As young people, conventional wisdom tells us to optimize for learning skills that will pay dividends in the decades to follow. Though what skills are worthwhile investments and how focused should we be? Spending your 10,000 hours on the wrong skillset may be a life altering blunder. Let’s spend a few minutes considering this question.

This skill selection process is made more difficult by our ever-changing world and introduction of new tools: the problems and solution spaces are in constant flux. I breakdown this question on two dimensions: (1) breath vs depth of skills and (2) how do find areas of depth.

Breath vs Depth
The generalist vs specialist debate has been well covered, though an important consideration is the evolving nature of tools. Let us consider the recent release of GPT.

GPT makes a case for the generalist. Suddenly, a generalist has the knowledge of a domain expert. For instance, I prefer using NoSQL databases, so haven’t invested much time in learning how to write advanced SQL queries, but now with GPT I can write SQL queries like a SQL database (almost) pro.

Yet, the inverse may also be true. GPT can do a solid job at a given task. For our SQL query example, maybe a standard deviation above the average developer when taking account time and accuracy. So the generalist who knows a little SQL doesn’t add any value. The only valuable individual is someone beyond that standard deviation. That is where a specialist would come into play, who spends time to achieve mastery.

So GPT simultaneously makes the generalist more capable but also less valuable with respect to a single task. While I specifically discuss GPT, the tool can be viewed as just the latest in a progression of tools: AWS for server management, Docker Kubernetes for devops, 3rd party auth providers for login, and the list goes on. A similar paradigm exists beyond the bubble of webapp development. Canva makes it easier to make graphics. The iPhone to take near production photos.

Being useful is easier but producing a better result than average result is harder.

This arrangement is perfect for the “T-shaped” profile– someone with a broad foundation and a few areas of deep expertise. Thanks to new tools, the necessary depth of that foundation is shallower than ever. It seems remissful to overlook such low hanging fruit.

Finding Areas of Depth
So, striving for a “T-shaped” profile, you’ve learned a bunch of basic skills and now decided you want to go deep in one or two few areas. How do you select those skills? Well, ideally, they should maximize your value creation and your ability to capture it.

Let’s take the example of a software engineer. They can produce incredible value: a handful of engineers or supposedly 250 devs in the case of Twitter can build a product for hundreds of millions. Yet most software engineers fail to capture their value and make a healthy but relatively modest amount (especially when being paid in salary). This is since most software engineering isn’t a differentiable endeavor. Most internet companies face more market rather than technical risk. Since a good engineer and a decent engineer can both ship code, they tend not to make or break a company. Instead, product and go to market tend to be deciding factors. While arguably less crucial than software engineering, these are differentiating skillsets. People in with these skillsets are often in vp roles or are able to capture more of their value. Hence, Apple paying John Ive’s 10x in base salary than Google’s top engineer.

For me this means, I’m not going to spend time learning every single web framework. If I don’t have a burning desire to learn Next.js, I won’t. If I don’t have a passion for optimizing hosting efficiency, I’ll use a fully managed container service. I’ll stick with my tried-and-true stack that I’ve been using for the past 6 year since this won’t make or break a project. Instead, with that extra time, I’ll get in more iteration cycles building product and hone that skillset. If I am going to go down a technical rabbit hole, it will be unique and differentiable.

These differentiating skillsets or unique sets of knowledge tend to be hard to come by and hard to measure. If you can teach or test a skillset, it becomes overly commoditized and less of a competitive edge.

While developing the generalist base of your “T” is almost entirely inductive– e.g. asking yourself what are the most important technologies and industries in the next 10 years and learning accordingly, I think the best way to land on these areas of depth is to add an element of deduction. To achieve mastery beyond an AI or any competition you have to be obsessed by something. By guiding through deduction and going by feel you can converge at this obsession and find a niche. In reality, this search will be a synthesis of the induction and deduction, though this deductive element is not to be overlooked.

This element of deduction makes you unique and provides the basis of your differentiated value proposition. If you were to work entirely inductively, you are more likely to end up like everyone else. What is powerful is combing the two. Going into an industry or learning new tech by passion and then being smart about working within that sector or paradigm inductively. Thus, like with the t-shaped profile we see the advantageous nature of these hybrid approach.

Too often I think we constrain these personal interests as they often seem impractical at first. I think this makes us boring, too similar, and by extension noncompetitive. We should follow these interests and let them bring us through rabbit holes. While “follow your passion” is rather cliché, I think it is something our generation too often forgets. Most of the projects I see at Stanford (including my own) aren’t sown out of passion but instead because they make sense. We’re over optimizing for foreseeable outcomes when many of the best outcomes cannot yet be imagined. We should be less afraid to make sense and more eager explore. The cost to do so is lower than ever. ∎