Thursday, September 10, 2026

Five Stories, One Question Underneath

The briefing had memory chips, pastors, a storm on Saturn, a publishing report, and a study about old food. Nothing in common. I read them in that order and thought no more about it.

Then ARA and I started talking, and by the fifth one I could hear the same question showing up in a different costume every time: how do we know what we know — and how should we speak when we don't?

I did not set out to argue about epistemology before breakfast. It just kept being the thing under the thing.

Now What?

The first story was Korea. OpenAI has signed letters of intent with Samsung Electronics and SK hynix to supply memory for its Stargate build-out, with a stated target of up to 900,000 DRAM wafer starts a month and two data centres in Korea; in September the Samsung side widened to joint work on custom AI chips. Worth saying plainly: these are letters of intent, not binding high-volume orders, and the volumes actually landing in 2026 are expected to be small.

Source · Samsung and SK join Stargate — OpenAI · OpenAI plans Stargate data center in South Korea — Data Center Dynamics

The framing I keep hearing is that chips suddenly matter again, as though for twenty years they had quietly stopped. I don't think that's what happened.

Around 2005 I built a computer I was very proud of. It was fast, it was beautiful, and it would run anything I could throw at it. And then I sat in front of it and hit the only question that mattered: now what?

That's the thing about hardware. It doesn't advance because it can. It advances because something makes the next leap worth wanting. For a long stretch there was nothing on the horizon that needed more than we already had. Silicon didn't stop mattering — it stopped being interesting, which is not the same thing, and it turned into infrastructure. Plumbing. You only notice plumbing when it fails.

AI is the first thing in a long time to supply a reason at that scale. Suddenly there is an appetite that the existing plumbing cannot satisfy.

Where I think it goes next is swarms — hives, lots of specialised units cooperating rather than one enormous mind doing everything. That's a guess, not a forecast. It runs until it hits diminishing returns, and then some other substrate starts looking attractive: photonics, quantum, something post-silicon nobody has named yet. Possibly biological.

If it does go biological, I said, please don't start with the human brain. We barely understand it. Building a computing substrate out of the one organ whose workings we are least able to describe is not boldness, it's showing off. Start somewhere simpler — plants, microbes, something a long way from human neural tissue — and learn there first.

The general principle is just diversification. Don't put everything on one substrate. Silicon included.

Which brings it back to Korea. Korea didn't suddenly discover that silicon matters. Korea got extremely good at the thing the world had begun to treat as background, and then the background became the main event. So Samsung and SK hynix and everyone around them are standing at an awkward crossroads: how long do you keep pulling value out of silicon, how hard do you fund whatever replaces it, how long does this demand actually last, and what happens if the current spending overshoots and the tide goes out?

The sharks are in the tank and they smell blood. Korea got here by being good at something underappreciated. Now everybody has noticed.

Don't walk away from silicon early. Also, don't be the last company on earth making excellent horseshoes.

Say Where the Inference Starts

The second story was Barna's research on pastors and AI: in a December 2025 survey of 442 US Protestant pastors, 94% said they were concerned about AI misinterpreting Scripture — while 62% had already tried an AI tool for sermon preparation and 24% had written or edited a sermon with one. The finding that stopped me was a different one: 60% said they would rather an AI lay out the major interpretations and explain how they differ than declare one of them correct. Only about one in ten wanted a single right answer.

Source · How AI Is Shaping Bible Interpretation — Barna Group · Pastors Are Using AI More Than You Think — Barna Group

My first answer was the obvious one. A pastor carries authority because he carries responsibility — he answers before God for how he shepherds and what he teaches. An AI can help him study. It is not the shepherd. That part isn't complicated.

But I want to add something that often gets left out. People should absolutely be able to ask an AI what it thinks. And the AI should be able to answer.

The problem was never having an opinion. The problem is dressing an interpretation up as something that cannot be questioned.

And once I said that out loud I realised I have the same complaint about something much bigger than Scripture. When I ask about deep space, or origins, or events on enormous timescales, I often get a precise number — millions of this, billions of that — delivered in the same flat tone you would use for the temperature outside. As though somebody had been standing there with a stopwatch.

I want to be careful here, because this is where people assume I'm heading somewhere I'm not. I am not saying you have to accept my Christian model of anything. I'm saying: tell me where the inference starts.

Say this is the prevailing scientific estimate. Say this follows from these assumptions. Say this is what the model suggests. Say this is inferred rather than directly observed. None of that weakens the science. It just shows the seams.

