AI is disrupting career progression by eliminating the learning opportunities at each rung while simultaneously enabling people to perform above their experience level, Alasdair Allan explained in his talk Engineering Progression When AI Ate the Middle at QCon London. Fewer junior developers join the industry, and AI slows hiring at the entry level.
AI is writing a lot of code, and the landscape of what being a developer is has massively changed, Allan said. But writing code was never the point of our profession, he added.
Using AI requires supervision, and supervision requires coding skills. If AI handles the work that used to train engineers, Allan questions where the next generation of engineers will come from. AI productivity benefits may come at the cost of the skills necessary to validate AI-written code, if junior engineers’ skill development has been stunted by AI use, he said. The pattern recognition veteran coders carry, how systems should be structured, where complexity hides, and what breaks at scale, can no longer be acquired traditionally, Allan said. Junior engineers are not building up the needed intuition for what is and isn’t right about the code they’re looking at:
Junior engineers won’t spend years reading legacy codebases or debugging production incidents at 3 a.m. They’ll just point the AI agent at it and ask it to summarize.
Most engineering is what people call blackfield: legacy systems under high load, on a deprecation path everyone agrees on, but no one has time to execute, Allan said. The spec was never written, or was written and made irrelevant by decades of undocumented decisions. Business rules are encoded in conditions that outlived everyone who understood them.
AI agents can read the code, test the documentation, but what they can’t do is read production, Allan said. They can’t look at years of request patterns and understand which code paths are load-bearing in ways the code itself does not reveal.
The hiring of young workers has slowed in exposed occupations, Allan mentioned. There is no decrease in work for workers over 25. This means fewer junior developers join the industry, he explained:
People weren’t being sacked; they just weren’t being hired in the first place.
AI stunts skill formation, AI transforms work into supervision, and now AI slows hiring at the entry level.
Organizations racing to adopt AI coding tools are simultaneously degrading the pipeline that produces people capable of supervising those tools. The people who can build good contexts for AI are people who have context in their heads through years of craft. The pipeline that produced these people is breaking, Allan argued.
Like medical residencies, you do the scut work because it teaches you, not because it’s efficient, Allan said. Companies that do this will be the ones that have senior developers left in ten years, as the rest of us will have retired by then, he concluded.
InfoQ interviewed Alasdair Allan after his talk. InfoQ: What are the conclusions from studies of how AI impacts development time and quality of generated code?
Alasdair Allan: The research is more contradictory than either side wants to admit. METR’s randomised controlled trial found experienced developers were 19% slower with AI, while believing they were 20% faster. Anthropic found junior engineers using AI scored 17% lower on mastery without finishing any faster. They traded learning for nothing.AI is genuinely useful for short, well-defined tasks, but the gains collapse as complexity increases and the bottleneck migrates upstream to code review, exactly where senior judgment lives. Zero out of fifteen AI-generated PRs were mergeable on real open-source projects, despite passing automated tests. AI implements functionality. It fails at craft.
InfoQ: How does the work that computer programmers do change when they use AI tools?
Allan: Programming is becoming supervisory. Anthropic’s engineers use Claude in 59% of their daily work, but can only fully delegate around 20%. That gap is where the jobs will live in the future, between generation and judgment. The AI writes code; the human decides whether it’s the right code.What concerns me is that 80 to 90% of engineering questions now go to AI rather than colleagues. The incidental learning from struggling through a problem with a mentor gets bypassed when AI gives you the answer without the journey. The career pipeline was built for people who write code. We haven’t rebuilt it for people who supervise systems.
InfoQ: What are your suggestions for building engineers in the AI era?
Allan: I don’t have the full answer. We’re at the diagnosis stage, not the solution stage. But we can act now: structured learning paths with deliberate rotation through fundamentals, like medical residencies, because the scut work builds judgment. We need to measure understanding, not velocity. Watch how people think, not what they produce. Treat context as infrastructure, write documentation that reads like onboarding for a senior engineer who knows how to code but doesn’t know the codebase. We need to set honest expectations. AI is a tool, not a teacher. Using it to skip understanding is borrowing against your future.