The junior engineer is dead. Long live the junior engineer.
Every six months a VC declares the entry-level role over. They're wrong about the timing, the cause, and what comes next. From the inside of an actual hiring loop, here's the real picture.
Every six months, a venture capitalist publishes a viral blog post declaring the entry-level engineering role over. Three months later, the same firm posts a junior engineer opening. By my count, this has happened eight times in the last two years. The cycle is so reliable that I'm tempted to set up a Slack alert.
I've also hired a lot of junior engineers in this period. I run a small team at an early-stage company. I've reviewed roughly 400 resumes for our last two openings, and I've watched what the actual entry-level interview market looks like from the inside.
Here's the honest version. The junior engineer is not dead. The job is changing — meaningfully and visibly — but the death narrative is wrong in three different ways, and the people repeating it are mostly trying to sell you something. Including, sometimes, the model that wrote their viral post.
What changed for real
A junior engineer in 2022 spent a measurable percentage of their first six months on what I'll call "boilerplate work". CRUD endpoints. Form validation. The same React component everyone has written a hundred times. This was annoying for the junior — they hated it — and annoying for the senior — we hated reviewing it. Nobody loved this phase. We tolerated it because it was the cheapest way to teach someone the shape of the codebase.
That phase is mostly gone. A junior can now have an AI write the CRUD endpoint in 90 seconds, the form validation in 60 more, and the React component in another 90. Three hours of "learning" compressed to ten minutes.
What that means in practice:
| For juniors in 2026 | What changed | What didn't |
|---|---|---|
| Typing speed | Mostly irrelevant. The model types. | — |
| Boilerplate code | Disappeared from the day-one workload. | — |
| Code review | Now most of the work, on day one. | — |
| Reading unfamiliar code | — | Still the bottleneck skill. |
| Communicating clearly | — | Now visible from day one. |
| Debugging | — | Still the senior skill juniors need. |
| Asking the right question | — | Worth twice what it used to be. |
| Shipping | — | The whole game. |
The dispiriting part is that "what didn't change" is much shorter than "what did". The skills now most valuable in a junior — reading code carefully, communicating clearly, debugging like a detective, knowing when not to ship — were always supposed to be valuable. They were just less valuable when ten hours a week of someone's time went to typing.
Why the death narrative keeps coming back
Three reasons, in order of how much they irritate me.
One: it's a great viral hook. "X is dead" is one of the oldest tropes in business writing. It moves traffic. The author gets attention. The actual content of the post almost never sustains the claim.
Two: it's true if you squint. Some of what juniors did is automated now. If you define the job by the tasks rather than the trajectory, you can build a case that the job is hollowed out. You'd be wrong, but you'd be wrong in a way that gets engagement.
Three: it sells things. The same people declaring the junior dead are running bootcamps, selling courses on "AI-native engineering", or pitching products that promise to replace junior labor. Their stake in the narrative is not abstract.
What I actually look for now
When I read a junior resume today, here's what I'm scanning for, in order:
- Have you finished something. Anything. A real project that someone actually used. The size doesn't matter. The finished-ness does.
- Can you read code. I'll ask you to walk me through someone else's open-source project in the interview. I want to hear you notice things.
- Do you ship. Tell me about a time you got something into prod, what broke, and what you did. If you've never shipped, I can teach you. If you can't tell stories about shipping, I can't teach you taste.
- Can you talk to people. Can you ask for help in a way that doesn't waste a senior's time. Can you write a Slack message that doesn't make me re-read it three times.
I do not particularly care: what your GPA is, what college, how many leetcode problems you've ground through, or whether you can write a binary tree traversal on a whiteboard. None of these were good predictors before. They're now actively misleading, because AI can solve all of them and your prep regimen tells me more about your tolerance for tedium than your software taste.
The advice for somebody starting today
I get asked this a lot, so the short version.
Build something that someone other than you uses. Ship it. Watch it break. Fix it. Write about what you learned. Repeat. This is the same advice it has always been. The AI has not changed it. The AI has just made it easier to build the first version, which means the bar to "I shipped something real" is genuinely lower than at any time in the last twenty years.
If you cannot show me a project that someone other than your mom has used, in 2026, that is on you. The tools are free. The friction is gone. The bottleneck is taste and finish.
What the death narrative gets right (a little)
The narrative does get one thing partly right. The path is harder now in a specific way. The traditional "we'll hire ten juniors, half will wash out, the rest will become useful in two years" model is harder to make economic sense of when so much of the early work is automated. Companies are hiring fewer juniors, more carefully, and with higher expectations.
That isn't death. That's career compression. The runway for figuring out whether you like this work is shorter. The people who'll do well are the ones who use the AI's force-multiplier nature to grow up faster — to take on harder problems, sooner, with more support.
The role isn't going away. The role is just earlier. The cohort that gets hired in 2026 will be more independent, faster, and a little more nervous than the 2019 cohort. The good news is they have better tools. The honest news is they'll have to use them.
The cheap takeaway
Stop reading "X is dead" posts. They're written for engagement. They're not written for you.
Build something. Show it to people. Ship it. Notice what breaks. That advice has been correct for thirty years. The AI hasn't changed it. The AI has just made it slightly more obvious which advice was always the real one.