As expectations accelerate regarding the ability of AI to write code entirely, a deeper issue emerges concerning how human talent evolves within technical organizations. In a recent presentation at the QCon AI Boston conference, Alasdair Allan presented a critical perspective on the gap left by AI in the professional development path of engineers, describing it as "missing rungs on the career ladder."
Between Ambitious Expectations and Digital Reality
Allan pointed to the clear discrepancy between optimistic expectations and actual data; while Dario Amodei, co-founder of Anthropic, predicted that AI would write 90% of code within a year, analyses have shown a different reality. According to research conducted by Redwood Research, the percentage of code written by AI in software repositories is closer to 50%, a figure that only reaches 90% if we include temporary text and transient scripts.
Figures recorded at major companies reveal a varied landscape:
- Google: Reports indicate that just over 25% of internal code was generated by AI.
- Microsoft: The intelligent generation rate is around 30%.
- GitHub Copilot: The acceptance rate of suggestions in enterprise environments is around 30%.
- Anthropic: The company's engineers use the Claude tool in 59% of their daily work, with reports indicating a 50% increase in productivity.
The Supervision Paradox: The Hidden Threat
The core of the problem raised by Allan does not lie in productivity, but in what he called the "Supervision Paradox." While AI tools increase the speed of task completion, they may deprive junior engineers of the opportunity to build the foundational skills that usually come from solving complex programming problems manually.
When AI takes over tasks that represented the "intermediate rungs" on the learning ladder, the junior engineer finds themselves able to produce quick outputs, but they lack the cognitive depth necessary to understand how the system works as a whole. This leads to a knowledge gap that makes it difficult for them in the future to take on leadership or senior engineering roles that require supervising autonomous systems.
What Does This Mean for the Future of Software Engineering?
This shift imposes strategic challenges on technical companies, summarized in three main axes:
- Erosion of Institutional Expertise: Excessive reliance on ready-made tools may lead to the loss of the deep technical context that engineers build through direct interaction with code complexities.
- Threat to the Technical Supply Chain: If juniors cannot develop their skills through intermediate stages, companies will face a shortage of expert engineers capable of managing and guiding complex AI systems.
- Redefining the Engineer's Role: The goal of the profession is no longer "writing code" per se, but the ability to design systems and supervise the quality of outputs, which requires a solid technical foundation that AI cannot provide to the learner.