Where are the junior engineers?

Over the last two years, the industry has pulled back from hiring juniors. You might be thinking, “Oh Sapan, how can you think that? We’re hiring X and Y number of junior engineers.” So are we, and I’ll come to that later.

The reality is that arithmetic doesn’t lie, and a lot of the numbers are worse than people guess.

We are about to hire 12 to 15 junior engineers, roughly what we hired last year, maybe even a few more.

I want to be fair to the people who stopped, because their logic isn’t incorrect. If one engineer with AI does the work that used to take three, why pay for someone who slows your seniors down for a year first?

Put that way, cutting junior hiring sounds responsible. But let me tell you why I think it’s a mistake. And after hearing me out, if you land on an answer different from mine, rest assured that’s absolutely fine.

The gap you’re picturing is really three gaps

When people say a junior can’t do what a senior does, they treat the distance between them as one big thing. It’s three things stacked together, and this is my own read, not a law of nature. Most of that distance has shrunk a lot over the last couple of years. The part that’s left is still standing strong like a rogue Bruce Banner on a rainy night, and no one is worried about it.

Gap 01 · Discipline

The first thing AI is solving is discipline

Most people assume the junior is the one who cuts corners. We’ve found it’s often the other way around. We put our freshers through test-driven development properly, from day one. They then worked next to senior hires who had never really done TDD and didn’t believe in it, and the freshers did it better.

That’s not a knock on senior engineers. It is simply much easier to build a habit in someone who is still forming how they work than to change someone who has done it a certain way for a decade. A junior hasn’t formed those habits yet, which is exactly what you want when you’re the one doing the forming.

There’s a line from Naval Ravikant I like.

On learning

“The smartest people are all self-taught, even if they went to school.”

That is really what you’re hiring for in a junior, curiosity and the appetite to learn. Get that right, give them good habits early, and the discipline gap closes faster with a junior than with a senior who is set in their ways.

Gap 02 · Knowledge

Second is the knowledge gap

The knowledge gap was real and expensive for a long time. It’s the one AI has shrunk the most. Knowing something used to mean being able to write it. Today you mostly need to know a thing exists and what good looks like when you see it. You don’t hand-write the migration anymore. You ask for it, you read it, and you judge it.

So the job we’re training juniors for has changed. We’re teaching them to be good delegators by knowing enough architecture to tell whether the shape they got back is right, describing what they want in the language that gets it built, and catching the thing that looks fine but isn’t.

You have to know it. You just don’t have to type it.

This is where our craftlets come in. A few times a week we run one-hour sessions on architecture and design concepts, and most of them are about awareness, which is what helps juniors the most. You can’t judge a design whose shape you’ve never seen, so we build that recognition early.

There’s also something people miss when they count years on a résumé. Experience doesn’t add up smoothly, year after year. It jumps, and before you know it the inexperienced fresher is an experienced professional handling complex problems.

Once you see that, you stop waiting five years for a junior to become useful and start closing the gap on purpose.

Gap 03 · Pragmatism

Pragmatism is the one gap that stays

The third gap is real, and we haven’t closed it yet. Pragmatism is the judgment call. Take two more weeks or ship what we have now, depending on what breaks if you’re wrong. When to bend a rule and when to hold the line. The honest answer in software is almost always “it depends,” and you can’t teach that on day one, or day five, or with a training module. It comes from having been burned, and getting burned takes time.

So this is the gap you staff around, and it comes down to your ratio. The senior’s judgment rubs off the only way it ever has, by sitting shoulder to shoulder on real problems. The learning that used to happen in code review happens now in the pairing and the mob. That only works if the ratio is right.

Put 12 juniors around one burned-out senior and you’ll prove every skeptic correct.

A growing company is a training company

The math never misses, except one thing. The engineers who will run your systems in three years, the ones who stay calm when production is on fire at 2AM, are the juniors you train now. There’s no other supply of them. You can’t post a job for a senior who already knows your systems and carries your scars and have someone show up. You grow them, or you go without them at the moment you need them most.

A growing company has always been a training company. AI didn’t change that. What it changed is which gap you spend your training budget on, and it happened to shrink the two that used to cost the most.

Sandro Mancuso has a line I’ve repeated since 2020.

On experience

“Repeating one year of experience ten times is not the same as ten years of experience.”

The opportunity in front of us is the reverse of that. A junior with the right habits, the right training, and a good senior beside them covers ground far faster than you ever thought they could.

So we’ll do what we did last year. We’ll hire 12 to 15 juniors and train them the way we always have.