Alasdair Allan's QCon London talk lands at a strange intersection: AI isn't just changing how we write code, it's rewriting the career ladder itself. His observation that AI "eats the middle" of engineering progression while simultaneously letting people perform above their experience level is more than a clever phrase. It's a structural shift that should worry anyone who cares about the long-term health of the profession. If junior developers are being hired less, and the learning opportunities at each rung are vanishing, then we're not just flattening a curve. We're hollowing out the pipeline that produces senior talent in the first place.
There's a tempting counterargument here, one that says "good, let AI handle the drudgery, and humans can focus on creative problem-solving." That's fine in theory, but Allan's point cuts deeper. The mundane tasks aren't just chores; they're the scaffolding for intuition. You learn to debug by debugging. You learn to design systems by breaking them. When AI removes those early reps, you get engineers who can produce impressive output without understanding why it works. That's not empowerment; it's dependency. It's the difference between knowing how to drive a car and knowing how to change the oil. Both get you to the store, but only one prepares you for the breakdown on the highway.
This is where the conversation gets practical for our readers. If you're a manager, the temptation to lean on AI for speed is real, but you're also the steward of your team's development. If you're a junior, the advice to "just use AI to learn faster" misses the point. You need to ask harder questions: not just "what does this code do?" but "why does this design pattern exist?" The related pieces we've run on Talking to My AI Clone Taught Me to Question the Tech and Navigating AI/ML Job Requirements: A Shift in Expected Skills reinforce this. One shows the danger of trusting an AI's output without skepticism. The other highlights how job postings now demand a hybrid of skills that didn't exist a few years ago. Both point to the same truth: the tools are changing faster than our mental models for how people learn and grow.
So what's our honest take? Allan is right to sound the alarm, but he's not being alarmist enough about the second-order effects. The real risk isn't that junior devs lose jobs to AI. It's that the ones who do get hired are being trained in an environment where their first few years of struggle, the ones that build resilience and judgment, have been replaced by autocomplete. We'd tell any reader who asks: don't wait for the industry to fix this. Take ownership of your own progression by deliberately seeking out the hard problems that AI can't solve for you. And if you're leading a team, build a culture where asking "why" is still a required skill. Because the next generation of engineers won't be judged on what they can build with AI. They'll be judged on whether they can fix it when it breaks. That's the learning opportunity we can't afford to automate away. Watch for the first wave of senior engineers who've never had to debug a memory leak without a copilot. That's the moment we'll know if we got this right.