A New Seniority: The Ladder and the Landscape

The question everyone is asking

If companies stop hiring junior developers, where do the senior developers come from in the future? It's a fair question, and it's been everywhere this week. I want to suggest it may be the wrong one.

Not because the concern is silly. Big companies really are hiring fewer juniors, and the ladder they used to climb really is missing its lower rungs. The problem is what the question assumes: that seniority is the thing at the top of that particular ladder, and that the only way to make more of it is to rebuild the rungs. I think the ladder was never the point. It was a byproduct of the tools we had, and the tools have changed.

The ladder was built by the toolchain

Think about how the ladder actually worked. Year one, you did the grind: pulled the reports, wrote the first drafts, fixed the build. Year three, someone more experienced tore your work apart and you fixed it, and then did it again. Year five, people started handing you the problems nobody knew how to solve. Somewhere after that you noticed you could tell why something broke before anyone else did, and that was seniority.

We tell that story as if the slowness were the ingredient. As if judgment were a sediment that only settles if you leave it undisturbed for a decade. But look at the mechanism underneath: attempt, consequence, explanation, retry. Nothing in that loop requires five years. What made it take five years was the toolchain. The feedback came from one manager, with the bandwidth of one person's attention and the scope of one person's experience, at a latency of days or weeks. The grind was slow because the tools were slow. The correction loop was narrow because the only correctors available were the people standing next to you.

So the ladder wasn't a theory of how skill is made. It was the shape skill-building took when the loop was that slow and that narrow. Climbing a ladder of competence by repeating mistakes until you stop making them is one way to get good at something. It is not obviously the best way, and it is certainly not the only one. It happened to be the dominant loop the old workplace made available.

Columns and gaps

Here is the image I keep coming back to. Picture every discipline, every role, every tool as a vertical column. The ladder measures how high up your column you've climbed. Junior is near the bottom, senior is near the top, and "5+ years" is a rough proxy for altitude.

That was a reasonable measure when the columns were short enough that someone standing on top of one could see into the next. It isn't anymore. Specialization made the columns taller. The explosion of platforms, frameworks, roles and tools made them more numerous and pushed them further apart. The person at the top of a tall column can see less of the whole than a mid-height person could in the past. Height stopped correlating with the thing we actually wanted from seniors, which was never really height. It was the ability to see why something broke before anyone else did, and that is a matter of sightlines across gaps.

I hear this every time a business lead calls an employee "dumb" or a salesperson calls a tool "smart." They're making the same mistake from opposite sides: grading the columns. The failure they're actually living with almost always sits in a gap. Between what the employee was told and what the system assumes. Between what the tool does and what the workflow around it needs. Between two people who each did their job correctly and produced something broken together. Nobody owns the gaps, so nobody gets blamed for them, and the blame lands on whichever column is standing nearest.

It's not that the individuals are wrong, or ineffective, or any of the uglier words. It's that the columns have grown too tall and the spaces between them too wide. The people who are valuable now are the ones who can work in the spaces.

The consequences didn't go away

The strongest version of the worry is this: judgment comes from consequences. You build the thing, it fails, and someone shows you why. If AI does the building, the junior skips the consequences along with the keystrokes, and never develops the instinct.

I think this gets the direction exactly backwards. What AI can remove is much of the grind. I don't have to make the bricks any more and it turns out, thankfully, that the ability to make bricks is not the skill that produces an architect. The grind was never where judgment lived. What AI compresses is the correction loop, which is exactly where judgment lives. The consequences still arrive. They arrive in minutes instead of weeks, and they arrive from across a dozen domains instead of one, because the tool doesn't care which column you were hired into. Someone who ships twenty things in a year with these tools meets more consequences, in more places, than someone who shipped two by hand. While it's still true that you can ship 20 bad things and never learn from the experience, AI increases the opportunity for high-frequency learning if you're actively interrogating the results. the loop didn't vanish. It tightened and it widened at the same time, for people who make the effort.