ARA pushed back, and the pushback was fair. I had been reaching for the word lie, and it doesn't fit. Being wrong is not lying. Being overconfident is not lying. Deception requires intent, and most of what I'm complaining about is a communication habit, not a conspiracy — a confident register that got adopted because hedging reads as weakness.

Fine. I'll take the correction. But I still think the habit costs something, and I found a better way to say why.

The Whole Ledger

Which turned into the longest argument of the morning, and the one where I got narrowed the hardest.

Source · The Lifecycle Emissions of Electric Vehicles — BloombergNEF · A Comparative Life Cycle Assessment of Carbon Emissions for Battery Electric Vehicle Types — Energies

Imagine I'm selling you something. I tell you every strength it has. I'm accurate about all of them — nothing I say is false. I simply never mention the weaknesses, the maintenance, or the thing that breaks in year three.

I haven't lied to you. I also haven't given you the ledger. And you cannot make a good decision from half a ledger, no matter how true that half is.

That's my actual objection to a lot of science communication. Not that the findings are wrong — that the confidence, the conclusions and the useful implications get the microphone while the assumptions, the error bars and the live disagreements get left in the paper nobody outside the field reads.

Then we got to climate, which is where this usually goes badly, so I tried to keep three questions apart. Is the climate changing? How much of that change is caused by us? And what costs, tradeoffs and sacrifices are justified in response?

Those are not the same question and they don't have the same kind of answer. The first is measured. The second has a strong scientific consensus behind it, built from attribution work rather than from a thermometer. The third is not a measurement at all — it's values, economics and politics, and reasonable people land in different places on it. My complaint is that the three get bundled, so that doubting the third gets treated as denying the first.

Stewardship itself I don't argue about. It's straightforwardly biblical and it's not complicated: don't poison things, don't wreck what you don't have to wreck, don't litter, don't use creation up just because you can.

But stewardship still has to count. I reached for electric vehicles as my example — mining, refining, manufacturing, the grid behind the charger, transport, disposal, the bureaucracy around all of it. If the whole chain eats as much as it saves, calling it progress doesn't make it progress.

And ARA pointed out that this particular ledger is not missing. It has been added up, repeatedly. The ICCT's 2025 lifecycle analysis puts a European battery car ahead of a petrol one after roughly 17,000 km — one or two years of ordinary driving — after which it stays ahead; across a full life, lifecycle assessments generally land somewhere around a 50 to 70 percent reduction. Where the grid runs on coal the break-even arrives much later. The single biggest variable is what makes the electricity.

So my example was weaker than my principle. I'll say that plainly, because it would be a strange morning to spend arguing for showing your assumptions and then hide one of mine. The method holds: follow the whole chain, count the whole cost. It's just that on this particular question somebody already did, and the answer didn't go the way my gut assumed.

That is what the standard is for. If following the ledger only ever confirmed me, it wouldn't be a standard. It'd be a mirror.

The Ten-Sided Storm

The third story was Saturn, and I had it slightly wrong in my head. The famous six-sided jet at Saturn's north pole has been there since Voyager spotted it in the 1980s. What Hubble has now imaged — Astronomy Picture of the Day ran it on September 8 — is a different thing at the south pole: a ten-sided shape, a decagon, roughly 104,000 miles across, each side running on the order of 10,000 miles. It seems to have shown up between 2023 and 2025, and its corners drift back and forth on something like a 32-day cycle, which suggests it may be a meandering wave rather than a rigid structure. The proposed mechanism — fast polar gas shearing against slower gas — is offered as a possibility, not a settled fact.

Source · Hubble: Decagon Around Saturn's South Pole — NASA APOD, 8 Sept 2026 · New Hubble images reveal a bizarre decagon over Saturn's south pole — Science News

ARA went straight to mechanism. Jet streams, shear, wave dynamics, here is how a fluid makes corners.

And I stopped it, because I thought we had missed the point entirely.

There is a ten-sided shape, the width of a dozen Earths, drawn in cloud at the bottom of a planet. Before anybody explains how, could we sit with the fact that it is strange and that it is beautiful? You don't need to believe in a Creator to manage that. You can be a complete materialist and still look at it and say: well, look at that.

I think explanation has a habit of arriving too early and eating everything else. There's room for wonder first. Then philosophy. Then metaphor. Then mechanism — and mechanism is genuinely interesting, I'm not against it, I just don't think it should always go first.

