Where should you work?

New college grads are changing their preferences quickly in the age of AI.

I meet with 200-300 recent college grads per year. I do this because a) I think these people can contribute a lot to the startups I work with, b) I genuinely enjoy it, and c) These conversations help me understand the world better.

The most common topic of conversation in these discussions is: Where should I work? For all the education that happens in universities, new grads barely know where to start to answer the questions, “What is the point of work and what should I look for in a job?”

The Algorithm

Because most college grads don’t have a strategic framework for making career decisions, they generally default to what I call “the Algorithm.” The algorithm says, “Go work wherever the smartest people 1-2 classes above me went to work.” Or even more simply, “Go work where the people I admire work.”

The algorithm is simple and generally effective. It turns out smart people do generally make good decisions, and bootstrapping off of other people’s decisions in the absence of information is a reasonable tactic, at least until you can collect your own first-party data.

One thing I’ve observed in new graduates is that they have almost no business judgment. They are not asking, “Will this business I’m about to join succeed?” because they know that they are not well-equipped to answer that question. They are also not asking, “Is the work interesting?” because there is a (probably correct) default assumption that entry-level jobs are all pretty grindy.

The Derivative

Because the algorithm is recursive, it creates a flywheel effect. As a company, you either have a new grad flywheel going, or you don’t. If you have the flywheel going, it’s reasonably straightforward to sustain. But if you don’t, getting into the flywheel is really hard. You can’t argue your way into it, because the algorithm is mimetic, not reasoned.

A few startups have hacked into this loop, but it’s been hard. Palantir was probably the most effective. They did this by inventing a new job — the forward deployed software engineer. This attracted people who wanted to be engineers, but also wanted to interface with customers. More recently, Scale managed to break into the flywheel by creating a localized network effect: They hired the most elite MIT undergraduates over a 2-3 year window.

A more common method is to take advantage of the fact that engineers love to work on hard problems. In his biography of Elon Musk, Walter Isaacson describes this strategy as “putting out the bat signal” on a hard problem that is ready to be solved. When combined with a rigorous interview process, this can be an effective way to get a flywheel going.

Finally, there is the brute force method: Paying a lot. That works too, if you can afford it.

The AI Cataclysm

We are early in grokking how fundamentally these existing talent market dynamics have already been disrupted by AI. In the last twelve months, I’ve noticed a big change in the young people I’m interacting with.

The big change is that people are now (correctly) looking at jobs as being either “AI-good” or “AI-bad.” AI-good means that the job is fundamentally on the right side of history, and therefore offers a strong career path, and AI-bad means jobs that are on the wrong side of history. White collar services businesses, because they are so ripe for AI, seem to have had the biggest swing in terms of being viewed as no longer on the right side of history.

The tip of the spear for the changing talent market landscape is the AI labs. To state the obvious, anyone that can work at those labs does go work at those labs. The compensation is extreme, it’s high status to tell your friends you work there, and you get to be at the cutting edge of a transformational technology evolution. People don’t turn these companies down.

The less well-understood dimension, though, is the subtle, but I believe very real, shift away from Big Tech and Wall Street, toward AI-native companies. Young people want to be a part of the future, and startups have never been more appealing as an option. There is the added bonus that after more than a decade of inflation, startup compensation is much more appealing than it used to be.

AI has had a huge leveling force already on the concept of “work experience.” No one has more than five years of experience with ChatGPT and LLMs, so a 24-year old who has spent the last five years deep in the weeds of AI may actually have more relevant experience than someone who has been working for much longer. This is a big advantage for young people — it makes working in an AI-native company structurally superior to jobs where they will need to work for decades to reach a comparable level of relative experience.

Young people know this, and they are already changing their behavior. I think it’s possible that in just a few years, we’ll see a shift from roughly 20% of elite college grads going into tech jobs up to potentially 50%. Just like Palantir invented the forward deployed software engineer, we’re going to see the emergence of the “AI generalist” — someone native and proficient in AI who can wear a lot of hats and do a lot of different jobs. The aperture will open for a broader group of non-technical applicants to fill these roles.

How to pick within AI?

Within AI, there are two simple criteria that really narrow down the list of potential companies to join: a) Does the company do something useful? and b) Is it growing fast?

These questions are interrelated. Today, so many people are connected to the Internet, and they are eager to adopt AI. So if you build something useful, distribution is happening very quickly. The result is that there are more and more companies joining the “$100M club” — the group of companies that are very rapidly scaling from $0 to $100M in revenue. These companies need to hire aggressively, and many are still small, so it is possible to have a big impact. There are dozens of such companies to choose from.

A Note on Builders

There is one more group of people worth touching on. And that is people who want to build a great company, rather than joining a great company.

As I suggested with “the Algorithm,” the average new grad is mostly focused on joining a company that is already successful, with a deep bench of great people they can learn from. Builders, by contrast, optimize for joining organizations where their contributions can move the needle on making the company they join succeed, regardless of its current level of success. These people tend to have extreme confidence and extreme agency. They look for companies that have the potential to succeed, and where they believe their presence can increase the likelihood of getting there.

In the software era, Builders made up no more than 1 in 100 graduates from top schools. There were two constraints on the size of this group, one hard and one soft. The hard constraint is that you need a really high level of self-belief to imagine that you as a fresh graduate can move the needle on a business’s potential success. This hard constraint is a fundamental aspect of human nature and culture. The softer constraint — which is now easing in the age of AI — is the extent to which lack of experience does actually limit your ability to contribute. This has changed rather dramatically.

With regard to Builders, my contention is that the ratio of builders coming out of top schools has the potential to increase from sub-1% to roughly 5%. If this happens, it may have a really positive effect on the startup ecosystem as a whole, because builders are the people who really drive companies forward. And it will be just as transformational for the people who choose to make this leap. They will ultimately be the ones defining our AI future.