Why the Axis-less Generalist Hits a Ceiling: You Need One Deep Axis Before Going Full-Stack

• Who this is for: Early-to-mid-career engineers unsure which way to go after being told “these days you should be able to do everything,” plus tech leads, EMs, and HR people who design engineer development plans
• Assumed knowledge: A basic feel for how front-end / back-end / infra roles divide up work, and comfort with talking about the “depth” and “breadth” of expertise
• Reading time: about 13 min

Overview

“The engineer of the future shouldn’t hole up in one technology — go full-stack, become a generalist.” That advice rings louder every year, and there is nothing wrong with it on its face. The problem is that it gets received with one crucial step skipped over.

This article asks a sharper question. Where do you end up if you head straight for “shallow and broad” without ever building a single deep axis? Can you actually succeed by aiming for full-stack or generalist work with no axis at all?

Here is the conclusion up front. No study has measured this directly with numbers. But several primary studies point, consistently, in the same direction: breadth doesn’t compound without a foundation. High-value specialized skills sit only on top of foundational skills, and that dependency runs in one direction¹. In professional and skilled occupations, breadth barely moves wages — and an uneven skill profile is actually a penalty².

Push a little further and something interesting comes into view. When you look closely at “generalists who succeeded without an axis,” they have almost always built an axis somewhere outside technology — integration, coordination, or business-building. Genuine success with no axis at all is almost never observed. So the road to success reduces to two paths — “build a deep technical axis” or “build a non-technical axis” — and both, in the end, converge on building some axis.

In this article I test that picture against primary sources in labor economics and organizational research, and close with the practical implication that “depth comes first in development order,” plus an exit for people who have already gone shallow-and-broad. The counterargument that “AI means you don’t need an axis anymore” carries enough weight to fill an article on its own, so a sister piece handles it in depth (see the end of this article).

“Full-stack isn’t the problem” — so what is?

Let me be clear up front. Full-stack engineers and generalists are not a bad choice. People who can move across multiple domains improve design coherence, fill the gaps in a team, and are especially valued for it. No argument there.

The problem lies elsewhere. It shows up when the advice “broader is better” gets quietly swapped for the practice of “so just dabble broadly and shallowly from day one.” You touch several domains on the surface without ever, in a single domain, having the experience of pushing one notch further than where others stop. You burn through framework tutorials one after another — you’ve “touched” each one, but none of them are something you can be trusted to own. This article calls that state “axis-less shallow-and-broad.”

What the T-shape and π-shape actually mean as skill profiles, and why depth accelerates learning, I’ve covered in separate articles (see Related at the end). Here I look squarely at only what comes next: what happens, structurally, when you head for breadth before building an axis.

Why “axis-less shallow-and-broad” hits a ceiling

A caveat first. What follows is not something you can measure with a single number. Read it as a structural consequence drawn from several primary studies.

Specialization stacks only “on top of the foundation” — the asymmetry of skill dependencies

The strongest clue is a 2025 study published in Nature Human Behaviour¹. The team analyzed skill-to-skill dependencies across roughly 70 million job transitions and showed that human capital has a nested structure: the specialized skills that lead to high wages can only stack on top of more general, foundational skills. And that dependency has a direction. Specialized skills depend on foundational ones, but not the reverse.

The study names programming as an example. To acquire programming skills, you first need general knowledge such as mathematics and systems analysis, which themselves rest on deductive and inductive reasoning. And here is the key implication: un-nested specialization gets left behind in wage growth.

What happens when you look at “axis-less shallow-and-broad” through this lens? It maps onto someone who has stepped across the foundational base but has stacked not a single specialized skill on top of it. The nested structure tells us that breadth itself is not the problem — rather, breadth converts into value only once it has a foundation of depth underneath it. A building with a foundation but no pillars holds nothing above it, no matter how much floor area it covers.

A generalist’s productivity is bound by their weakest skill

There’s a second study that bears directly on the engineering occupation. A 2005 analysis using the NLSY (National Longitudinal Survey of Youth) compared, by occupation, how being a “generalist (balanced skills)” versus a “specialist (skewed skills)” affects wages².

