Level 5 in the age of Altman and Musk: Does the humble CEO still make sense?

Collins called it Level 5 Leadership — the combination of "personal humility" and "professional will." He wrote it 25 years ago. Re-reading it in 2026, I think he observed the right thing but gave it the wrong name.

What Collins calls "humility" isn't really humility. It's integrity — and this isn't a semantic issue. The name determines whether Collins's theoretical frame still applies in an era where celebrity CEOs are the norm, not the exception.

This is post 2 in the series reading Good to Great with AI. Transparency warning (per the commitment in Post 0): I still haven't read the original book. The Collins summary below is based on AI, with risk of distortion — wherever it matters, I'll mark it for you to verify.

What Collins said

In the chapter on Level 5 Leadership, Collins ranks leaders across 5 levels, each building on the previous. Level 1 is a capable individual. Level 2 is a contributing team member. Level 3 is an effective manager. Level 4 is a visionary leader driving high standards. Level 5 — the highest — is an executive who builds enduring greatness through personal humility and professional will.

Collins's most controversial finding: in the 11 companies that went from "good" to "great" that his team studied, every single one had a Level 5 leader at the moment of transition. No exceptions.

The two faces of Level 5 per Collins:

Personal Humility:

• Avoids public praise, doesn't boast
• Directs ambition toward the company, not the self
• Prepares a successor who can surpass them
• When successful: looks out the window — credits team, luck, circumstances
• When failing: looks in the mirror — takes responsibility personally

Professional Will:

• Produces superior results, the catalyst for moving from good to great
• Iron determination to do what's necessary for company endurance
• Sets high standards, doesn't accept less
• When facing difficulty: doesn't blame, doesn't look for windows

Collins's canonical example: Darwin Smith of Kimberly-Clark — CEO for 20 years, wore cheap clothes, drove an old car, did farm chores on weekends. The man who decided to sell off the company's entire core paper-mill business to go all-in on consumer products in direct combat with P&G. Wall Street ridiculed. The press ridiculed. Smith didn't respond, just executed. Result: outperformed the S&P 4x over 20 years.

Colman Mockler of Gillette: refused three hostile takeover bids to protect the company long-term, even though each refusal cost him personally. Described as a "perfect gentleman," always made time for family.

Collins is blunt about an opposing observation: hiring outside celebrity CEOs is negatively correlated with good-to-great transitions. Of the 11 good-to-great companies, only 1 hired an outsider as CEO. The comparison group (companies that didn't make the leap to great) had a much higher rate of outsider star CEOs.

His argument: celebrity CEOs optimize for personal image, not durable company. They're good at creating "first peaks" but can't build a self-sustaining machine.

The problem with reading Collins in 2026

In 2026, the list of most influential CEOs in tech is almost entirely celebrities. Jensen Huang at Nvidia. Sam Altman at OpenAI. Dario Amodei at Anthropic. Elon Musk at Tesla/SpaceX/X. All public-facing, all with personal followings exceeding their companies, all on magazine covers and keynote stages monthly.

By Collins's standard, this is a warning sign. But the companies they lead are creating the largest economic value in industry history.

There are two ways to read this:

Reading 1: Collins is outdated. Context has changed. Celebrity CEOs work now because the AI era changes the incentive structure — talent recruitment needs a public founder, fundraising needs a narrative, regulatory engagement needs a face.

Reading 2: Collins is still right. We don't know the ending yet. 25 years after the book, we've seen Fannie Mae collapse, Circuit City go bankrupt, Wells Fargo's scandal. Collins warned about endurance, not short-term peaks. Today's celebrity CEOs may be at Collins's "first peak" — not yet at the durability test.

Both readings have merit. But I think both ask the wrong question. Because the question isn't "are celebrity CEOs okay" — it's "is humility really what Collins observed?"

Renaming: not humility, integrity

Reading Collins's descriptions of Level 5 CEOs carefully, there's a different pattern from "humility":

• Darwin Smith wore cheap clothes — not to appear humble, but because he genuinely didn't see expensive clothes as making him more valuable
• When asked about success: "I was lucky to work with talented people" — not a polite phrase, he actually believed it
• When the company failed: "that was my decision and it was wrong" — no one had to force him, he volunteered the admission first
• Joseph Cullman wrote his autobiography, dedicating the opening chapter to listing decisions of his own that he believed were wrong — while the company was at its peak

The common trait isn't "I am small" (humility). It's "I say what I see, even when it's bad for my image." That's integrity with truth — intellectual honesty.

I've written a blog post about two kinds of integrity in engineering (The Invisible Ceiling). One definition from that post applies here:

Integrity with the self is the ability to look at your real capabilities without distortion. It's daring to say "I don't know" when you don't. It's daring to acknowledge that the person next to you is better than you at something — and going to learn from them.

Applied to Level 5: what Collins observed in Darwin Smith wasn't that Smith thought he was small. It was that Smith didn't distort cause-and-effect about Kimberly-Clark's wins and losses. When the company won, he didn't fool himself into thinking it was because of him. When it lost, he didn't distort to blame elsewhere. That's integrity, not humility.

