Forever Junior: The Skills AI Can't Develop for You

Software engineering is shifting into an AI-assisted era, but that shift risks turning new engineers into "forever juniors" if they let agents do the learning for them. The author confronted that risk by asking what AI actually changes about the job and what concretely makes someone a Senior. Observation and practice revealed Seniors aren’t just coders: they defend shipped code, argue design trade-offs, detect when an agent is wrong, delegate appropriately, and teach others. With a promotion to Senior as a concrete milestone, the author mapped the human skills needed and began deliberately training them rather than outsourcing learning to agents.

Practical practices follow: force Socratic questioning and understanding checks (explain in your own words, predict failure modes, defend trade-offs), cultivate curiosity (ask agents to justify and cite), and treat reviewing as a core craft by emulating team patterns. A reusable workflow - discover, align, anchor, plan, implement - helps reuse existing code and justify deviations. Technical constraints in a personal skill.md (read-only exploration allowed; mutations require explicit confirmation) preserve opportunities to practice autonomy. Occasionally doing the work without an agent builds judgment. These habits turn agent-assisted drafts into training data for real growth, making the goal to write the reviewing rules rather than rely on them.