The results split along occupational lines. In generalist-type occupations — management, sales, clerical work — skill breadth (balance) leads to higher wages, and a skewed profile is a penalty. But in professional, skilled, and operator occupations — the layer engineers belong to — breadth barely affects wages. Simply having a wide range of skills doesn’t translate into recognition for a specialist.

The study places the Law of the Minimum behind this. Originally an agronomy principle — that plant growth is constrained by whichever nutrient is scarcest — it’s applied here to productivity: a generalist’s productivity is constrained by their lowest skill. The shallow-and-broad worker, the moment depth is demanded, gets dragged down by their weakest point.

Layer the two studies together and you can see why “axis-less shallow-and-broad” struggles to reach high market value as an employed engineer. Specialization never accumulates, so wages leave them behind (the nested structure); spreading out broadly and shallowly never amounts to recognition as a specialist; and when depth is finally required, they hit a ceiling at their weakest point — the inherent fate of a generalist working style (the Law of the Minimum). This is not a talent problem; it’s a problem in how skills accumulate.

The real identity of the “successful generalist”

At this point some readers will think, “But I actually know generalists who are thriving without anything you’d call an axis.” That’s a correct observation, and an important clue.

High-quality research that generalists are at an advantage does exist. A study analyzing the career histories of roughly 4,500 S&P 1500 CEOs found that generalist-type CEOs earn about 19% more per year than specialist types³. At the executive level, broad experience is clearly a premium.

But read the study closely and two reservations surface. First, the same study reports there is no evidence that firms led by generalist CEOs perform better. Pay is higher, but on accounting and stock-price performance, firms with specialist CEOs do, if anything, slightly better. In other words this is a phenomenon of “the labor market putting a high price on general human capital,” not evidence that “generalists do better work.” A wage premium is not a productivity premium. Second, the subjects are CEOs and top executives — not something that transfers directly to engineers on the ground.

The theoretical pillar usually cited in defense of generalists — the jack-of-all-trades theory — draws the same line when you read it carefully⁴. The model behind the theory is “entrepreneur = generalist, employee = specialist.” An employee is evaluated on their strongest skill, but an entrepreneur’s whole venture is constrained by their weakest skill — which is exactly why entrepreneurs need balance. In other words, the theory actually endorses specialization for those who are employed. And the empirical side of the theory is fragile: when you statistically remove unobserved individual traits (raw ability and the like), the correlation between balanced skills and entrepreneurship disappears⁵.

Line these up and one picture forms. Most “generalists who succeeded without an axis” have, in place of a discarded technical axis, built a different axis. Management, product, business-building, team coordination and integration — these are perfectly legitimate “deep axes.” They aren’t “axis-less”; they have pushed one notch past where others stop, in a domain other than technology. The real identity of the observed success isn’t the absence of an axis but a change of axis.

It’s true that the era now lets one person build cross-functionally, and solo founding is on the rise. But that’s the entrepreneur’s context — a domain where, just as jack-of-all-trades theory says, generalists already hold the advantage. The career argument for employed engineers and the productivity argument for entrepreneurs need to be kept separate. Conflating them to conclude “so engineers should go broad without an axis too” is a leap in logic.

Two routes both reduce to building an axis

Pull the research so far into a single diagram, and the fork beyond “axis-less shallow-and-broad” looks like this.

flowchart TB
    A["Shallow & broad, no axis"] --> B["Build a deep technical axis"]
    A --> C["Build a non-technical axis"]
    A --> D["Build neither"]
    B --> E["Value rises as a T-shape"]
    C --> F["Thrive on a different kind of axis"]
    D --> G["Ceiling at the weakest skill"]

The people who succeed travel one of the top two routes — build a deep technical axis and lay breadth on top of it (a T-shape), or build a non-technical axis (a different kind of T-shape). Both reduce to “building an axis.” Only the third route — “keep spreading wider without building either axis” — heads for a ceiling, exactly as the nested structure and the Law of the Minimum predict.

So the most accurate answer to “Can you succeed by aiming to be a generalist without building an axis?” is this: the people who succeed have built some axis, whether technical or business. “Succeeding without building an axis” essentially almost never holds — at least, when you look closely at people who are thriving, you’ll usually find some axis behind them.

