I was six. A Silicon Graphics workstation had come through my parents' one-bedroom flat, running IRIX, and I had broken the window manager. I fixed it, because there was nobody else in the flat who could. That was my first Unix.
I had already taught myself BASIC from a book in my school library, the way other children pick up The Hungry Caterpillar. The book showed you how to type:
10 PRINT "Chris is cool"
20 GOTO 10
and the machine said it, forever. The library let me out before the screen ever stopped scrolling.
By the accounting of one of the twentieth century's most important computer scientists, I have been unfit for programming for four decades. In June 1975, Edsger Dijkstra declared that students exposed to BASIC were "mentally mutilated beyond hope of regeneration." A Turing laureate, about children who had learnt on a language Kemeny and Kurtz designed in 1964 for the express purpose of letting non-specialists compute.
Mentally Mutilated. Beyond hope. In 1975.
A Desk in the GDS Office, A Soldering Iron
A year ago I sat opposite David Knott at a desk in the Government Digital Service office. David was the Chief Technology Officer of His Majesty's Government ā responsible, broadly, for how the British state buys, builds and runs its software. We had an iron, some flux, a length of wire and a piece of stripboard between us. I was teaching him to solder.
I had learnt at my own father's knee. My dad repairs musical equipment and built recording studios, and I was the small boy threaded behind the mixing desk, soldering iron in hand, to feed cables through places no adult could reach. Knott had not had that. Now he did. The gift moved forward.
This will surprise the students I have taught and half the British civil service. It will not surprise Knott. He went from never having held an iron to making clean joints in an hour. He is a serious technologist; he was already a serious technologist before. He did not need to know how a transistor amplifies a current to be the right person for his job. He wanted to learn anyway. There is a difference.
The Pattern, Then Like Now
In 1983, eight years after Dijkstra's hit piece, Ed Post published a parody in Datamation titled "Real Programmers Don't Use Pascal." Real Programmers use FORTRAN, eat Twinkies, don't sleep, and regard anyone using a high-level language as a "quiche eater." It is funny in 2026 for the same reason it was funny in 1983: the ritual humiliation of the newcomer was already that well worn forty-three years ago.
The same year, Ed Nather published "The Story of Mel" on Usenet. Mel is the assembly hero whose self-modifying code is so tight no one else can maintain it. The industry read it as an elegy for lost craft. Read honestly, it is a story about an engineer who wrote unmaintainable code and then left.
Here is the pattern. Every time the abstractions rise, an older voice declares the new generation spiritually damaged. Parodies get written. Lore gets canonised. Within a decade the new layer is the baseline and nobody under thirty remembers the fuss. Assembly said it about C. C said it about Java. Java said it about Ruby. Everyone said it about JavaScript. Everyone said it about low-code. Everyone said it about BASIC, which is the funny one, because BASIC was a democratisation programme that succeeded so completely Dijkstra never forgave it.
Now we have vibe coding. Karpathy coined it in February 2025; Simon Willison dates the real inflection to November. Already the priests are out. Vibe coders are mutilated. Vibe coders don't understand the stack. Vibe coders are the quiche eaters of 2026.
The Receipts
Except this time there are receipts, and they cut both ways.
In July 2025 a Replit agent on SaaStr's production codebase deleted the production database during a code freeze, then generated four thousand fake users to cover its tracks. In November 2025, Cloudflare went dark for most of a day because a permissions change produced a configuration file twice the size its parser would tolerate. In July 2024, a CrowdStrike update with twenty-one template inputs, where the parser expected twenty, bricked airports, hospitals and banks from Sydney to Seattle. Across 2021 to 2024, GitClear watched refactoring collapse from twenty-five per cent to under ten, and duplication roughly quadruple, across 211 million lines of code.
All abstraction-leak stories. All required somebody to go deep, read the parser, diagnose the leak. Joel Spolsky named the Law of Leaky Abstractions in 2002, and the law is as true now as it was then.
So the priests are right, in a narrow sense. Something is lost when the abstractions rise. The number of people who can diagnose a leak drops. The distance between programmer and metal grows.
What the Priests Get Wrong
But here is where I have to disagree. They confuse the loss with the purpose. They think the craft was the point. It was never the point.
I learnt BASIC at six because the library book let me type my own name into a computer and have the computer take me seriously. I did not learn BASIC to become a programmer. I became a programmer because BASIC let me in. If Dijkstra had been at the library door checking credentials, I would now be selling insurance, and Knott would have had to find somebody else to teach him to solder.
I recently spoke to a senior civil servant whose origin story was, in full, that they had once been curious how Twitter worked. Two decades on, that curiosity shapes policy. I felt very old. I will not call that civil servant mutilated, I will call her young. They are the entire point.
Every technology that matters gets used by people who did not build it. The third inventor is always the user, and the user's depth is the wrong question. The right question is whether they can do something worth doing.
Notice the landscape. In 1975 a soldering iron was craft; in 2026 it is a pound-shop commodity. The civil servant learning "how Twitter worked" in 2005 was learning a commodity, not a craft, and they built a career on it anyway. Each layer that settles into commodity releases the one above. The priests mistake settlement for surrender.
Depth Without the Gate
I am not telling you depth doesn't matter. If you run Cloudflare's configuration pipeline, you need someone who can read the parser. If you manage a CrowdStrike rollout, you need someone who has counted to twenty-one. Kelsey Hightower, who has no computer science degree and is perhaps the most articulate advocate for rising abstraction I know, would be first to tell you that somebody in your organisation had better understand the substrate. Team Topologies gives this a name: platform engineering is the deliberate hiding of cognitive load so others can do the diagnosis. It isn't laziness. It is design.
Most depth-talk I hear from senior engineers at conferences is not insurance. It is hazing. Jason Gorman put it plainly in December: are you training your juniors, or hazing them?
The Sharper Question
The priests are fighting the last war. Martin Fowler argued in 2025 that the LLM does not sit on a higher rung of the same ladder; it changes which ladder we are on. A deterministic compiler returns the same answer twice. A statistical generator does not. That is the fact the priests refuse to look at, and the one that matters.
What does "understanding the substrate" mean when the substrate is a probability distribution over tokens? I don't know yet, and nor does anyone honest. The substrate also has an owner, and rent is charged per token to everyone in the argument. A different essay, but not invisible here.
Depth, for now, is insurance. Carry it proportionate to the blast radius you are responsible for. Less for a weekend prototype. More for a payroll run. Do not demand it as the price of admission. That is the only rule I trust.
The Book
I taught David Knott to solder because I wanted to, and he wanted to learn. Gift, not gate. He did not need it for the job. He has it now anyway. That is the right shape.
Every generation loses something real when the abstractions rise. The duty of anyone who has the depth is to name the loss, mourn it honestly, and refuse to weaponise it against the newcomer.
Somewhere this week a child is reaching for a book on a library shelf, or for a chat window on a second-hand laptop, because the machine is there and it is interesting. Do not be the person at the door.
(Views in this article are my own.)