I'm not convinced that the CTO role has ever really had the right definition. Not least because it's different across orgs & industries and constantly evolves as the tech changes and orgs mature. If you see the CTO role as accountable for the technology operating model to meet business outcomes there is no need for AI to change that.
You still need strategic tech choices, architectures and manage risks around security, fraud etc. A CTO who codes all the time isn't really a CTO regardless of whether they're vibe coding or trad. coding.
It's the accountability that really matters. Sure everyone with an AI chat now thinks they're an expert, but those people existed before GPT.
What I find most interesting: the Microsoft and Carnegie Mellon study cited says higher confidence in AI is associated with less critical thinking. If one really sits with that for a moment, it means that this finding doesn’t validate going deeper. Instead, and more interestingly, it warns against the confidence that depth produces. The CTO who understands the substrate well enough to trust their own architectural judgment is exactly the profile the study flags as the risk. The evidence used to build this case cuts in the opposite direction from the conclusion being drawn.
The six habits, directionally, are not wrong. But they are incomplete without the one discipline that makes them worth anything: sustained critical thinking applied against tools explicitly designed to make us trust them, feel productive using them, and come back for more. Depth without that discipline doesn’t produce better judgment. It produces more sophisticated confidence in a system optimizing for our dependency.
The study is about confidence-during-use, not confidence-from-study, but you're right that I leaned on it as if those were the same thing and they aren't. Depth alone doesn't immunize anyone from the confidence trap. It probably makes the trap more elegant.
The discipline you're naming, sustained critical thinking against tools designed to feel trustworthy, is the missing layer. I'd argue the papers help build that discipline rather than erode it, but only if read with the posture you're describing. Read passively, they become another form of consumption. The substrate matters; the stance toward it matters more.
You are spot on with this article. One of the big personal shifts I had to do was to start rolling up my sleeves and getting into the “substrate” this last year. I was doing a lot of that, and then I realized, without sharing back, that learning is kind of a dead end. Because of how fast things are moving, you almost need to have a moment to stop, put down your thoughts on paper, and continue to accelerate.
Your blog has been a big inspiration to start mine, so thank you for this.
This piece prompted something I've been thinking about.
You name two patterns — the Token Monkey and the Vibe Coder. I think there are more. The depth-seeker who reads the papers, the evangelist who sells the vision, the governor who retreats into frameworks, the abstainer who waits it out. Six patterns, six different expressions of confidence — and each one has delegated the thinking somewhere.
The uncomfortable part is that "read the papers, learn the taxonomy, go deeper" — the prescription here — produces its own certainty. Earned certainty, yes. But certainty nonetheless. And certainty is where scrutiny stops.
I'm not convinced that the CTO role has ever really had the right definition. Not least because it's different across orgs & industries and constantly evolves as the tech changes and orgs mature. If you see the CTO role as accountable for the technology operating model to meet business outcomes there is no need for AI to change that.
You still need strategic tech choices, architectures and manage risks around security, fraud etc. A CTO who codes all the time isn't really a CTO regardless of whether they're vibe coding or trad. coding.
It's the accountability that really matters. Sure everyone with an AI chat now thinks they're an expert, but those people existed before GPT.
No they didn’t.
What I find most interesting: the Microsoft and Carnegie Mellon study cited says higher confidence in AI is associated with less critical thinking. If one really sits with that for a moment, it means that this finding doesn’t validate going deeper. Instead, and more interestingly, it warns against the confidence that depth produces. The CTO who understands the substrate well enough to trust their own architectural judgment is exactly the profile the study flags as the risk. The evidence used to build this case cuts in the opposite direction from the conclusion being drawn.
The six habits, directionally, are not wrong. But they are incomplete without the one discipline that makes them worth anything: sustained critical thinking applied against tools explicitly designed to make us trust them, feel productive using them, and come back for more. Depth without that discipline doesn’t produce better judgment. It produces more sophisticated confidence in a system optimizing for our dependency.
Josh, fair hit.
The study is about confidence-during-use, not confidence-from-study, but you're right that I leaned on it as if those were the same thing and they aren't. Depth alone doesn't immunize anyone from the confidence trap. It probably makes the trap more elegant.
The discipline you're naming, sustained critical thinking against tools designed to feel trustworthy, is the missing layer. I'd argue the papers help build that discipline rather than erode it, but only if read with the posture you're describing. Read passively, they become another form of consumption. The substrate matters; the stance toward it matters more.
Going to chew on this.
You are spot on with this article. One of the big personal shifts I had to do was to start rolling up my sleeves and getting into the “substrate” this last year. I was doing a lot of that, and then I realized, without sharing back, that learning is kind of a dead end. Because of how fast things are moving, you almost need to have a moment to stop, put down your thoughts on paper, and continue to accelerate.
Your blog has been a big inspiration to start mine, so thank you for this.
I relate very strongly to this and I plead guilty as well... we need to talk about it...
This piece prompted something I've been thinking about.
You name two patterns — the Token Monkey and the Vibe Coder. I think there are more. The depth-seeker who reads the papers, the evangelist who sells the vision, the governor who retreats into frameworks, the abstainer who waits it out. Six patterns, six different expressions of confidence — and each one has delegated the thinking somewhere.
The uncomfortable part is that "read the papers, learn the taxonomy, go deeper" — the prescription here — produces its own certainty. Earned certainty, yes. But certainty nonetheless. And certainty is where scrutiny stops.
I wrote about this at more length here: https://kinarey.com/the-confidence-trap/