Why the name matters

If the underlying trait is humility, Collins's paradox is real: how can a CEO be both humble and iron-willed? This is where Collins gets stuck — he calls it a "compelling combination" but can't explain why two opposites coexist.

If the underlying trait is integrity, the paradox disappears. Both faces are natural consequences of the same thing:

• Integrity with truth → looks like humility. Because the truth is that no one single-handedly builds a company. People with integrity see this, say this. People without integrity distort toward self-credit.

• Integrity with commitment → looks like iron will. Because once you've promised yourself the company will succeed, you can't simultaneously promise and lie to yourself about progress. Iron will isn't a superpower — it's the consequence of not letting yourself self-deceive.

• Integrity with the team → looks like succession planning. Because people with integrity see their limits clearly. They know who can do role X better than them. They aren't threatened by that truth — they leverage it.

In other words: Collins observed two expressions. Both come from one cause. When he named the cause "humility plus will," he had to call it a paradox. When you name it "integrity," they're just two faces.

A test: two people both "humble"

The test for whether this reframe has value: find two people who both appear humble, one Level 5, one not. If the "humility" frame can't tell them apart but the "integrity" frame can, the reframe is worth something.

Person A: A CEO who publicly says "we have an amazing team, I'm just lucky to be here," but in internal meetings consistently takes all the credit, blames subordinates for failures, never admits any decision was wrong.

Person B: A CEO who openly says "there are two people on this team I respect more than myself at X and Y," admits in internal meetings that decision Q from last year was wrong for reasons A, B, C, and this time they'll do it differently.

By the "humility" frame: Person A looks more humble — because they say humble things in public. This frame can't distinguish who is Level 5.

By the "integrity" frame: Person B is Level 5. Because integrity isn't what you say, but whether your behavior is consistent with truth. Person A has a humble attitude but un-integrated behavior. Person B doesn't need a humble attitude — their behavior already has integrity.

This is the decisive test. Humility can be performed. Integrity leaves traces: decisions, explanations of failure, actual succession plans, response when contradicted.

Testing the framework on 4 AI-era CEOs

This is where Post 2 has practical meaning. Let's apply the reframe to 4 names defining the AI era — and see how the result differs from the "humility" frame.

Jensen Huang

By the "humility" frame: fail. Huang is a public figure, leather jacket has become a personal symbol, he appears on Nvidia keynote stages with full showmanship. Collins would warn.

By the "integrity" frame: closest to Level 5 among the four. Reasons:

• Huang openly acknowledges Nvidia almost went bankrupt several times in 30 years — doesn't hide, doesn't reframe as "we strategically pivoted"
• When talking about AI compute, he explains technical detail at a level engineers can verify — not sales talk
• In keynotes, he credits Nvidia engineers very specifically by name, no generic "our amazing team"
• When the market doubted (crypto crash 2022, dot-com style sentiment 2025), he didn't panic-respond with buybacks or hype — he explained directly why demand is structurally different

What Huang is showmanship-y about isn't who he is. It's about the product and technology. This is an important distinction. Steve Jobs was the same: showmanship about the product, not about himself.

Caveat — I need to verify: Does Huang have a succession plan? Currently unclear. This is a test Collins considered very important. If Huang doesn't prepare succession by age 70, that's a bad signal.

Dario Amodei

By the "humility" frame: near Level 5. Amodei is moderately public-facing, not showy, speaks academically.

By the "integrity" frame: near Level 5, for a different reason. Reasons:

• Founded Anthropic by leaving OpenAI on safety grounds — act of integrity with mission, despite personal and financial cost
• Public about AI risk timelines in ways that disadvantage short-term business (worries investors, attracts regulators)
• Anthropic's "Responsible Scaling Policy" and public commitments — verifiable, with cost if violated
• In interviews, willing to say "we don't know" about Claude's capabilities — rare in the industry

Both frames agree: Amodei might be Level 5. But the reasons differ.

The humility frame says: he doesn't boast, so he's okay. The integrity frame says: he's willing to say things that hurt business to stay consistent with mission. Reason #2 is the one that matters.

Sam Altman

By the "humility" frame: borderline. Altman is public but not showy in Musk's way.

By the "integrity" frame: borderline, for more specific reasons.

Pass:

• Public "I was wrong" after the OpenAI board incident
• Public "we were not transparent enough" about GPT-4 capabilities
• Speaks directly about AGI risk despite the regulatory attention it draws

Fail:

• Multiple accusations of saying different things to different stakeholders (Helen Toner case, Ilya Sutskever's departure)
• OpenAI board firing — the board's stated reason was "not consistently candid in his communications with the board"
• Equity and governance arrangements not transparent

This is where integrity must have an operational definition. Mine:

1. Consistency between speech and action
2. Willingness to admit being wrong when it has cost
3. Prioritizes truth when it harms the self

Altman passes (2) and partially (3). Open question on (1). If (1) doesn't pass, he isn't Level 5 — regardless of public mood.

