Firing Minds to Hire Mobs
We stand on the shoulders of giants and then shut them up.
We cite Aristotle in conference talks and train juniors breadth-first. We name-drop Seneca and hand new developers Copilot. We revere the lineage and reject the lesson, which has been the same lesson since Athens: go deep before you go wide, or you'll never go deep at all.
Seneca, first century: "Nusquam est qui ubique est." Everywhere means nowhere. He compared shallow learning to food that leaves the stomach before it's digested. Linger with a few master thinkers. Go deep. Breadth without depth is tourism.
Plato prescribed ten years of mathematics before you touch dialectic; try to reason before you're ready and you get corrupted (Republic VII). Aristotle said you can't profit from advanced work without foundational experience (Nicomachean Ethics I.3). The medieval guilds enforced it: seven years as apprentice, then produce a masterpiece judged by masters, or you don't advance.
Dijkstra said it about software specifically in 1975: "The tools we use have a profound (and devious!) influence on our thinking habits, and, therefore, on our thinking abilities." He argued that teaching programming languages instead of teaching thinking was the foundational error of CS education. The field ignored him for fifty years.
Once you see that code is residue and reasoning is the scarce asset, the rest of the industry's behavior starts to look deranged. What I see now is a field of opportunity and horrors. The opportunity is later in this series. The horrors are here.
When digital computers arrived at Los Alamos, the human computers didn't get a clean transition. Nobody sorted them into "keepers" and "redundant." The war ended. Spouses relocated. Funding shifted. MANIAC I wasn't even operational until 1952; seven years of overlap where the old and the new coexisted and nobody made a plan.
What happened was erosion. Peter Lax had been paying attention to the physics underneath his calculations; he became one of the most important mathematicians of the twentieth century. Marjorie Devaney understood what the machines were actually doing; she spent forty years building Los Alamos's computing infrastructure. They didn't transition because someone promoted them. They transitioned because they'd always been doing more than cranking numbers.
The rest just stopped being needed. Not fired. Not replaced. The work moved on and they didn't move with it, because nobody taught them how and nobody thought to try. Lax survived because he'd been doing mind-work inside a mob-work job title.
No plan. No pipeline. Just a widening gap between the people who understood why and the people who only knew how.
We're watching the same split happen right now in software. Mind-work and mob-work aren't castes of people; they're modes of work. But the industry keeps optimizing for one and gutting the other.
This would be concerning as philosophy alone. Unfortunately, it's also measurable.
First: the junior-to-senior pipeline is eroding. Stanford's Digital Economy Lab tracked it in ADP payroll data; employment for software developers aged 22-25 dropped nearly 20% from its 2022 peak while workers over 30 grew 6-12%. A Harvard study found that AI-adopting firms reduced junior hiring by 7.7% within six quarters. SignalFire's 2025 talent report showed Big Tech's new-grad share of hires fell from 32% to 7% between 2019 and 2024. The juniors who should be learning why software is built the way it is aren't getting the chance. The apprenticeship is being skipped.
And here's the part that should embarrass us: the models those hiring managers trust to replace junior output? Stack Overflow's 2025 survey found developers only trust AI-generated code 29% of the time. Two-thirds report spending more time fixing "almost-right" AI output than it would have taken to write it. We're dismantling the pipeline based on a bet we're already losing.
Second: the seniors who are supposed to drive AI are learning the tools faster than they're learning the thinking. BairesDev's 2025 survey found two-thirds expect their roles to be redefined by AI; only a third have learned any AI-specific skills. They built careers on applied logic. They know how to build. Many of them have never had to articulate the first principles underneath their decisions. They've never traced their design instincts back to the novel sources — the domain knowledge, the user research, the academic work, the hard-won failures — that shaped those instincts. They know the right answer. They can't always explain why it's right; and "because I've seen this before" doesn't transfer to a model.
So we're firing the next generation of minds before they develop, and expecting the current generation to do work they haven't been trained for.
Every junior developer writing CRUD endpoints and gluing APIs together is doing mob-work right now. The machine does it cheaply enough to change hiring decisions. The question is whether anyone is teaching them mind-work before the transition finishes. Right now, the answer is mostly no. The pipeline that would teach them is the one we're dismantling.
The tool shapes the thinking; the thinking shrinks to fit the tool.
Every senior who can't explain why — who built on instinct and never traced it back to the source — is the next human computer. Every junior who never gets the chance to learn why is already gone.
The machine is the mob now.
Are you a Mind?