PK Paul Kedrosky Co-founder and managing partner at SK Ventures, an early-stage venture firm, and a contributing editor at Bloomberg covering early-stage technology and startups...

“Software is really unusual in another way, which is that the feedback loop, as you let it work in its own undirected way, … you can let it loose and do that because there's ground truth. The software works or it doesn't. Now, try doing that with most of white-collar work. I defy you to give a PowerPoint presentation to a model, let it loose for 12 hours, and say, I bet what it comes back with is better. It won't be, not because models suck, but because there's no ground truth. … If you let models loop recursively on things that don't have ground truth, you get exploding errors. You don't get better product.”

Odd Lots · Money & Markets · November 2026

“Software is really unusual in another way, which is that the feedback loop, as you let it work in its own undirected way, … you can let it loose and do that because there's ground truth. The software works or it doesn't. Now, try doing that with most of white-collar work. I defy you to give a PowerPoint presentation to a model, let it loose for 12 hours, and say, I bet what it comes back with is better. It won't be, not because models suck, but because there's no ground truth. … If you let models loop recursively on things that don't have ground truth, you get exploding errors. You don't get better product.” — Paul Kedrosky, Odd Lots

Context

Paul Kedrosky argues that AI's early adopters, coders, look nothing like the mainstream customers the revenue forecasts depend on. This is his reason coding gains may not carry over to other white-collar work.

Transcript

Odd Lots Around 43:50 into the episode
Speaker 2

Normals look nothing like that, right? We're not at the frontier. Things are good enough. I talk to coders who are baffled that people use Meta's Muse. They're like, don't you realize that's a horrible model? People don't care. It got them a restaurant reservation. So this is the. It's fine.

Joe Weisenthal

…emails that tell you you're going to be there at supposed to be somewhere at 3 p.m. tomorrow. You might want to. That's a totally fine. Use case. I do not need Fable 5.2 to tell me I have two emails that are like to tell me that I have two reservations conflicting with each other. Exactly. And so,

Speaker 2

and it gets worse, though, because software is really unusual in another way, which is that it has what's called a tight gradient descent, which means that the feedback loop, as you let it work in its own undirected way, at whether it's RSI, recursive self-improvement, or it's something more mundane, like just improving some old financial code, you can let it loose and do that because there's ground truth, right? The software works or it doesn't. And so it feeds back to itself and it can improve. Now, try doing that with most of white-collar work. I defy you to give a PowerPoint presentation to a model, let it loose for 12 hours, and say, I bet what it comes back with is better. It won't be not because models suck, but because there's no ground truth. Who knows what better means in that context? So if you let models loop recursively on things that don't have ground truth, you get exploding errors. You don't get better product. And this is a deep problem as you transition from coders to normal users, which is exactly what we're facing right now.

Tracy Alloway

Well, this is something that I've been thinking about even in the context of coding, because people were very excited about AI and its ability to code early on. But now, with all the outages that we've seen this week, I mean, like, you know, is that productivity if all you're doing is like creating…

Speaker 2

in the numbers, yes, and the numbers are staggering, right? I don't know if you've looked at it, but like the CVs in terms of critical vulnerabilities, in the first three weeks of October, where you identified, I'm going to go think of the numbers, it's like 8% of the critical software…

Tracy Alloway

yeah, that's right. Yeah. Right,

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

More from Odd Lots