The fact that we're slowly removing the apprenticeship layer and passing it off as "productivity gains" should be far more alarming.
Just imagine. The kind of work that is best suited for AI, such as boilerplate code, simple CRUD endpoints, and basic component wiring, is actually similar to the work that helps in training junior developers. It was not that we were working on those tasks because they were difficult, but rather because they are the tasks that build our understanding of how software works.
The pipeline problem
Every senior developer alive today was once a junior who learned by doing boring work. You wired up forms. You wrote repetitive tests. You copy-pasted patterns until they became intuition.
Today, companies perceive equivalent work and ask βAI is able to do this quickly, why recruit a junior? Job ads reveal the reality: relative to senior posts, the number of junior developer positions has decreased by 2024 and has continued to decline to 2025. The access ramp gets steeper as we act as if the road is in good condition.
We're eating the seed corn π½
What concerns me the most is that the industry is focusing on short-term gains, without realizing that it is actually leading to a long-term crisis.
β Fewer juniors hired today means fewer mid-levels in 3 years
β Fewer mid-levels means fewer seniors in 6 years
β The "talent shortage" everyone complains about? We're actively manufacturing the next one
Companies report hiring fewer juniors because AI handles their traditional workload. Online dev communities are full of people debating whether entry-level frontend roles even exist anymore. These aren't hypotheticals. This is happening now.
"Just learn AI tools" isn't an answer
A recurring statement I come across is that "Juniors should just learn to use AI and they'll be fine." Okay. But could you be more specific about what they should use AI for?
If you have never constructed a mental model regarding how state is propagated in an application, asking an LLM to make one for you doesn't help you learn. You are a passenger, not a driver. You can't debug what you don't know. You can't architect what you have not crafted from scratch.
Using AI to assist in development is great for seniors because of course, the mental models are already there. They know when the AI is wrong. A junior using AI without the context of experience is just writing code they can't conceptualize. That's not learning. That's copying homework. π
What would actually help
β Intentional apprenticeship programs β companies investing in juniors despite the short-term cost, because they understand the pipeline
β AI as training wheels, not replacement β let juniors write it first, then compare with AI output as a learning tool
β Senior devs mentoring deliberately β code review matters more now, not less
This issue cannot be solved by the market alone. Markets are designed to perform best in the short term. The junior pipeline requires a ten-year commitment that ultimately results in institutional expertise, preservation of the organizational culture, and the presence of individuals who truly comprehend your codebase.
The uncomfortable truth
We got ours. We had the luxury of learning by doing tedious work in environments that tolerated our inexperience. Pulling that ladder up behind us β even unintentionally β is a choice with consequences.
Each time we throw a "we replaced two junior roles with Copilot" party, we are also dancing on the grave of a hypothetical future senior developer. π
So Iβd love to know: What are you doing if youβre in a position of hiring or decision making to ensure the junior pipeline is running? Or are we all crossing our fingers someone else figures this out?
Top comments (0)