Escaping the Jack-of-All-Trades Trap: The Late Specialization Path

• Intended readers: People who feel “I’ve tried a lot of things, but none of them deeply.” Anyone navigating a career pivot or unsure where to head next.
• Prerequisites: None.
• Reading time: About 11 minutes.

Overview

“I’ve dabbled in lots of things, but never gone deep on any of them.” A lot of people feel this way. The dominant cultural story rewards those who found their One Thing early and dug relentlessly into it.

Research, however, paints a slightly different picture. Many top performers actually share a pattern of diverse experiences followed by late specialization. This is the “sampling period → specialization” model David Epstein synthesized in Range (2019). Especially in modern problem domains where the rules are fuzzy (Wicked environments), people with cross-domain experience often have an edge.

There is, though, a critical caveat. A sampling period and “permanently broad and shallow” are not the same thing. The first comes paired with a decision, somewhere along the way, to commit to a deep axis. The second keeps spreading without ever committing — and ends in the “jack of all trades, master of none” pattern.

This article is for people who have already “tried a lot of things.” It walks through (1) why the sampling period is not wasted, (2) why late specialization is rational once you understand Kind vs. Wicked environments, (3) the single difference that separates a “Kiyou Binbou” (器用貧乏 — literally “skillful but poor”: jack of all trades, master of none) from a late-specialization candidate — the decision to build an axis — and (4) a four-step practical guide for what to do next.

I happen to be someone who landed on a technical axis early, so I’m going to be careful not to write from above. Rather than “here’s what worked for me,” the stance is “here’s what the research and others’ stories suggest about a different path.”

“I’ve tried a lot, but never deeply”

In career and life-pivot conversations, the same phrasing keeps coming up:

“I’ve done a lot of things, but none of them deeply.” “Whenever something catches my eye, I jump in — but I never master anything.” “It feels too late to become a specialist now.”

A lot of people will recognize that voice in themselves. Social media and the bestseller shelves are dominated by stories of people who “found their thing early and went deep.” The more of them you read, the more your own résumé starts to look uselessly scattered.

That feeling shouldn’t be shamed. If anything, it’s evidence that you’re trying to evaluate your own career honestly.

But the punchline up front: it isn’t necessarily wasted.

The idea of a “sampling period”

In sports, music, and academia, the popular image of world-class performers is “they picked one thing early and grinded only that, forever.” This is the picture Malcolm Gladwell popularized as the “10,000-Hour Rule”¹.

Once you actually look at the data, a different picture emerges.

In Range: Why Generalists Triumph in a Specialized World (2019), David Epstein argues²:

Early specialization is the exception, not the rule, even among top performers. Many of them passed through a sampling period across multiple domains before settling into their eventual specialty.

The signature contrast is Tiger Woods vs. Roger Federer². Woods is the icon of early specialization — the kid who reportedly held a golf club at age two. Federer, meanwhile, played soccer, skiing, wrestling, swimming, skateboarding, basketball, and tennis as a child, and only narrowed his focus to tennis relatively late.

The same shape repeats across genres:

• Vincent van Gogh: before becoming a painter, he worked as an art dealer, teacher, missionary trainee, bookseller, and trainee preacher².
• Gunpei Yokoi (Nintendo, designer of the Game Boy): his philosophy of “Lateral Thinking with Withered Technology” combined mature, off-the-shelf components rather than bleeding-edge ones to build the Game Boy (including Color), which sold roughly 118.7 million units worldwide². It’s a textbook case of combining knowledge from multiple domains in a deliberately late, integrative way.

The takeaway: the sampling period itself is not failure or detour. For people who eventually settle into a specialty, it tends to be a necessary investment in figuring out what fits and what doesn’t.

Kind vs. Wicked: where late specialization actually works

That said, “sample first, then specialize” isn’t a universal rule. Whether it works depends on the nature of the domain.

The most useful frame here is the Kind learning environment / Wicked learning environment distinction from Robin Hogarth, Tomás Lejarraga, and Emre Soyer (2015)³.

Characteristics of a Kind environment

• The rules are stable and don’t shift much.
• Feedback is fast and accurate (the consequences of an action come back quickly and correctly).
• Similar problems recur.
• Examples: chess, golf, typical programming exercises, the basic technique of an instrument.

In Kind environments, starting early and going deep tends to pay off. The rules don’t change, so accumulated knowledge doesn’t decay. The classic 10,000-hour repetition story works in domains shaped like this.

Characteristics of a Wicked environment

• The rules are unclear, or change over time.
• Feedback is delayed, or arrives in misleading forms.
• Each situation is novel; past patterns don’t transfer cleanly.
• Examples: executive judgment, launching new businesses, complex client work, politics, hard medical diagnoses, big life decisions.

