The ladder is missing rungs – Engineering Progression When AI Ate the Middle

7 hours ago

AI is automating many coding tasks traditionally done by junior engineers, which were crucial for building judgment and intuition.
Despite predictions, AI-generated code comprises around 30-50% of commits in some companies, not 90%, but its impact is real and productivity gains are reported.
Studies show AI can stunt skill formation, with AI-assisted junior engineers learning less and scoring lower on debugging and conceptual tasks.
Experienced developers may work slower with AI in familiar codebases, but they increasingly depend on it, refusing to work without assistance.
AI success rates are high for short tasks but drop sharply for complex tasks requiring senior-level judgment, highlighting a supervision paradox.
Incidents like Amazon's outages illustrate risks when AI lacks context and judgment, relying on outdated or incomplete information.
AI adoption is shifting engineering work towards supervision, review, and architecture, with less net-new coding by humans.
Labor market data indicates reduced hiring of young workers in AI-exposed occupations like programming, threatening the pipeline for future engineers.
Solutions include structured learning paths, measuring understanding over velocity, and treating context as infrastructure to preserve institutional knowledge.
The long-term challenge is building engineers capable of supervising AI systems without the traditional rungs of career development.

Hasty Briefsbeta