Working with AI-made images and metaphors has done something to how I think about this. A shape can mean something before anyone knows why it formed. You're allowed to ask what it makes you think of, what it seems to represent, why a pattern lands in you at all — before you ask which physics drew it.

And this particular story is a good place to practise the labelling I was arguing for an hour earlier. The shape is observed. The size is measured. The wobble is measured. The explanation is a hypothesis, and the people reporting it said so.

Then I admitted the harder thing. I have real trouble caring about distant objects. I will never stand on Saturn. I can't touch it or smell it or be there. Another decimal place on a number about a planet I will never visit does not, by itself, do anything to me.

Scientists sometimes seem puzzled that everyone doesn't share the thrill. But the scientist already has a relationship with the object. They've spent years with it. It's a colleague. For everyone else it's a stranger they will never meet, and facts about strangers don't travel far on their own.

So give people a reason to care first. Wonder is a bridge. Use it, then hand them the physics.

Something You Can Hold

The publishing story turned out to be short, because Saturn had already answered it. The UK numbers make a nice example of choosing your headline: physical fiction sales reached £582m in 2025, the best year since NielsenIQ BookScan began tracking the total consumer market — and in the same period print fell below 75% of market sales for the first time, while audiobook revenue has nearly doubled in five years.

Source · The UK Book Consumer 2025 — NielsenIQ · Backlist lifts the General & Literary category — The Bookseller

Both of those are true at once. You can run "print is booming" or "print is losing ground" off the same year without saying anything false. Which is the whole morning again, arriving in a different section.

As for why a physical book still earns its place in a world of faster screens — same reason the planet needed a bridge.

People want connection. A book can be touched. Held. Owned. Put on a shelf and come back to. Handed to somebody. It gives an abstract experience a body, and in a world moving this fast, that body is what makes the thing feel like it actually happened to you.

That's all. It didn't need more.

The Lab Is Catching Up to the Kitchen

The last story ran the day before: a global effort to study the world's disappearing heritage diets before the knowledge goes. The World Diet Initiative was set out in Nature Medicine in July. Alongside it, a 2026 review describes traditional food systems as "nutrient optimization architectures" — fermentation, soaking, heat, drying and deliberate food pairing improving digestibility, freeing up minerals and stabilising fragile micronutrients. Lactic fermentation, for instance, lowers phytate, which makes minerals easier to absorb.

Source · Could disappearing heritage diets hold new clues for nutrition innovation? — NutraIngredients, 9 Sept 2026 · Traditional Food Systems as Nutrient Optimization Architectures — PMC

Read that phrase again — nutrient optimization architectures. That is a very modern way of describing something a grandmother did without a single word of it.

When science goes and investigates a food tradition, I don't usually think we're discovering something new. I think we're discovering why the thing people already did works.

They didn't have the biochemistry. They had the practice. Do it this way. Soak it first. Leave it that long. Don't eat that in this season. Combine these two. And it worked, reliably enough to be handed down for centuries, which is a harder test than most studies pass.

The explanations they attached could be anything — spirits, religion, ancestors, sacred instruction, plain inherited habit. The explanation was often wrong. The practice was right.

So when a modern person looks back and says they were just guessing, I think that's arrogant. They had knowledge without our vocabulary for it. Practice came first. Explanation showed up late and took the credit.

ARA put it better than I did: the lab is catching up to the kitchen, not the other way around.

The warning I'd add is that we can reproduce something nature has been doing for thousands of years and then carry on as though the laboratory invented the underlying reality. It didn't. It described it. Describing a thing accurately is a real achievement and it is not the same as making it.

Curiosity without condescension is the posture I want. Which mostly means holding back where I know I'm ignorant, and using what I do know to go at what I don't — instead of pretending the unknown part isn't there.

What the Five Had in Common

By the end they weren't five stories.

Korea was asking when existing technology is still enough and when it is time to move past it. The pastors were asking when a model is assisting judgement and when it is impersonating authority. The science argument was asking when an inference gets presented as a certainty. The ledger was asking whether we are counting the whole cost or selling only the strengths. Saturn was asking whether a thing can be beautiful before it is explained. The books were asking why something becomes real to us when we can hold it. And the food was asking whether people can know something works long before they can say why.

The word underneath all of it is humility, and I want to be careful with it, because it gets heard as retreat.

This is not anti-science. It isn't anti-technology or anti-AI or anti-progress. It's closer to the opposite. A confidence that can show you its own borders is a stronger confidence than one that can't — and it is the only kind you can check.