I expected to hear something about picking the right role, or the best partners, or the place where I’d get the most reps. Instead, his logic was simple:

Early in your venture career, you’re tuning the algorithm in your head for what a truly great company looks like. And like any algorithm, it’s a garbage in, garbage out problem. If you join a top firm, you’re more likely to see what “great” actually looks like. If you join a weaker one, the data you’re exposed to might steer you toward false conclusions.

I’ve been thinking about that a lot lately as I watch investors try to pattern match their way into finding the next super-compounder.

We look for what we’ve seen before

One of the most consistent lessons I’ve learned in VC is that people tend to look for whatever they’ve personally witnessed.

If the biggest company you’ve seen went from seed to a $1B exit, that’s the pattern you look for.

If you’ve seen something go from Series B to a $20B+ outcome, those characteristics become your pattern.

If you worked inside a rocket ship during its scaling phase, you start looking for echoes of that experience everywhere.

When I joined Spark, I was lucky. I got to see a seed investment go from sub-$5M valuation to a billion-dollar outcome. And I also watched another business go from Series B into what ultimately became a $20B public company. Whether I realized it or not, those companies shaped the contours of my mental model.

And I’ve noticed that other investors’ pattern-matching is almost entirely defined by whatever they’ve been exposed to, too. We all tend to look for what we found before.

The problem with secondhand pattern recognition

Pattern matching gets especially tricky when the “data” wasn’t experienced up close.

For example, I see investors who have only witnessed companies raise their way from seed to a high-priced Series C—without seeing how those companies actually performed afterward. So their pattern becomes: great companies raise fast and raise big. But that’s not a pattern of a great business… it’s a pattern of fundraising traction.

On the flip side, you can study super-compounders from afar—companies that went from zero to $100B—and think you understand the pattern. But if you haven’t seen the messy middle up close, you’re often pattern-matching against a very shallow silhouette of the business, not the actual contours.

Shallow data leads to shallow conclusions.

The AI era has amplified this effect

Right now we’re in a moment where AI model companies are all the rage. And it’s incredibly tempting to draw narrow, simplistic conclusions from the outside.

For example: Anthropic and Mistral were both founded by world-class research scientists from famous labs → therefore, to find the next one, just back elite research scientists from famous labs.

Simple. Clean. Wrong.

Nobody on the outside truly knows what those founders were like—how they thought, how they operated, what their conviction looked like, or how the companies were shaped in their earliest days. And those companies represent a very specific type of business model and market timing.

Super-compounders of tomorrow will almost certainly look different. They always do.

A shallow pattern-matching mindset will anchor you to what worked before, not what will work next.

The self-reinforcing trap

The dangerous part is that shallow pattern matching is self-reinforcing.

If investors all look for the same founder profile as the last successful company, the entire funnel narrows around that profile.
If they only look for companies that resemble the last super-compounder, the market quickly over-rotates.

I think that’s exactly what we’re seeing today: extraordinarily narrow founder archetypes getting all the attention, and everything on the edges getting ignored.

Blind spots

I’ll be the first to admit: I’ve never been involved in a true zero-to-50B+ super-compounder. I want to be.

But I don’t think the path is to pattern match the last generation’s winners using a shallow understanding of what they really were. And I also don’t want to limit myself to only the kinds of companies I’ve personally seen—those that reached a billion or ten billion in enterprise value.

Those are great companies… but they’re not necessarily the legendary ones.

In the end, my professor was right: your algorithm is shaped by your data. But the data you’ve seen—up close or from afar—can also box you in.

We tend to look for what we found before.

And maybe the most important investor skill is to train ourselves to look beyond that. Because the next generation of super-compounders won’t look like the last one. They’ll almost certainly be extraordinarily different.

 

NOTE: This post was written through a combination of my own dictated thoughts and a GPT trained by many years of my prior writing. This was a bit of an experiment, and I’m posting it with minimal edits.