What Have We Actually Made?

With Bruce Lloyd in London and Frank in Cleebourg.

The Curious AI Chronicles — conversations with Bruce about humans, machines and the increasingly strange territory in between.

Bruce and I were supposed to be talking about artificial intelligence.

We did, eventually.

First we had to get the audio working. Then Bruce told me that he had tried to install the Pineapple app and it hadn’t worked. He had followed the instructions, which is usually a sensible thing to do, except that modern technology has developed the unfortunate habit of regarding instructions as more of a general suggestion.

Bruce made the point that the instructions had probably been written by somebody determined to be thorough rather than somebody hoping to make life simple.

There is a difference.

It was a suitably messy beginning to a conversation which eventually became about The Pineapple. That happens quite often in our chats. We start somewhere sensible, notice something at the side of the road and, twenty minutes later, find ourselves somewhere completely different.

The journey may be the point.

Bruce had previously asked an AI to look at The Pineapple. I do the same thing rather obsessively. ChatGPT gets it. Claude gets it. Grok gets it. DeepSeek and Kimi have their turn. I usually ask them to criticise rather than admire it, which can occasionally become uncomfortable.

Bruce’s version had produced a kinder observation. It suggested that The Pineapple was about paying attention to real life, listening to people whose experiences differ from our own, questioning what we take for granted and turning conversation into better questions rather than rushing towards answers.

Bruce thought something was missing.

“The journey is the reward,” he said.

That stopped me, because AI is very good at shortcuts. Give it a problem and it will try to get you to the answer. That is partly why we use it. But sometimes the interesting bit is everything that happens before the answer arrives: the wrong turn, the irritation, the thing discovered at 2.17 in the morning, the solution which creates another problem, or the sudden recognition that something built six months ago is now preventing something entirely different from working.

The Pineapple has become a rather good example. Issue One and Issue Thirty belong to the same publication, but the machinery underneath them barely belongs to the same species. I have learnt an extraordinary amount in between, mostly because things keep going wrong.

The latest problem sounded harmless. We now have the Pineapple app and a public archive of earlier issues. Excellent. Except the articles for the next issue also live in that archive, which means that if they are public, people can read next week’s Pineapple before we publish it.

So we hide them.

Also excellent.

Except the software which assembles the magazine can no longer see them.

The obvious solution is to let the book-building system access the hidden articles. Unfortunately, the book builder involves Python. The articles live inside WordPress. WordPress sits on a server. Permissions are involved. URLs behave differently depending on whether an article is public, private, pending or draft.

I know just enough about these things to be dangerous and not enough to be relaxed.

So the current workaround is magnificently primitive. Before building the magazine, I make the articles public. The machine finds them and assembles the magazine. Then I hide them again.

This is not the digital revolution I was promised.

Bruce’s reaction was entirely reasonable. Surely, he said, there should be a timer.

Yes.

There should.

There should also probably be somebody I can telephone. Instead, there is me, a server, some Python code and ChatGPT, which sounds like the beginning of a joke and occasionally feels like one.

The machinery nobody sees.

A few weeks earlier, another change had killed public access to the Pineapple archive altogether. ChatGPT and I spent about five hours trying to find the problem. Eventually I stopped typing explanations and simply took screenshots.

Screenshot.

Answer.

Try something.

Next screenshot.

Another answer.

Try something else.

At some point the conversation became so long that the thread itself ran out of space and we had to continue in another one. Eventually we found the culprit: a design decision I had made months before. It had worked perfectly until another change exposed the weakness.

Once we found it, the solution wasn’t terribly dramatic.

Finding it was.

Much of The Pineapple has developed like this. The original book-building code was perhaps a few hundred lines. It is now somewhere around two thousand. Bruce asked how much of it I had written.

None.

I cannot code in Python.

I describe what I want. AI writes something. We test it. It breaks. We work out why. We change it. Something else breaks. Eventually it works.

Until Thursday.

The absurd part is that all this complication exists in order to make the finished result feel simple. Open The Pineapple. Read something. Listen to something. Skip something. Smile at a cartoon. Perhaps recognise somebody. Close it again.

The reader should not have to care about Python, server architecture, databases, permissions or the difference between a private URL and a public one. That is our problem.

Perhaps most simple things are like that. The supermarket cashier scanning a packet of pasta is sitting on top of a remarkably complicated system. The waiter carrying two coffees is connected to suppliers, refrigeration, accounting, staffing, tax, card payments and an espresso machine which will choose 07.42 on Saturday morning to require emotional support.

The machinery disappears when it works.

Which leads to the more interesting question.

Why build it in the first place?

What happens after the conversation?

Bruce eventually asked what The Pineapple was actually for.

