AI ate my learning curve
If you want great seniors, you first have to produce good juniors. And producing good juniors in the AI era means thinking carefully about what AI should accelerate, and what juniors still need to learn the hard way.
Details
AI can be a huge productivity multiplier, but much of that multiplier depends on the person using it.
For seniors, a prompt can amplify years of experience. What looks like “minimal checking” is actually built on mental models formed through failed deployments, strange bugs, bad decisions, and long nights debugging systems that refused to behave.
Juniors do not have that yet.
AI can make them appear productive while quietly bypassing the experience they need to become independent. As long as the agent stays on track, everything looks fine. But once it starts drifting, building on its own assumptions, or obsessing over the wrong problem, that productivity can collapse very quickly.
At that point, the junior may be left with a large codebase they cannot properly explain, debug, or even hand over to a senior: they know what they asked the AI to build, but not necessarily what it actually built.
AI can be a learning multiplier, but it can also hide missing understanding behind apparently working software. So where is the line between useful acceleration and outsourcing the very skills juniors are supposed to develop?
This talk explores that tension, and what it means for juniors, seniors, and teams trying to grow developers in an increasingly AI-assisted world.