In Wicked environments, early specialization can become a trap. Hogarth and colleagues describe specialists who, through repetition in distorted-feedback environments, end up “successfully” learning the wrong patterns. One of their illustrations: a physician highly regarded as a typhoid specialist who turned out to be “a more productive carrier of typhoid than Typhoid Mary” herself³. He was confident in his diagnostic specialty, while his own asymptomatic carrier status — outside that specialty — sat in a complete blind spot. A vivid case of expertise inside a Wicked domain producing a self-flattering yet dangerously incomplete map.

Which environment are you in right now?

This becomes the operative question.

• Software syntax, specific algorithm implementations, an instrument’s basic fingerings — these lean Kind.
• Organizational management, launching new services, complex clinical judgment, big career decisions — these lean Wicked.

In Wicked environments, pattern recognition built up across multiple domains often beats single-domain depth. The patterns may not repeat within a domain, but they often rhyme across domains. Epstein calls this “conceptual reasoning skill that can connect new ideas and work across contexts”².

The strengths of late specialization

People who specialize after a meaningful sampling period sometimes carry strengths that single-track specialists don’t.

• Cross-domain pattern recognition. A pattern seen in one field gets reused in another. Customer psychology learned in sales feeds product design; feedback design learned in teaching feeds management; etc.
• Tolerance for uncertainty. Repeatedly diving into new territory builds an attitude that doesn’t panic in unfamiliar situations. In Wicked environments, this often matters more than any specific technical skill.
• Higher-resolution metacognition. You know from experience what you’re good at, what you’re bad at, and which environments let you concentrate. This is hard to acquire if you only ever lived inside one domain.
• Communicative range. Speaking the languages of multiple industries and roles makes you a translator. You become valuable as a bridge between teams, departments, or disciplines.

None of this happens automatically, though. These strengths only switch on once they’re paired with the “decision to build an axis” — which is the next section.

The one thing that separates “Kiyou Binbou” from a late-specialization candidate

This is the heart of the article.

The sampling period itself isn’t the problem. The question is whether it stays permanent or eventually converts into an axis.

	“Kiyou Binbou” (jack of all trades)	Late-specialization candidate
Background	Diverse experiences	Diverse experiences
Breadth of shallow knowledge	Wide	Wide
Deep axis	Never built	Built at some point
End state	“Can do anything, expert in nothing”	Multi-domain experience + deep axis = strong

The only thing separating them is whether you decide to build an axis.

That’s a sharp framing, but it’s also hopeful. Whether your sampling period was wasted or not is decided after the fact, by whether you go on to build that axis.

For people thinking “it’s too late to become a specialist now”:

• Plenty of people set their axis in their 30s and reach the front line in their 40s.
• Plenty of people reset their axis in their 40s and become respected senior figures in that field in their 50s.
• More fundamentally, in Wicked environments, the structure actually favors people who sampled long and broad before committing².

It might not be that you’re “too late.” You may simply not have made the decision to build an axis yet.

What to do next — a four-step guide

To keep this from staying abstract, here’s a practical four-step process.

Step 1: Inventory your past experience

Paper or digital — doesn’t matter. Write down every job you’ve held, every topic you genuinely studied, every hobby that stuck. Filter for things that lasted at least three months; that cuts most of the noise.

For each item, jot down “why I started doing this” and “what I got from it.” Whether it was money, curiosity, or a personal connection changes the meaning.

Step 2: Find recurring patterns

Lay it all out and look for themes that show up again and again. For example:

• Different industries, but you always ended up working on “designing how the system runs.”
• Different roles, but you were always “the translator between people.”
• Different domains, but you kept getting called in during the “build the new structure” phase.

These themes are usually your hidden axis candidates. Because they feel completely normal to you, you tend not to notice them — they need to be pointed out from the outside. Showing your résumé to someone you trust and asking what they see is a useful shortcut.

Step 3: Decide where to put your axis

Once axis candidates surface, ask the next question:

• Is this axis in a Kind-leaning domain or a Wicked-leaning domain?

This distinction changes how you build the axis itself.

Building a Kind-leaning axis: Pick objects whose depth can be measured externally. Examples: certifications in a specific domain, benchmark exams, deep reading of canonical libraries or codebases, competitions, systematic reading lists for the canonical books of a field. You’re aiming to produce visible artifacts — “I can demonstrably operate at this level in this field.” Repetition in a fast, accurate-feedback environment is the efficient path.

Building a Wicked-leaning axis: Stacking individual facts works less well here. Invest instead in conceptual frames that cut across the domains you’ve already lived in. Systems thinking, organizational development, strategy frameworks, the vocabulary of complex systems — tools that retroactively put an “interpretive layer” on top of your past experience. Use case studies (your own and others’) as raw material, and practice re-reading them through these frames. The depth indicator here isn’t a certificate or a test score; it’s whether you can take someone else’s problem, lift it one level of abstraction, and explain it back.

