The Toilet Roll, the Admiral, and the Age of AI

History has always worked this way. We just weren't paying attention.

In 1543, Copernicus said the earth moves around the sun. Everyone knew he was wrong. You could see the sun moving.

Wittgenstein later asked: what would it have looked like if the sun really did go around the earth?

The answer? Exactly the same.

The problem was never stupidity. It was the box. Every era has assumptions so baked in that leaving them feels like insanity.

Gutenberg thought he'd print a few Bibles. Someone at IBM supposedly said the world might need five computers (he didn't, but the story sticks). People in 2019 called LLMs a research curiosity.

They weren't fools. They were in a box..

Accidents Make Everything

On October 22, 1707, Admiral Shovell ran his fleet into rocks off the Scilly Isles. Nearly 2,000 men drowned. He ignored a navigator's warning—the hanging story is probably legend, but the disaster is not.

Parliament offered £20,000 for a way to find longitude at sea. The answer was time. That required better clock springs. A clockmaker named Huntsman made crucible steel that worked exceptionally well for springs. That steel also made cutting hard metal easier, which led to better cannon barrels. Wilkinson bored them. Napoleon loved light artillery and did very well from them, except perhaps his last battle... Driven by his success and the needs of his army he offered a prize to anyone who could preserve food on his long campaigns. A champagne maker named Appert won it for bottling and boiling food.

An English company bought Appert's patent—and his patent for a continuous papermaking machine. Cheap paper was flooding the world because linen was becoming far cheaper to make because of the loom, which led to waste linen and therefore almost free raw materials for paper. Eventually, Seth Wheeler patented the toilet roll.

Admiral Shovell cannot navigate. Therefore: toilet roll.

That chain is how everything works. No one planned it. And it wasn't a one-off. Maybach made the carburetor from a perfume atomizer. Domagk spilled red dye and discovered Chemotherapy.

The 2017 AI Transformer paper was not written to pass medical exams or fold proteins. But here we are.

We are still on the first page of that chain.

Descartes Sold You a Lie

Descartes gave us a powerful tool: break problems into pieces, doubt everything, build from first principles.

It gave us modern science. It also gave us silos.

Now you have PhDs on Milton's comma. Rocket Scientists who never talked to ethicists. Policymakers who never talked to teachers. Each group is solving its own problem. None sees the whole chain.

Reductionism made us great at depth. It made us terrible at connection.

That is changing. Not because we got smarter. Because the machine handles the linear work, and your brain can finally do what it was built for: seeing across the seams.

Today, the talk is about where AI can save 20% on headcount.

Those people are the Dutch weavers. They have misidentified the problem.

The Other 99% Are Coming

In 1400, maybe 5% of people could write. Michelangelo, Newton, Jefferson—they were the one‑in‑a‑million who got access.

What happens when 5, 7, perhaps more, billion people get access?

AI does not just democratise information. It democratises capability. A student in Lagos with a phone now has research help that a well‑funded professional could not get five years ago.

You have no idea what the other 99% can do. Neither does anyone else. But history says: quite a lot.

The Real Problem Is Trust, Not Intelligence

In the 1990s, Patti Maes wanted software agents that could tell margarine from butter. Today, the same machine that gives you the right answer also gives you a perfect lie. Same fluency. Same confidence.

The problem is not capability. It is provenance.

Also: your AI is not neutral. It was trained on English, on Western data, on WEIRD psychology. It works better for a lawyer in London than for a farmer in Lagos. The tool you think will liberate the excluded speaks the language of the included.

Solvable. Not being solved fast enough.

So: Pause Before You Kick

You know where this is going. The Admiral leads to the toilet roll. The loom leads to cheap rag paper, the printing press, and the Reformation. The AI Transformer paper leads somewhere you cannot see.

That is precisely why you should not run headlong.

Many organisations right now are asking: Where can we deploy AI by next quarter? Will AI be replacing all our jobs? Where can we get ahead? Will we be put out of business? Those are the wrong questions. 

You do not know where this chain ends. Neither does anyone else. That is not a weakness. It is a reason to stop, think, and choose deliberately.

A human pause. Not a moratorium. Not fear. Just a deliberate breath.

• Map the chain you are already inside.

• Ask what kind of future you want to accelerate toward.

• Then decide what to automate, what to augment, and what to leave entirely alone.

The Dutch weavers did not pause. They smacked the loom, then adopted it blindly. You can do better.

When the Checklist Ends, Think. Only Then Kick.

On the moon, the cuff checklists had a final instruction: If all technological repair options have failed, try kicking it. Percussive maintenance. A lunar boot.

That is not a licence to act without thought. It is a licence to act when you have thought carefully.

You are not at the bottom of that checklist yet. There is still time to pause.

The loom is being smashed. The paper will be cheap soon. The new world is further off than it looks and closer than it seems.

So take the pause. Ask the boring, slow, human questions. Decide what you actually want.

Then, when you are ready, kick with a lunar boot.  But only as a last resort.