Fair question.

Brida began with conversation. People sit around a table and bring their experiences, opinions, jobs, families, irritations, mistakes and small bits of knowledge gathered over a lifetime. English happens to be the shared language.

Then the meeting ends. Everybody closes the screen.

And what?

That question bothered me long before The Pineapple became a magazine. A good conversation seemed too valuable simply to vanish because somebody pressed Leave Meeting.

So I started making a record.

Not a transcript. Nobody needs sixty pages of “umm”, “yes”, “sorry, you go first” and somebody discovering that they have been muted for four minutes. I wanted something that caught the experience without pretending that the original conversation had been a polished essay. Sanitised where necessary. Reshaped. Written properly.

Those records eventually became The Pineapple.

Then something changed.

The Pineapple stopped being simply the place where conversations were stored. The conversations began travelling. Someone in Portugal could encounter an experience from Lithuania. Someone who had never met Bruce could read one of our conversations. A member could return to something discussed two weeks earlier and see it differently. An observation in one meeting could start a completely different conversation somewhere else.

People do not have to become friends. They do not have to follow one another. They do not need to exchange photographs of lunch.

Their experience can travel without them.

That interests me enormously.

Proof of life.

Bruce still wasn’t entirely satisfied. Was The Pineapple really connected to the commercial part of Brida, or had I accidentally built myself an extremely elaborate intellectual hobby?

Another fair question.

I said it served two purposes. The first was the record: people who had taken part could see something lasting come out of the conversation.

The second was harder to describe.

I called it proof of life.

Perhaps proof of concept. Perhaps simply proof that Brida exists.

Brida is difficult to explain because the moment I start explaining it, it begins sounding like something else. A language school. A networking group. A discussion club. A magazine. A community.

None of those descriptions is completely wrong.

None quite gets it either.

The Pineapple comes closer because instead of explaining what happens when people bring their lives to a conversation, we can show what came out of those conversations. Read it. Listen to it. Have a wander around. If what you find interests you, perhaps one day you will sit at the table too.

Then Bruce asked what percentage of Pineapple articles had originated in conversations involving paying clients.

I thought about it.

Probably around eighty per cent.

There was a short silence.

That changed the question.

The magazine is not sitting beside the real activity as a piece of marketing added afterwards. It grows out of the activity itself. The conversation produces The Pineapple. The Pineapple carries the conversation further. And somebody who encounters The Pineapple may eventually arrive at another conversation.

There is a circle there.

I am only beginning to understand it.

The human model.

Earlier, Bruce had suggested two familiar routes for growth. One is to build something repeatable and scale it. The other is to keep pushing the frontier through innovation.

I suggested a third.

The human model.

I am not sure the Harvard Business Review will be calling for the paper just yet, but I know what I mean.

On the surface, Brida is deliberately messy. People interrupt. They hesitate. Somebody misunderstands something. Someone changes the subject. A story appears that nobody expected. A person who has barely spoken suddenly says the most interesting thing of the evening.

That messiness has value.

Underneath it, however, the system is becoming increasingly structured. AI helps build questions and process material afterwards. Software assembles the magazine. Audio is generated. The archive is maintained. The app distributes it. Databases help us find people who might bring something interesting to the table.

Technology does more and more of the machinery so that the machinery itself can recede.

The human part remains the point.

We are not using humans to feed the machine. We are using machines to give human experience a longer life.

At least, that is the intention.

We will undoubtedly break something along the way.

Something slightly bigger.

In an earlier note about The Pineapple, Bruce suggested that it might be an experiment in what human learning could look like in a world where answers are becoming abundant but attention, judgement, wisdom and conversation are becoming relatively scarce.

That is an awfully large job to give a small magazine called The Pineapple, so I shall resist putting it on the masthead.

But there is something in it.

We have spent centuries building systems to distribute information. AI has just made information astonishingly cheap. Ask a question and an answer appears. Ask again and another appears. Ask for ten alternatives and they arrive before the kettle has boiled.

Perhaps the scarce thing is no longer the answer.

Perhaps it is having an experience worth talking about. Listening properly to somebody else’s. Noticing the difference. Changing your mind. Or simply asking a better question afterwards.

Those things remain stubbornly human.

Which brings me back to The Pineapple.

It started with a very modest irritation. A conversation happened and then disappeared. I didn’t like that, so we made a record. The record became a magazine. The magazine became an archive. The archive began talking. Then we built an app.

That broke something.

We fixed it.

Something else broke.

And underneath all those increasingly ridiculous layers of technology sits the same little idea that started the whole thing.

A conversation between a few people might contain something worth carrying forward.

Everything else — including two thousand lines of Python — came afterwards.