasn't everybody that agreed with him. Over on X, Yan LeCun, predictably, has been dunking on the AGI claim. He's been basically doing this all week. You also have Andrej Karpathy, who was a lot more, I guess you could say, like, measured. Basically, the thing that I've heard him talk a lot about is he says that, I'm sure some of these coding writing percentages are like huge, right? Going from 20 to 80 percent.
But what he said is, if you measure tokens emitted versus what is so, and that's kind of where you're getting this 20 to 80 percent jump. But he says the difference is, and a harder question is, what fraction of all of these tokens, all of this code that's being admitted are going to be merge commits that a senior engineer would actually approve? And I think a big point of that is, a lot of this is getting written by AI, but a lot of it is still going to have to be edited or not approved or redone. And personally, I'm going to be honest, I am not a developer, but I've created a bunch of software apps over the last couple months using these AI tools. It's been so fun. It's super addictive.
But basically everything I write, it's like I feel like I'm starting out with a giant block of granite and I'm chiseling it down to be what I want it to be.
And so it's not like, hey, Claude, make me a perfect statue, and it chisels everything.
You give it a direction and then you say, I change this, change that, and you're refining the code, you're rewriting it because it has bugs and glitches and errors. And so anyways, I think people can be pessimistic about that process. But I think at the end of the day, it's for people that aren't developers that don't know how to code. I mean, even for developers who are rejoking, you can just go way faster. So I know people, him and Ha about the exact percentage and well, you know, it technically can't, you know, write all of your code perfectly. Like you have to go and edit it or monitor it or fix the bugs or rework it. Like, OK, whatever. But that's life. And you're still going 100 times faster than if you're doing it from scratch. So in my opinion, it's exciting. The most recent project that I personally have been vibe coding doing all of this is called AI Chat Daily Dotcom. It's basically a news website that goes along with this podcast. So every time I drop a podcast episode, that website will go grab the transcripts from the episode. And all of the news stories I talk about, it will write an article about them, and it will put quotes from me in the podcast and all my opinions basically in there. It will also go and scrape 20 different news sources and get more facts and pieces of information and kind of embed those and has kind of bullet points in there. So if you hear me talk about a story here on the podcast and you're interested in getting a deeper dive into all of the facts in the background, while also getting quotes, my opinion on it, you can go check out the associated. And usually I'm going to start putting them in links in the description. It's not already there. This is all part of the this vibe coding project. So it should all be automated.

16 more minutes of transcript below

Feed this to your agent

Try it now — copy, paste, done:

curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000766120210

Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire.

Using your own key:

curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000766120210