The Broken Ladder

Early in my career I worked on a work order automation system. A senior developer started it. I came on to help, and over a few years I became the point of contact, and then the owner of this project.

Nobody handed it to me. There was no transition meeting, no document, no moment where ownership formally moved. It arrived because I had been there the whole time, and at some point I was simply the person who knew why the thing worked the way it did.

A lot of knowing that involved flying to warehouses.

We would sit through customer meetings and then go walk the floor. You watch how a picker actually moves through a rack system. You see the label printer that jams on the third pallet, the clipboard that exists because a screen ended up in the wrong place, the guy who has a workaround for a step nobody ever wrote down. Then you go back and look at the business rules, and half of them stop being arbitrary. On paper they look like somebody's preference. Standing in the building, they look inevitable.

I did not learn that system by reading it. I learned it by being present while it was built, and by seeing the thing it was modeling. The code was downstream of both.

Years later I could still tell you why a particular validation existed. The reason was not in the code. It was not in a ticket. It was in a warehouse in another state, in a conversation with a floor manager who explained why his crew did one step before another in a sequence that made no sense until it did.

What a Rung Actually Was

We talk about the career ladder as if it is made of titles. Junior, mid, senior, staff, principal. The industry sold that version hard, with timelines attached. Two years to mid. Five to senior. A promotion cycle you could plan a life around.

Those were the visible part. They were never the load-bearing part.

Every rung I actually climbed had the same shape. Work small enough that I could hold all of it in my head at once. A system I was present for while it got built. Proximity to the people whose problem it solved.

Nobody was teaching me. It was not a training program, and nobody in the company would have described it as development. It was a job that happened to also be an education, and the second function was never written down anywhere, which is why nobody noticed when it went away.

The ladder people are grieving right now is the visible one. The titles. The timeline. The promotion that was supposed to arrive at eighteen months. Those were scaffolding, and the industry sold the scaffolding as the building. The rungs were made of something else entirely, and they were made of it accidentally.

Where I Am Writing From

My current work is doubling down on AI. Agentic workflows, new projects, the whole direction. The legacy loan origination and servicing system I have spent the last year on is moving into maintenance while new systems get built to replace it. Eventually my team will start working with those new systems. That is the plan, and I do not have a complaint about the plan.

Warehouses then, lending now. Different industry, different decade, and the shape underneath has not changed at all.

I have not inherited anything yet. Nothing bad has happened to me. This is not a war story about a black box that landed on my desk.

I can already tell you what I am not going to know when those systems arrive, and I can tell you why, and the why has nothing to do with anyone hiding anything from me.

What Comes Through the Wall

Demos. Slack chatter. Daily meetings with product owners.

Take an honest inventory of what those carry. Demos carry finished behavior. This is what it does. Slack carries status. This is where it is. Product owners carry intent. This is who it serves and what problem it solves.

Now take an inventory of what none of them carry.

The approach that got abandoned in the second week. The argument about the data model, and who won it, and why. The constraint someone discovered halfway through that quietly reshaped everything built after it. The thing that was tried and did not work, which is almost always the most load-bearing knowledge in any system and almost never survives into documentation.

Every channel available to me is a what channel. The how channel is not restricted. It does not exist.

I do not know how that team works internally, and I am not going to guess. Maybe they have rich internal discussion I never see. That is entirely possible and it does not change my situation, because the question is not whether the reasoning happened. The question is whether it reaches the people who will own the result. On my side of the wall, it does not.

And being remote is not the cause. I have worked remote for years. Remote had pull requests you could read, design threads you could argue in, review comments where somebody explained why your approach would fall apart under load. Those were how channels and they worked fine across three time zones. What I have now is not a distance problem. It is a different problem wearing the same clothes.

The Speed Is the Mechanism

When a system takes eighteen months to build, being nearby is an education whether anyone intends it or not. You are present for the reversals. You watch a decision get made badly and then get made again. You absorb the reasoning in fragments, over lunches and in hallway arguments and in the six weeks where everyone is stuck on the same thing. Nobody plans this. It is a side effect of duration.

Compress that to weeks, and route a large share of the construction through a model responding to prompts, and there is no period left to be nearby for. The reversals happen in a session that closes. The abandoned approach never becomes a conversation because it never took long enough to be worth discussing. The reasoning is real, and it is fast, and it is gone.

The hype about AI was right about the speed. Genuinely right. I use these tools every day and I am not going to pretend the acceleration is imaginary.

The cost that never showed up on any invoice is that the build stopped being observable. Not secret. Observable. There is a difference, and the difference is the whole argument.

It Is Breaking At Both Ends

Almost all of the conversation about AI and careers points at the bottom of the ladder. Juniors not getting hired. The entry-level rung thinning out because the work that used to justify an entry-level salary is now the work the tools do best.

But it is one rung. What I am describing is not happening to juniors. It is happening to me, and it will happen to people with more experience than I have.

The first rung was small work you could understand end to end. The tools absorb that. Every rung after the first was proximity to construction. Speed absorbs that.

The ladder is not getting shorter. It is breaking along its length, and the bottom is just the only section anyone is measuring.

And nobody owns fixing it, because nobody ever owned building it. No company set out to create a formation pipeline. They set out to ship working software, and the formation happened as a byproduct of how the work was organized. When the organization of the work changes, the byproduct stops, and there is no department responsible for noticing. Every individual decision that removed a rung was correct on its own terms. The aggregate is nobody's job.

That is what makes this different from the usual complaint about the industry. There is no villain and no reversal coming. The rungs were free, and free things do not get rebuilt when they disappear, because no line item ever showed what they cost.

Naming It Before It Lands

Seneca wrote to a grieving woman that what wrecks us is not misfortune but misfortune we never considered could reach us. His line: he robs present ills of their power who has perceived their coming beforehand.

The Stoics called the practice premeditatio malorum, and it is regularly mistaken for pessimism. It is closer to specification work. You are not imagining disaster to feel bad in advance. You are converting a vague dread into a concrete list, because a list is something you can act on and dread is not.

So here is mine.

When those systems come to my team, I will not know what was tried and abandoned. I will not know which constraints are statutory and which are somebody's risk decision from four years ago that hardened into gospel. I will not know why the data model looks the way it does, which means my instinct about what is safe to change will be worth less than it has been worth for twenty years. I will not have stood in the warehouse.

None of that is a complaint, and none of it is a prediction that the systems will be bad. They may be excellent. Excellent and opaque are fully compatible, and that combination is going to be more common than either the fanboys or the doomers have accounted for.

What the list does is change the shape of the problem. If I know the how channel does not exist, I can stop waiting for it and start deciding what to build in its place, on my side, for my team. That is a much smaller question than what AI is doing to the profession, and it is the only version of the question I actually have any purchase on.

The ladder I climbed is not coming back. I am not sure it should. It was inefficient in ways I was too close to see, and some of that inefficiency deserved to die.

But it did something we never priced, and it did it for free, and the bill for that arrives late and gets paid by people who were not in the room when the decision was made.

Over the next several pieces I want to work through what that actually means. Who stops arriving, and what happens to a team when they do. What it costs to depend daily on a tool you do not trust. Where the value goes when everyone shipped fast and someone has to own the mess. And what you build for yourself when the structure that used to build you is not there.

The ladder was never yours. What you learn climbing without one is.

Thanks for reading my articles! If you enjoyed this article, please clap many times. It would mean a lot to me and help other people see the story.

Follow me on X!

You can also find me at The Stoic Coder! Hope to see you over there.