The argument is straightforward, and it deserves to be stated plainly: if the industry produces no juniors today, it will have no seniors tomorrow.
You cannot skip an entire tier of the engineering pipeline and expect competence to materialise from elsewhere. Senior engineers are not conjured — they are the product of years of learning, error, correction, and accumulated context. Remove the years, remove the engineers. At present, senior engineers are carrying a disproportionate load. They are reviewing code, designing systems, and managing the consequences of AI-generated suggestions that frequently produce confident-sounding output and structurally questionable results. The industry responds to this situation by encouraging them to move faster. The models producing the questionable output are, increasingly, trained on previous AI output. The feedback loop is not encouraging. There is a further problem that rarely gets addressed directly. A language model's working context has practical limits. A human engineer who has spent months inside a large codebase carries an understanding of its history, its failure modes, its architectural decisions and their reasons — an understanding that cannot be reconstructed from a context window. That knowledge lives in the engineer. When the engineer is gone, it is gone. If the industry continues overloading its remaining experienced engineers while systematically refusing to develop new ones, the trajectory is foreseeable: slower systems, compounding technical debt, and a generation of developers left to maintain and debug codebases that nobody around them fully understands — because the people who understood them were never replaced.
The short-term cost saving has a long-term invoice. It will arrive!