In the AI era, what should junior engineers really learn?
By Midi Yang, InsurPal CTO
As the CTO of an early-stage product startup, I work closely with new grads, and together with the team we've been thinking a lot about how AI is reshaping what "entry-level" engineering should look like.
Many large companies have frozen hiring for entry levels, implicitly assuming that junior engineers' skill sets can be replaced by AI. I don't believe that.
I don't think people are replaceable, nor is such a practice sustainable. I think people have been trained, and used, in the wrong way.
Thesis: Entry-level engineers should not be optimized for typing speed. They should be optimized for ownership, judgment, and end-to-end problem solving.
Traditionally, entry-level developers are focused almost entirely on implementation: writing code. With coding AI, that is no longer where most of the leverage is. But it's not because juniors can only do implementation, it's because we placed them there.
What we are experimenting with
On our team, entry-level engineers own two core deliverables. Paired with review cycles and low-friction access to senior expertise, this becomes far more accessible than before, and new grads improve very quickly.
Design docs
This is usually considered senior-level work. But seniority often comes from:
- Deep understanding of existing systems (AI can help accelerate this)
- Understanding trade-offs between design choices (AI can help surface options)
- Tight review cycles with quick senior feedback
The key is not that juniors become instant architects. The key is that they build architectural judgment early, with guardrails.
Implementation, largely supported by AI
AI accelerates execution, but the engineer still owns the intent, structure, and decisions behind the code.
It turns out that when new grads are trusted with real ownership, they step up and improve much faster than we tend to expect.
A split that may be historical, not natural
This made me question a long-standing assumption: seniors do architecture, juniors do implementation. That split may come less from how people naturally grow and more from how organizations optimized for older tools.
In an AI-first world, engineers can grow more holistically, with exposure to both design and implementation, and ownership of the problem end-to-end from the very beginning.
Why this matters for startups
As a resource-constrained startup, we don't embrace these new broad roles just because we're small. We do it because AI has made them possible. The bigger the company, the slower the shift, but the direction feels inevitable.
Open question
Curious how others are rethinking junior engineer growth in an AI-driven world.
Want to compare notes?
If you are experimenting with junior engineer training, design doc templates, or AI-assisted workflows, I would love to hear what is working.