We stopped hiring juniors. We're eating the seed corn.

I already argued that the bottom rung is gone — that AI ate the grunt-work tasks juniors used to cut their teeth on. This is the sequel, and it's the part that should scare the people making the cuts: when you stop hiring juniors, you're not saving money. You're eating the seed corn.

The cut is real, and it's rational — locally

The numbers aren't subtle. Entry-level developer postings are down something like 40% or more from their 2022 peak, and the junior share of hiring has roughly halved. Employment of developers aged 22–25 fell about 20% from the 2022 peak by mid-2025. And on a spreadsheet, the decision makes perfect sense: a mid-level engineer with a good agent now does what used to take a senior plus two juniors. Why pay for the juniors?

Because that spreadsheet only has this year's row on it.

Seniors are grown, not found

Here's the thing every "we don't need juniors anymore" plan quietly assumes: that senior engineers will keep materializing from somewhere. They won't. A senior is just a junior who spent five years making mistakes on real systems with someone watching. There is no other supply. Cut the junior pipeline and you're not trimming headcount — you're defunding the only process that produces the judgment you're about to depend on more than ever.

We've run this experiment before. After 2008, a lot of firms stopped hiring and training entry-level staff; a few years later they hit a wall of missing mid-levels that no amount of budget could instantly fill, because you can't buy experience that no one spent the years creating. Some companies see it this year — IBM reportedly tripled its junior intake in early 2026, restructuring the work rather than deleting the rung.

You can't skip the junior years and still have seniors in five. The ladder only works if someone's on the bottom of it.

AI makes this worse, not better — and here's the cruel twist

The usual response is: "AI will just make everyone senior faster." Backwards. Judgment — knowing which of the model's ten plausible answers is the right one, when to stop, what smells wrong — is built by doing the work AI now does for you. Take away the reps and you get people who can prompt fluently and can't tell when the output is subtly, expensively wrong. AI raises the value of taste and judgment while quietly removing the on-ramp to acquiring them. That's the trap: knowing what to build is the scarce skill, and it's exactly the skill the missing junior years were supposed to teach.

What to actually do

• Keep the rung, change the work. Juniors shouldn't do the tasks AI does now — they should do the next tier up, under supervision: reviewing AI output, owning small real features, learning to judge.
• Treat mentorship as production, not charity. Transmitting judgment is the highest-leverage thing your seniors do this decade. Staff it like it matters.
• Hire for the pipeline, not just the quarter. The team that keeps growing juniors through the drought will have seniors when everyone else is bidding for a shortage they manufactured.

The bottom line

Firing the juniors looks like efficiency and it's actually consumption — you're spending a future supply of senior judgment to lower this year's cost. The rung feels optional right up until you need a senior and realize nobody grew one.

Don't delete the bottom of the ladder. Move the juniors up a step and keep the farm running — because seniors don't spawn, and AI just made the thing they know how to do the most valuable thing there is.