Entering AP US History in high school, my professor gave us some Summer reading before the start of the year. The book was “Lies My Teacher Told Me”, which detailed how US History classes washed over the truth of the American story. The most vivid being of Christopher Columbus.
It felt really incredible to think about a US History professor demystifying how to get to the “what really happened”, and how we really got here. Which then opened the class throughout the year to more rigorous discussion and freedom of thought. One classic debate was around the US dropping the atomic bomb in Japan twice. Was it to actually stop the war, or to show Russia who really ran the global stage. Doesn’t really matter what you think, it was more that we could debate this for an entire class period as the point of school. And it gave a structure to question authority constantly.
Fast forward to today. I was at a house party recently with other tech people and founders, and we were discussing some of the different micro-eras of the past 10 years. Specifically, the popularization of bad advice and what led to it.
THE BAD ADVICE OF PATTERN MATCHING INCENTIVES
Show me your last big win, and I’ll tell you about your next 5 failures.
The point of this is short. The new thing has to look like the new thing, which makes it hard to pattern match to the previous new thing. But pattern matching is at the heart of derisking. This creates a weird loop of advice.
The previous advice of zero interest rates and Snapchat as the last big thing, for consumer software, was to grow at all cost / monetize later / find novel behavior.
Nikita Bier showed all of this was wrong with Gas App. Monetize directly in the V1. Derisk through known behavior. Growth inputs must be advantageously proportional to the monetization in the V1.
There wouldn’t be a single VC that could share this advice. As a result, we’ve only had 1 US consumer internet win since Snapchat. Which I would argue is Discord.
MORE BROADLY OUTSIDE OF CONSUMER INTERNET
Sam Altman talks about how he regrets the advice that he gave while at YC. And how OpenAI is a counter signal to all of that previous advice.
Ran as a research lab. Insane amounts of investment poured in to results that were not even clearly valuable at the time. The breakout happened, not even with the first release of ChatGPT. But with the release of ChatGPT 3.
This was not the be nimble, iterate quickly, talk to customers approach that is often parroted as the only way to build.
So what then, as founders and builders in general, are we supposed to do?
This viral moment from David Holz (founder of Midjourney), counter to a VC post, stirred a lot of emotions recently.
As I wrote in “SAVE THE INTERNET”, we have to lean into serious alternatives if we want to discover new land. I also recently tweeted that every company I see coming out of SF looks exactly the same, because they believe in the same narrow cone of values that will progress their field / society at large.
David’s tweet was a more emotionally driven response to how he’s crafted Midjourney. But it’s clearly one of the truly different companies in SF. They haven’t raised a single dime from investors. They focused as a research org, and utilized Discord as their front end and network. Being quite profitable, they reinforce their research lab structure and have several hardware moonshot projects being explored internally.
This goes beyond any advice an investor or thought leader could share. Because it’s directly against their incentives of capital distribution.
And this is my final point. We all have to graduate away from “pattern matching” and move towards nuanced referencing.
The more founders that can develop this skill, will also pull the capital class with us. Arguably this is significantly more important at the pre-seed and seed stage. But I do believe it becomes incredibly critical in a world of AI distributed advantages around the mean. There won’t be many easily earned wins, and not having differentiated references will hurt your chances even more.
The image above from my essay “The Body As Mixer” shows how DJ’ing is structured in relationship to mined references. And it’s the interaction with the body, that turns it into a new idea.
If you were to apply pattern matching to DJ’ing, you wouldn’t produce a compelling effect on the audience. Instead they’d groan, knowing every turn the set is taking because they’ve seen that set before. They’ve felt that set before, even if there is a 3% difference.
Nuance referencing is producing a new potential history. It’s excavating something previously secret, that we haven’t felt in a while. And then the new question emerges → do you share what you’ve found? Someone important once said companies are built on secrets…
I don’t do edits really, so excuse typos and things that don’t make sense.
Thanks so much for giving me your attention. I hope it was worth it, if not… unsubscribing will not hurt my feelings, and will give you back time you literally cannot have back.
Much love.
Live in the light