There are senior engineers pushing back on AI with real skepticism and real frustration.
These engineers get it. They've watched every hype cycle claim transformative results using sloppy metrics, seen Goodhart's Law eat entire organizations[1], watched teams credit a new tool for gains that really came from finally fixing their deployment pipeline, and seen a new flashier technology get the attention and investment when "boring" technologies are already doing the job.
And then there's the axe porridge problem. In the folktale, a traveller convinces villagers he can make porridge from an old axe head -- he just needs a little water, some oats, maybe some butter... AI adoption gives teams the cover to make changes they'd wanted for years. Better test suites, cleaner CI, smaller PRs. AI is the excuse, and the fixes were overdue anyway. The improvements are real, but people rightly rebel against the attribution. This was the most robust objection to my recent post about early interview observations with teams adopting AI.
The irony is the grizzled senior engineer encountering the next hype cycle is itself history repeating itself. Early in my career I was replacing Mainframe code with the hype at the time, Java. Mainframes were and still are an incredibly mature and versatile technology[2]. The pushback was warranted. But delivery of new features was at a standstill. Three decades on, Java remains deeply embedded in organizations and I suspect the vast majority of the Mainframe code is still around to boot. Both sides were half right, and nobody remembers the debate.
In one exchange an engineer compared the AI hype to the Kubernetes hype. It's an apt and recent comparison. Kubernetes was useful technology that got wildly misapplied. The question is whether AI is more Kubernetes or more Internet.
The skeptics say Kubernetes. I think Internet. And every hype cycle argues that it's different, so I do not pretend to have an airtight rebuttal. But the distinction matters: Kubernetes changed how we run software, while the Internet changed who could build it and how value moved through organizations. AI looks a lot more like the latter.
My workflow today bears little resemblance to how I worked even a few months ago. It's unrecognisable from 12-months ago. I describe what I want, review what comes back, and steer. My default mode is parallel exploration rather than sequential construction. The cost of trying something dropped so far that I try three approaches before I'd have previously finished planning one. That rate of change in how I work -- not just the tools I use, but the shape of the work itself -- is not something I experienced with Kubernetes, or containers, or cloud, or agile. None of those changed the shape of my day the way this has.
And it's not just changing how I deliver software. AI is in the marketing team's weekly standup too, at Internet scale and at a pace that's making it hard to adapt.
Skepticism is warranted here. People are building faster horses all over the place. Social media presents a vibe-coded-a-million-bucks view that is distorted and easy to dismiss. I have zero doubt there will be some sort of correction.
The scale of the pushback to a hype cycle is proportional to the level of hype. And the hype is insane. This only amplifies the core "we've seen it all before and it's never true." But opting out is not zero cost either. For all the flashiness, every hype cycle washes out to some new, inherent value. Ironically, I think the skeptics -- with their hard-won instinct for separating signal from noise, for spotting axe porridge -- are the most likely to unlock the real benefit, if they engage.
It's not just an intellectual debate. People's careers are being upended. Software engineers are being made redundant, and that threat only seems to be growing. There are people with conscientious objections to AI getting mandates they have to use it to keep their job. That reality deserves more than a dismissive "get on board." For many, it's about livelihood, identity, and decades of craft devalued overnight. That deserves compassion and respect.
In some circles there is an unsubtle subtext of "you'll come crawling back." And to some extent, they're right, all hype cycles correct. We will solve one set of problems and create a bunch of new ones. The pattern repeats. The size of this one is the open question.
Both the skeptics and the enthusiasts are running on faith right now, just pointed in different directions. I've picked mine, and I might be wrong.