Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

What the Erdős Problem taught me

Straightforward problems may need these three questions.

[Photo: Getty Images]

Profile photo of Neel Somani

Founder · Eclipse Labs

Consider this simple question.

You are standing in front of an ordinary house. How many doors does it have?

Most people answer immediately. One. The front door.

But pause for a moment. Does a locked side entrance count? A cellar hatch? A doorway that was sealed years ago and can no longer be opened?

The question sounds trivial until you realize the hard part was never counting. It was deciding what counts.

Often, AI fails to answer questions like these. These are known as “Erdős problems,” named after the Hungarian mathematician Paul Erdős. These problems are designed to look deceptively simple. The problems are easy to state informally, easy to agree on, and often framed as a single, clean question.

Once you notice that pattern, you start to see it everywhere.

Recently, I’ve spent a lot of time using AI to work on Erdős problems. The work itself is technical, but the lesson it reinforces has very little to do with math.

Simple concepts became genuinely hard, and in many cases, the difficulty wasn’t obvious from the way the problem was written. The complexity only revealed itself upon manual analysis of what the problem really meant.

That experience changed how I think about problems more broadly. I now see the same pattern show up in leadership, strategy, and decision-making, particularly in how AI changes what looks obvious versus what actually matters.

Some of the most dangerous problems aren’t those that look complicated, but the ones that look obvious.

The hidden constraints people don’t see

Usually, when you have a problem that is easy to explain but extremely hard to solve, the issue is that you don’t fully understand the problem yet and aren’t framing it correctly. 

You may think everyone is aligned because they agree with the problem’s wording. In reality, people are often solving different versions of the same problem.

In math, this appears when a problem is written clearly without explicit constraints. Some solutions technically answer the question, but they are considered trivial and therefore out of scope. 

That distinction is understood informally by people who work on these problems, but it isn’t obvious to outsiders.

Leaders often run into the same issue. A problem can look fully specified when it isn’t. Without clarity on constraints, teams spend a lot of time working on something that technically answers the question but misses the point.

3 questions that reveal the real problem

When a problem looks obvious, that’s often a signal to slow down and ask more questions:

1. What solutions would technically answer the question but are considered trivial or out of scope implicitly?

2. Do I truly understand the problem, or am I only making progress on it?

3. Can I consider multiple reasonable ways to approach this problem on the spot?

Those questions are often more informative than early execution.

Progress isn’t always understanding

It’s easy to make progress by combining existing results to prove a slightly tighter bound or to show that some constraint to the problem exists. That kind of progress is limited.

True progress is when there’s no existing result that’s trivially applied. It’s subjective, and much harder to recognize. You have to understand a problem to make progress on it in that sense.

This distinction matters for leadership. Brute force and scale are a great way to pick low-hanging fruit. They’re also a way to identify interdisciplinary findings that might have been otherwise missed. But brute force, scale, or more effort does not automatically translate into insight.

I see this show up when leaders assume that more resources will eventually force clarity. Sometimes that works, but often it doesn’t.

Insight is a different property. It comes from identifying the right way to look at the problem, kind of like the right way to state it.

When you look at a problem the right way, solutions often become much more straightforward.

Why leaders misjudge difficulty

When viewing AI systems through the lens of Erdős-style problems, most non-technical leaders underestimate one thing in particular.

Often, they assume that all problems are equally difficult. They see a class of problems being solved and assume that it’s only a matter of time before they’re all solved.

Sometimes that’s true. In the case of math, it’s less true because there are problem categories much harder than non-mathematicians can imagine.

To decide whether a problem is genuinely tractable or just deceptively clean, I examine whether I can immediately think of multiple reasonable approaches.

If there are plenty of things worth trying, that’s usually a good sign. If I’m not even sure where to begin on a problem, that might be an indicator that it’s harder than it looks.

Leaders can apply this directly. When evaluating an initiative, ask whether there are multiple viable approaches that make sense or whether everyone is circling the same idea without real alternatives.

A lack of starting points is often more informative than a lack of answers.

Clarity before commitment

If advising a CEO evaluating an AI initiative, I would want them to internalize this from my Erdős work: Make sure you’re extremely clear about all the details and constraints of what you want to solve.

That clarity is about understanding what counts as a real solution, what is considered trivial, what constraints are assumed but not stated, and where insight actually needs to come from.

Straightforward problems aren’t always straightforward. In fact, the problems that look the cleanest are often the ones that deserve the most scrutiny. Recognizing this early puts leaders in a much better position to deploy talent and capital effectively, rather than discovering too late that the hard part was never written down in the first place.

Neel Somani is the founder of Eclipse Labs.