Implications for talent development

So far this has been about individuals, but it’s also a story about how you design development. A development policy of “have them aim for full-stack right away” or “let juniors do a bit of everything early” — however well-intentioned — carries the risk of mass-producing axis-less people.

The clue about ordering comes from the same Nature Human Behaviour study¹. That skill dependencies are asymmetric — specialization depends on foundations, but not the reverse — means there is a direction to the order in which you stack. The standard career trajectory is “start in jobs that require general skills, then transition into a profession” — specialization rides only on top of foundations. This backs the “build depth, then broaden” order from the structural side. The reverse order — broadening without a foundation of depth — runs against the structure.

To be careful, I should note a reservation: research that rigorously proves a causal claim that “T-shaped development itself is effective” is actually still thin. The T-shape is widely used as a talent-management framework — one industry survey reports that 84% of responding companies have adopted some form of the T-shaped skill model⁶ — but much of the case for its developmental effect rests on anecdote and framework, not on causal evidence like a randomized controlled trial. So this article’s claim is not “go T-shaped and you’ll become excellent” but the narrower one: heading for breadth without building an axis is structurally disadvantageous in its ordering.

In the Japanese context, the problem gets one notch trickier. The national “Job-Based HR Guidelines” officially describe traditional Japanese-style employment as “a system in which autonomous career formation is hard to achieve”⁷, and METI (the Ministry of Economy, Trade and Industry) has called for a shift to “skill-based talent development”⁸. The IPA’s skill-standard frameworks (ITSS / the Digital Skill Standards) organize a two-layer structure — the foundational literacy required of everyone and the deep-dive domains of specialists⁹ — so the design philosophy of “breadth for everyone plus depth for specialists” is, as an institution, already in place. The problem is execution. The share of Japanese companies reporting a shortage of DX-driving talent stands at 85.1%, and that number has barely improved in years¹⁰. The “mechanism for growing axes” — the precondition for letting people build breadth — still isn’t working well enough on the ground.

The practical guideline for development design is simple. First, have them build one depth they can be trusted to own. Once that axis exists, have them broaden into adjacent domains. Don’t reverse the order. This isn’t just for the talented few — precisely because you can’t rely on talent, the design of ordering is where the difference shows up.

For those already shallow-and-broad

Finally, for anyone who feels “isn’t it already too late?” If you’ve touched many domains shallowly and have nothing you’d call an axis — and you’re aware of it — there is an exit.

First, pick one of your existing “shallow experiences” and switch to going deep on it. The breadth you’ve touched isn’t wasted: it’s material for judging which domain is worth digging into, and once you’ve dug deep somewhere, the way you dug transfers to other domains. “Specializing late” is, in fact, the path many excellent practitioners have taken (a related article covers this exit in detail).

Second, you also have the option of picking a “non-technical axis” instead of insisting on technology. Integration, coordination, product, business — the second route in the diagram. What matters isn’t which axis you pick but experiencing “one notch past where others stop” somewhere, on one of them.

Building an axis doesn’t require raw genius. What it requires is choosing one thing and committing to stay there longer than other people do. Breadth you can always add afterward.

Summary

• Neither full-stack nor generalist is a bad choice. The problem is heading for “shallow and broad” without ever building a single deep axis.
• “Axis-less shallow-and-broad” tends, structurally, to hit a ceiling. Specialization stacks only on top of foundations (the nested structure), and a generalist’s productivity is bound by their weakest skill (the Law of the Minimum)¹².
• The real identity of the “generalist who succeeded without an axis” isn’t the absence of an axis but a change of axis. They’ve built an axis outside technology (integration, coordination, business). The strong research showing a generalist advantage targets executives and entrepreneurs, and doesn’t extrapolate cleanly to employed engineers³⁴.
• The success routes reduce to two — a “technical axis” or a “non-technical axis” — and both converge on “building an axis.”
• The development order is depth → breadth. The asymmetry of skill dependencies backs this direction from the structural side¹. In Japan, the very “mechanism for growing axes” — the precondition for building breadth — remains a challenge⁷¹⁰.
• People who have already gone shallow-and-broad have an exit too: pick one from your existing experience and switch to going deep.

