Every tool that was going to end programming

The first page

My first line of HTML was written by hand, in a text editor, into a page laid out with <table>. That is how you positioned anything. You drew an invisible grid and you put the text in the boxes.

There were no databases behind those pages and there were no content management systems. If the telephone number at the bottom of the site changed, you opened every file and you changed it in every file. If you missed one, it was wrong until somebody noticed.

I was writing it on a Bondi Blue iMac, the original one, in 1998. My son was born in 1995, and three years later that machine arrived and I stopped being a marketing media designer who wanted things built and started building them myself. I have never owned a computer that was not a Mac since.

I was also the first in my group of friends with a mobile phone, which was the size of a brick, and the first with a Nintendo. But I did not come to this because I loved computers. I came to it because I wanted to make things and the person who controlled the making kept saying no.

1996 and 1997, the first ending

Then the tools arrived.

CSS became an official standard on 17 December 1996, and suddenly you did not have to fake a layout with tables. A year later, in December 1997, Dreamweaver arrived, and you could draw a page on a screen while the program wrote the HTML underneath for you.

The talk at the time was that hand coding was finished. Why would anybody type angle brackets when a machine could generate them faster and without typing errors.

What actually happened is that a different kind of work appeared next to it. Somebody still had to make the generated page behave, and clean it up, and make it work in browsers that disagreed with each other about everything.

2003, the second ending

WordPress was released on 27 May 2003, and this one was bigger, because it did not take work away from programmers. It took a whole job away.

Suddenly a person with no technical skill at all could publish to the internet without asking anybody. No file transfer. No editing forty pages. No waiting.

And no webmaster.

I have written about that person before. He is the reason I learned any of this, because for years the answer to every idea I had was no. The first job the web genuinely deleted was the job of the man who told me I was not allowed to build things. I am not going to pretend that has never given me any satisfaction.

And then the work came back wearing different clothes

Here is the part that always gets left out of the obituary.

WordPress did not delete the work. It moved it. Within a few years there was an entire economy that had not existed before. Themes. Plugins. Extensions. Child themes. Translations. Hosting built specifically for it. People who had never heard the word framework were now commissioning, buying and customising, and every one of those transactions was somebody's job.

I made WordPress themes for a living, for years. That job did not exist in 2002. It existed because WordPress arrived and because enough people needed their site not to look like everybody else's.

I also worked with the company behind it for a while, and I was part of the translation group, so that people could use it in Spanish, which is my own language, in their own words rather than in somebody's approximation of them.

None of that was on anybody's list in 2003. The list said: the webmaster is finished. The list was right. The list was also tiny, next to what actually happened.

That is the part the predictions always miss. They are good at counting the jobs that go. They have never once been able to count the jobs that do not have a name yet.

The predictions are good at counting what ends. They have never once counted what had no name yet.

The 2010s, the third ending

Then the visual builders. Drag a box, drop an image, publish. No code at all, and this time they said so in the marketing, which is where the phrase no code comes from.

The same promise, for the third time, in the same words. And again, a new kind of work appeared beside it, because somebody has to decide what the thing is for, and somebody has to get it off the builder when it outgrows one.

What actually happened, every single time

Nobody went back. That is the first thing and it is worth saying clearly. Not one person who stopped editing forty files by hand ever asked to do it again. The drudgery went, it stayed gone, and that was a straightforward good.

Second, after each of these there were more people building software, not fewer. The tools lowered the cost of making something, and when the cost of making something falls, the world does not make the same amount of it more cheaply. It makes far more of it.

And third, the level moved. When layout stopped being manual, the interesting problem became interaction. When publishing stopped being manual, the interesting problem became systems. Each tool ate the floor, and everybody standing on it moved up a storey.

The question almost nobody is asking

So the useful question is not which jobs this one ends. We have asked that four times now and we have been wrong four times.

The useful question is what this one makes possible that was not possible before.

I do not have the whole answer and neither does anybody else yet. But I have mine, because I am the case study.

After fifteen years of a cancer that has not finished with me, I decided last year to apply this to my own life, as the subject, to get answers I had not been able to get any other way. Everything this company now builds came out of that. Governance, security, wellbeing. I did not sit down and plan a product line. It started with me, and the product is what I found.

That is one person. Multiply it.

What is genuinely different this time

I am not going to pretend this round is the same, because I do not think it is, and the honest version is more useful than the comfortable one.

Each of the earlier tools removed one specific task. Layout. Publishing. Deployment. This one is not task shaped. It reaches into the middle of the work.

And there is a difference I do not hear many people say out loud. This is not a chat. I know that is what it is called. For me it is an environment, and the nearest thing I can compare it to is the first time a real development environment arrived and everything was suddenly inside one window. People were astonished by that. This is larger.

But when I look carefully at what it has actually taken from my own days, it is still not the part I expected. Debugging and writing are two completely different activities, and what I have handed over is the searching. Reading four hundred lines looking for one wrong character is gone. Remembering the exact name of a function I last used in 2019 is gone. It took the looking. It did not take the deciding.