This split matters. In a Wicked domain, “I’ll just go pick up one more certification and that will deepen me” tends not to work as well as people hope. Conversely, in a Kind domain, “I want to be someone who can talk about their experience” alone stays shallow. Match the building method to the type of domain. That’s the core of Step 3.

Step 4: Set a focused window on the axis

Once you’ve decided, deliberately carve out a window where you concentrate there. This is the hardest step. The longer you’ve been a sampler, the more easily you derail when the next interesting thing shows up.

• Set a focus window of about 1 to 2 years.
• During that window, deliberately reduce new sampling.
• Inside that axis, produce a meaningful body of output (work, writing, a portfolio of cases you can talk about).

Once a real body of work exists, your own identity shifts: “I am a person of this axis.” External perception shifts too. That’s what produces the visible difference between “jack of all trades” and “T-shaped” or “π-shaped.”

Caveat: how to tell whether to keep sampling or build the axis

Finally, some criteria for when you’re stuck on “should I keep sampling, or commit?”

Signs that it might be time to commit:

• Each new domain you try gives you more déjà vu — you keep recognizing structures you’ve seen before. The marginal yield from sampling is diminishing.
• When asked “what’s your theme?” your hesitation is starting to feel like a real signal, not just modesty. That’s the absence of an axis showing up.
• A backlog of cross-domain output is forming in your head — articles you want to write, projects you want to run, structures you want to propose. That’s usually a sign the axis is ready to be declared.

Signs that more sampling is still legitimate:

• The new domains genuinely are new to you; déjà vu is still rare.
• You have the mental and financial slack to keep exploring.
• Committing to one axis right now would feel dishonest because more than one of your serious themes hasn’t fully ripened.

The exact call is personal, but the meta-rule is: consciously choose when to stop sampling. The most expensive failure mode is “I looked up and I’d just drifted into the next thing again,” over and over.

Summary — your résumé isn’t wasted, but build an axis

To recap:

1. Many top performers reach the top via a sampling period followed by late specialization. Early specialization is the exception, not the rule².
2. In Wicked environments (unclear rules, distorted feedback), multi-domain experience can outperform single-domain depth. Many of the most important problems in modern life have this structure³.
3. “Kiyou Binbou” and “late-specialization candidate” diverge on a single decision: whether you build an axis. Your past doesn’t change. What the past means does — and that meaning is set by the next decision.
4. Four steps: (1) inventory, (2) find patterns, (3) design the axis (Kind or Wicked), (4) set a focused window and produce real output.

If you’ve been feeling “I’ve done a lot, but none of it deeply” — your résumé is not waste. Your past won’t change, but the meaning of your past will be decided by the choices you make from here.

---

Series articles:

• From Game Hacking to Psychology: How One Specialty Opened Up the World — A first-person account of someone who set their axis early.
• I-shaped, T-shaped, Pi-shaped: The Depth × Breadth Skill Matrix — Conceptual organization of skill shapes by depth and breadth.

Related articles

If this topic resonates, you may also be interested in:

• Action Plan from Your 40s, Part 3: A Survival Skill Strategy — Axis-building strategy by life stage.
• A Cognitive Investment Strategy by Age in the AI Era — Designing learning around cognitive peaks and troughs.
• Is “I Love Programming” Still Enough in the AI Era? — Rebuilding T-shaped skills.
• Why the Market Value of Deep Specialists Is Exploding — The scarcity premium on deep expertise.

References

References cited in the text, numbered in order of appearance.

1. Outliers: The Story of Success — Malcolm Gladwell (2008). Little, Brown and Company. [Reliability: Medium] — The popular book that popularized the “10,000-Hour Rule.” K. Anders Ericsson, whose original research the rule draws on, later added significant qualifications to Gladwell’s interpretation. ↩︎

2. Range: Why Generalists Triumph in a Specialized World — David Epstein (2019). Riverhead Books. ISBN 978-0735214484. [Reliability: Medium-High] — #1 NYT bestseller. Argues against early specialization, makes the case for the sampling period, and uses cases such as Tiger Woods vs. Roger Federer. Because it synthesizes a wide body of research as a general-audience book, individual claims may warrant secondary verification. ↩︎ ↩︎² ↩︎³ ↩︎⁴ ↩︎⁵ ↩︎⁶ ↩︎⁷

3. The Two Settings of Kind and Wicked Learning Environments — Hogarth, R. M., Lejarraga, T., & Soyer, E. (2015). Current Directions in Psychological Science, 24(5), 379–385. doi:10.1177/0963721415591878. [Reliability: High] — The peer-reviewed paper that formalized the Kind / Wicked learning environment distinction. Includes examples of expertise mislearned via distorted feedback, including the Typhoid-Mary-style anecdote referenced above. ↩︎ ↩︎² ↩︎³