Yesterday, I was watching a reshowing of Gwynne Shotwell and Elon Musk at the 2026 All-In Summit. Perks of the attention economy are that almost all content these days no longer have exclusivity. And if so, there’s only a small window of that arbitrage. Nevertheless, I digress.
In it, Elon says, “Physics is a harsh judge.” And well it got me thinking. We live in a world where DPI (distributions to paid-in capital) is a harsh judge.
And I say DPI in a royal sense. It really is realized IRR. ‘Cause 3X DPI in 15 years is quite bad. Your Benjamins are better in the money market growing 8% a year. Not that much better but better. 3.17X to 3X. But benefit of the money market is your capital is liquid. But I will use DPI for the purpose of this piece because it rolls off the tongue better. Or rolls off the fingers…? You get my point.
But unlike the world where physics is a harsh judge, DPI is not instantaneous.
I was also catching up with a friend yesterday and he said something with respect to recursive self-improvement (RSI). For the layman, aka me as well, the process in which the AI learns by itself from itself. In other words, an AI capable of rewriting their own code. It’s a great theory and we see quite a few companies working on turning this into a reality. But the most interesting takeaway I got from that conversation was that “recursive self-improvement only works if it doesn’t take up time. If it takes time, you can’t do that many turns. Then the harness becomes the only thing you iterate on, but that undermines any moats.” In fact, building a Claude skill is the equivalent of retraining the harness after each task. Requires little time spent, but anyone can do so.
Similarly, because DPI is realized after an incredibly long feedback cycle, 5-8 years at best, you don’t have much time to retrain the model more than a small handful of times before, well, your life expectancy. So, in lieu of DPI, LPs look for other signals that act as a proxy to the harsh judge.
Mark-ups become the mild judge.
Revenue growth and talent density become the gentle judge.
And brand is the lenient judge.
And when things are too lenient, you no longer have one judge, but an army of judges. Anyone can judge you. Everyone is your jury. And all you need is one viral tweet to be your executioner.
So, when LPs say they underwrite you on discipline, at least this is what I look for, and my generous interpretation of what other LPs look for, they’re underwriting you to what can you do today that exists on a faster feedback cycle and what is easy to make evident your ability to compound your existing assets. In other words, between Fund I and II, what is the metric you would like to give LPs for them to assess your rate of change?
Now it’s up to LPs to decide if that metric holds gravity or not. But you at least have to provide one. In its absence, we are back at square one. DPI is a harsh judge.
Photo by Sasun Bughdaryan on Unsplash
Stay up to date with the weekly cup of cognitive adventures inside venture capital and startups, as well as cataloging the history of tomorrow through the bookmarks of yesterday!
The views expressed on this blogpost are for informational purposes only. None of the views expressed herein constitute legal, investment, business, or tax advice. Any allusions or references to funds or companies are for illustrative purposes only, and should not be relied upon as investment recommendations. Consult a professional investment advisor prior to making any investment decisions.