On a good day the two of us make extraordinary things, and they are fast and they are secure. And I still have to catch the mistakes, because the tool is not perfect, and anybody telling you otherwise is selling you something.

I have also paid for this round personally in a way the others never cost me. I cannot sit and type without watching my hands any more, because one of them does not work. So when I say these tools are an extension of my hands, I am being literal, and I am not being sentimental about what it cost to get here.

Who actually gets the most out of it

Last year I was called into a meeting I had not asked for, to present my own application to a twenty year old who had been introduced as the AI expert. Her qualification was that she typed into ChatGPT.

I have nothing against her. She was doing exactly what she had been told to do. The failure was the decision that put her in that chair, and that decision is being made in a great many companies right now.

Using a tool is not expertise. What I see over and over is that the people getting real leverage out of this are the ones who already understand how a computer behaves. Software engineers, IT people, engineers of any discipline, and also people who are simply very good with productivity tools and have a feel for how systems break. They know what to ask for, and more importantly they know when the answer is wrong.

My own example. For years I was frightened of the terminal. Properly frightened. Use the button instead, do not touch the black window. Then somebody told me to stop being afraid and just do it.

Now everything on this machine goes through the terminal, three or four of them open at once, and I have never once pressed a commit button inside an editor. I do not have to work that way. But I can, and knowing that I can is exactly what makes these tools worth something in my hands.

About the fear, honestly

I understand the fear. I share some of it.

This is one of the most powerful things we have built, and power does not arrive in a polite version. The same capability that lets one person build a secure product alone also lets somebody build something harmful. That is true and pretending otherwise would be dishonest.

But it is not new and it is not special to this. We could not stop the internet being used for the worst things people do. We cannot stop anyone using an encrypted messenger. An ordinary household contains a dozen objects that could hurt somebody, and we did not ban the kitchen. The question has never been whether a tool can be misused. It is who is accountable, and what they choose to put around it.

Which is exactly where I have watched things change. A year ago I was working inside one of the AI builders when it was still mostly developers building for developers, and I had to construct infrastructure inside it that simply did not exist. Today that is closed and reasonably secured. Not perfect. Not useless either. Still not where I would put a product that carries my company's name. But the direction is right, and that direction came from people deciding to be responsible. It did not arrive on its own.

And while we are being honest. I rebuilt my own infrastructure over the past week. It took good deployments and bad ones, reviews, rewrites, and a great deal of thinking. So when somebody posts that they built their first application in five minutes, we both know what that sentence is leaving out.

What has not changed at all

You still have to know what you are asking for.

That is the whole thing, and it is the link back to Wednesday. If English is now the interface, then a vague sentence produces vague software, and a person who cannot describe a system precisely will not get a precise system, no matter how good the tool is. The skill did not disappear. It moved into the sentence.

Which is why the people getting the most out of this are not the ones who never learned to build. They are the ones who did, and who now describe in English what they used to type by hand.

What I actually got back

Do I miss being the mysterious person who could read code at three in the morning. No. Nobody ever wanted to be at that desk at three in the morning. That is nostalgia, and nostalgia is a poor reason to keep doing anything.

Here is what I got instead, and none of it is abstract for me.

I cannot write by hand any more. My dominant hand has been injured for nearly two years and completely paralysed for one of them, and I am hoping this is the beginning of the year it heals. These words were spoken, not typed.

So I can leave this computer. The work carries on, my phone tells me if something breaks, and in the morning I am told what happened while I was asleep. I go to bed instead of staying up until four. I walk my dog early, before the street fills up. I sleep, which when your body is doing what mine is doing is not a luxury. It is treatment.

That is the trade. Not a faster way to type. A working life that fits around an illness and a hand that does not work, and that lets me be the engineer I could not be before.

What this means if you are deciding for a company

The pattern across thirty years is consistent enough to plan around.

The organisations that got the most out of each of these shifts were the ones that moved their people up rather than out. The work that disappeared was never the expensive work. It was the repetitive work that had been quietly costing them their best people's attention.

The ones that suffered were the ones who bought the tool, changed nothing else, and then concluded the tool did not work. The tool worked. Nobody had decided what it was for.

And nobody should be handed that decision because they are fluent in a chat window. Ask who in your organisation can tell when the answer is wrong. That person is your AI expert, whatever their job title currently says.

None of this is a technology question and it never has been. It is a question about what your people are actually for, and it is the question we spend most of our time on.

Sources

The World Wide Web Consortium Issues Cascading Style Sheets Recommendation, W3C. CSS level 1 became a W3C Recommendation on 17 December 1996.

Adobe Dreamweaver. Created by Macromedia and first released in December 1997.

WordPress, our journey. The first version, WordPress 0.7, was released on 27 May 2003.

The iMac G3 family tree, 512 Pixels. The Bondi Blue iMac was released in August 1998.