The "humility" frame can't analyze at this depth. The "integrity" frame can.

Elon Musk

Musk is the case where I lack information to adjudicate. Early Tesla/SpaceX phase (betting personal capital when the company nearly went bankrupt, all-in on mission) is meaningfully different from the post-2020 phase (public behavior has shifted markedly). By the "humility" frame, clearly fails. By the "integrity" frame, depends on which phase you weight. I'll leave Musk as an open question — readers who follow him more closely than I do should assess for themselves.

Comparing the two frames

Applied to the four people, the two frames give different results:

The integrity frame isn't easier to pass. It's strict in the right places. It allows a showy CEO (Huang) to still be Level 5 but doesn't allow a CEO who is inconsistent with truth (Altman partial fail) to pass.

This is what the "humility" frame can't do. It conflates public style with the core thing that matters.

Why this reframe matters for Vietnamese founders

Vietnamese founders live in a culture that treats humility as virtue. "Humble" is among the highest forms of praise in Vietnamese. The downside: it easily becomes performance — done in order to be praised.

I've seen Vietnamese founders:

• Say "I still have so much to learn" while not actually learning from anyone
• Publicly say "we got lucky" while internally taking all the credit
• Be "humble" with customers and the press, but praise themselves without restraint to their team
• Decline public praise, but crave private praise obsessively

This is performed humility. It doesn't create Level 5 — it creates performative humility, one of the most toxic things in leadership.

Reframing to integrity solves this problem:

Integrity can't be performed. Because integrity is behavior, not attitude. You can say humble things while still stealing credit from your team — that's no integrity. You can be confident, even fierce in argument, but change your mind when the other side has better reasoning — that's integrity.

For Vietnamese founders, I think the practical message is:

1. Don't learn to say humble things. Learn to not distort when reporting progress, when explaining failures, when assigning success.

2. Public presence isn't ego. In the global AI era, silence equals nonexistence. But public presence must be paired with integrity — not the hype cycle.

3. Most importantly: integrity with the self first. If you can't look at your real capabilities without distortion, you can't have integrity with your team. These are in order. The first must come first.

The last point matters because it's the real ceiling of leadership. A founder who can't look at the truth about themselves will always lead the company away from truth. That company can be "good" — never "great."

Back to the original question

Is a humble CEO still relevant in an era when every top tech CEO is a celebrity?

My answer: wrong question. A CEO with integrity is always relevant — independent of context. What changed between 1965 and 2026 is how many CEOs can look not-humble while still having integrity. In 1965, humility and integrity often overlapped (the context forced them to). In 2026, they can decouple — you can be public, showy, evangelist, and still have integrity with the core truth.

Collins isn't wrong. He named it suboptimally. With renaming, his framework becomes applicable to the AI era — and strict in the right places.

Methodology notes

As committed in Post 0, here's the transparency about how this post was made.

Actual process:

I didn't read Good to Great before writing this post. I followed the planned 5-step method:

1. Ask AI to summarize the Level 5 chapter
2. Brain dump before deep dialogue — this is where the reframe "not humility, integrity" emerged
3. Critique with AI — AI pushed back on my motivated reasoning, I considered and partially held my position
4. Distill into blog — this post
5. Audio later

Where AI pushed back on me and I held my position:

AI warned I might have motivated reasoning — because I'd just written a 12-minute blog post on integrity, I might be "seeing integrity everywhere." AI tested in reverse: if Collins were right and humility (not integrity) is the underlying trait, would I accept it?

I considered this. I kept the reframe for an operational reason: humility doesn't have a reliable behavioral signature, integrity does. A CEO can have a humble attitude but un-integrated behavior — and vice versa. I think the behavioral test matters more than the attitude test. I might be biased — readers should remain skeptical.

Where I trusted AI without verifying:

• Specific quotes from Level 5 CEOs (Darwin Smith, Joseph Cullman) — I paraphrased AI summaries, not verbatim from the book
• The "10/11 internal CEO" ratio — AI said it, I haven't verified
• The Joseph Cullman autobiography story — specific details could be off
• The list of 11 good-to-great companies and their outcomes — verifiable but I haven't done it

Where I'm verified or confident:

• The 5-level frame and "personal humility + professional will" — core concept, widely cited
• "Window and Mirror" metaphor — the most famous quote from the chapter
• Darwin Smith selling Kimberly-Clark's paper mills — historical fact, easily verifiable

AI bias warning:

AI may be deferential to famous CEOs because they're heavily mentioned in training data. I pushed AI to find contrarian cases (Adam Neumann, Travis Kalanick, Stanley Gault) to balance. But the analysis of Huang and Amodei may still tilt positive because of how much positive material exists about them in training data.

If you've read Good to Great in the original and see me (or AI) misunderstanding, please correct. This is what honest reading looks like in the AI era — not pretending to have read it carefully.

Post 3 next: First Who, Then What — Collins says choose people first, then choose what. The AI era poses the question: what if "people" can also be AI agents?