I can say this with some confidence because I'm living it. I spent twenty years as a developer the first time around. I've learned more, and built more, in the past year than in those two decades combined. Not because I'm smarter today than I was back then, but because the loop between trying something, seeing it fail, understanding why and trying again shrank dramatically, and because it stopped being confined to the one column I happened to be standing in. The twenty years weren't wasted; they built a column. But what's making the difference now is curiosity, range and perspective. The ladder can't measure those things because the years didn't produce them. I brought them with me.

That's the part that convinces me this isn't a story about AI at all. The things that made me capable of being a frontier full-stack developer in 1999 are the same things driving my scope and velocity today. They were in my toolkit long before I typed my first 10 PRINT "HELLO". The tools changed completely, twice. The thing applying them didn't.

I don't believe that the ten-thousand-hours idea was ever a universal law of expertise. It makes sense for skills that demand a finely tuned nervous system and precise hands, where the body itself has to be trained in a narrow groove. But much of what we call knowledge work is different. It is judgment, and judgment is made of loops... iteration, feedback, and introspection, and the willingness and ability to engage with them, not merely the number of hours or years it took to do so.

What seniority looks like now

Seniority has historically used time and vertical depth as proxies for judgment. AI fundamentally changes the economics of acquiring experience while increasing the value of horizontal awareness. The old model of seniority gave us a ladder. The new one may look more like a landscape. Climbing the rungs may no longer reliably indicate the capabilities we actually want and need.

If seniority is the ability to connect the spaces between things, it has a signatue that is not necessarily measured in years. Someone who asks what the system upstream assumes before they build anything. Someone who can hold the tool's model of the problem and the human's model at the same time and notice where they disagree. Someone who treats "that's not my area" as the beginning of a question rather than the end of one. Breadth of awareness, scope of curiosity, flexibility of mind, and the willingness to pick up a tool they've never seen and find out what it's for by using it.

Those traits are visible in a two-year person. They're also exactly what makes someone good alongside AI tools, because an AI tool is the ultimate tall column: deep, fast, and completely blind to what's on either side of it. The person worth having next to it is the one standing in the gap.

If seniority looks vertical, allow me to suggest there's another dimension to consider, one horizontal to time and the accumulation of experience through trial and error. Call it "horizontal seniority" if you must give it a name. What it measures isn't how long you've worked in this column, but how many columns you can see, how well you understand their interfaces, and whether you can recognize what happens in the gaps.

New questions, new answers

Here's deliberate oversimplification: if you wanted to fix your city's traffic over the next five years, would you hire a senior traffic engineer who has spent thirty years timing stoplights, or a hobbyist with five years of hands-on work in robotics and mesh networking? In a world where the cars talk to each other and negotiate speed, spacing and right of way among themselves, the stoplight is on its way out, and thirty years spent coordinating their timing from a central location is thirty years of depth in a column that's about to be demolished. I'd take the second, with one condition. They have to be curious enough to ask why the lights were timed the way they were before they change anything, because there are absolutely things in the world which the robot cars do not yet see and which nobody has documented. Depth still matters. It just stopped being the only way to measure.

Requiring senior depth in one column, and lamenting its impending demise as a way of knowing who is more capable, is a rule born of a specific set of circumstances. The "5+ years experience" filter lets the current players select for the things they themselves invested in, without noticing that the rules of the game, and the capabilities that define the next generation of players, may have changed.

A solution in disguise

When I hear someone say "Where will the next senior engineers come from?," I hear a question asked from inside a very specific picture of what skill looks like. The ladder made tall specialists, and occasionally, almost by accident, it made someone who had looked into the gaps for long enough to become a connector. We called those people senior and credited the years. The years were never the cause.

The people this moment needs may already be out there. They may not look like seniors. They may have three years and six columns instead of fifteen years and one. If your hiring filter can't see them, that isn't a skills pipeline problem, it's a problem with the filter. New tools and new tasks require new ways of thinking, and for the first time in a long time, perhaps ever, we're being forced to name what we were actually looking for all along. That doesn't look like a crisis to me. It looks like a solution in disguise.