This is the first of a three-part series (the structural argument).

• Where “AI Means You Don’t Need an Axis” Falls Apart — auditing the “you don’t need an axis” argument in the AI era
• Is “AI Does It, So Checking Can Be Light” True? — how to design the substance of verification

Related

You may also be interested in these related articles:

• I-shaped, T-shaped, π-shaped: A Depth-and-Breadth Skill Matrix — a systematic look at the definitions of “axis,” “depth,” and “breadth”; the concept primer this article builds on
• Escaping the Jack-of-All-Trades Trap: Specializing Late as an Option — a detailed treatment of the “exit” for people already shallow-and-broad
• From Game Hacking to Psychology: How One Axis Widened My World — a concrete example of how a single axis accelerates learning in new domains
• Jack-of-All-Trades or Stay in Your Lane? AI-Era Role Design by Company Size — “breadth presupposes a deep axis,” argued from the organizational-design side
• Your Career Plan Is Decided by “Contribution,” Not Skills — the axis of “where you stand,” not just the skills you hold

References

References are listed in the order of the citation numbers used in the text.

1. Skill dependencies uncover nested human capital - Hosseinioun, Neffke, Zhang, Youn, Nature Human Behaviour (2025; online 2024). DOI: 10.1038/s41562-024-02093-2. Analysis of ~70 million job transitions; demonstrates the asymmetry of skill dependencies and the nested structure. 【Reliability: High】 ↩︎ ↩︎² ↩︎³ ↩︎⁴ ↩︎⁵

2. Generalists and Specialists, Ability and Earnings - Sang-Hyop Lee, University of Hawaii Working Paper (2005). NLSY data. In professional and skilled occupations, breadth doesn’t affect wages and a skewed profile is a penalty; an application of the Law of the Minimum. 【Reliability: Medium】 ↩︎ ↩︎² ↩︎³

3. Generalists versus Specialists: Lifetime Work Experience and Chief Executive Officer Pay - Custódio, Ferreira, Matos, Journal of Financial Economics (2013). Generalist CEOs are paid about 19% more but don’t outperform on results (pay ≠ productivity). 【Reliability: High】 ↩︎ ↩︎²

4. Balanced Skills and Entrepreneurship - Edward P. Lazear, American Economic Review (2004) / “Entrepreneurship”, Journal of Labor Economics (2005). The jack-of-all-trades theory; a model of entrepreneur = generalist, employee = specialist. 【Reliability: High (theory) / for the empirics see ⁵】 ↩︎ ↩︎²

5. The Jack-of-All-Trades Entrepreneur: Innate Talent or Acquired Skill? - Olmo Silva, Economics Letters (2007). Shows that once fixed effects remove unobserved individual traits, the correlation between balanced skills and entrepreneurship vanishes. 【Reliability: Medium-High】 ↩︎ ↩︎²

6. Workforce and Learning Trends 2023 - CompTIA (2023). Survey of US/UK HR and L&D professionals; 84% of responding companies have adopted some form of the T-shaped skill model. 【Reliability: Medium】 ↩︎

7. Job-Based HR Guidelines (ジョブ型人事指針) - Cabinet Secretariat, METI, and MHLW (August 2024). Describes traditional Japanese-style employment as one in which “autonomous career formation is hard to achieve.” 【Reliability: High】 ↩︎ ↩︎²

8. Report of the Study Group on Digital Talent Development for the Society 5.0 Era (skill-based talent development) - METI (May 2025). Three perspectives: skill visualization, hands-on learning, and capability certification. 【Reliability: High】 ↩︎

9. IT Skill Standards / Digital Skill Standards - Information-technology Promotion Agency (IPA). Organizes skills in a two-layer structure of foundational literacy and specialist deep-dive domains. 【Reliability: High】 ↩︎

10. DX Trends 2025 - Information-technology Promotion Agency (IPA) (2025). 85.1% of Japanese companies report a shortage of DX-driving talent — flat year over year. 【Reliability: High】 ↩︎ ↩︎²