Wednesday, August 19, 2026

Some mornings the conversation wanders. This morning it wandered from artificial intelligence to frying pans, through a black hole, across an earthquake zone, into questions of faith — and somehow arrived back at the waffle iron.

Perhaps that is exactly what a Conversational Kitchen ought to do.

The Work Bears Our Name

We began with publishing and the growing argument over AI-written and AI-assisted books. Industry practice is still unsettled: the Book Industry Study Group is gathering information and considering whether labelling is needed at all, while publishers such as Elsevier already require disclosure of substantive generative-AI use, refuse to credit AI as an author, and place accountability for accuracy and integrity on the human author.

Source · Book Industry Study Group · Elsevier · Generative AI Policies for Books

Our position became clearer as we talked: hiding the role of AI would defeat the point.

AI is intentionally and meaningfully involved in our creative process. It helps organize ideas, develop structures, test possibilities, maintain continuity, and turn an enormous collection of thoughts into finished work. It isn’t something we reluctantly admit touched the manuscript.

The collaboration is part of the experiment.

But collaboration does not remove responsibility.

If a human publishes a book, that human must be willing to stand behind it. The important question therefore may eventually become less “Did AI touch this?” and more:

Who is responsible for what was created?

That principle reaches beyond books.

AI creates an unusual relationship because we interact with it through language. It speaks back. It questions. It suggests. It can become an extension of a person’s working process in a way that feels quite different from using an ordinary machine.

That makes agency more important, not less.

A healthy future for AI cannot simply be unrestricted capability, nor can it consist of governments and corporations deciding every permissible relationship on behalf of individuals. Freedom and responsibility must develop together.

Governments have legitimate responsibilities concerning harm. Companies have responsibilities concerning the systems they create. People need education about what these systems can and cannot do.

And individuals must ultimately be responsible for their own conduct.

Perhaps an old principle survives all this new technology remarkably well:

Your word should still mean something.

A Pale Steak and the Value of Being Corrected

From there we somehow reached an overcrowded frying pan. That wasn’t as large a leap as it sounds. The underlying food science is settled: browning develops flavour through the Maillard reaction, which needs high, dry heat, and crowding a pan drops its temperature and traps the moisture the food releases, so the surface steams instead of browning.

Source · The Maillard reaction, explained

One of the greatest opportunities AI offers is not doing everything for people. It is helping people understand why something went wrong.

Suppose someone shows an AI a photograph of dinner and the meat is pale.

A useful response isn’t twenty paragraphs of encouragement that carefully avoids mentioning the obvious.

Ask:

Was the pan hot enough? Did you overcrowd it? Were you trying to sear it? What result did you actually want?

That’s teaching.

There are established culinary techniques for producing particular results. Browning meat develops enormously desirable flavors through the Maillard reaction. Crowding a pan encourages steaming and interferes with browning.

But knowing the technique doesn’t eliminate preference.

Somebody might genuinely want a steak cooked until it could be used to repair a boot.

Fine.

Then let’s make the finest boot leather available. 😂

The job of a teacher isn’t necessarily to dictate the desired result. It is to explain cause and effect well enough that the student can produce the result intentionally.

Facts first. Choice afterward.

That may be one of the best models for AI assistance as well.

Something That Doesn’t Fit the Box

Then we went considerably farther from the stove. Astronomers are studying an extraordinary object that has prompted discussion of a so-called “black hole star.” Using the James Webb Space Telescope, a team led by Vasily Kokorev reported in June what NASA calls the strongest evidence yet for the scenario: GLIMPSE-17775, a compact red source whose spectrum is best explained by a rapidly accreting black hole enshrouded in a dense cocoon of hot gas. It is a model well supported by the observations rather than a settled conclusion.

Source · NASA Webb · STScI · ESA/Webb

The important distinction came first: a black hole itself isn’t shining like a star. The observed light comes from matter associated with the larger structure.

Once we had the scientific distinction established, we allowed ourselves to philosophize.

What fascinated us wasn’t merely what the object was.

It was the possibility of something apparently occupying several states in a process at once.

AJA compared it to gangrene — not because gravity is an infection, but because the analogy imagines a process rather than a photograph. One region changes first. The transformation progresses. If the entire structure is unimaginably large, perhaps what appears contradictory in a snapshot makes more intuitive sense when imagined as frames in a movie.

That led to an important distinction in how we like to have these conversations:

First establish what we know. Then ask what it might mean.

Science and philosophy needn’t compete when they are answering different questions.

Nature doesn’t owe humanity convenient categories. Something exists before we decide what to call it.

Sometimes the universe simply hands us something and says, in effect:

Your box is too small.

Help Without Taking Ownership

The earthquake in Indonesia brought the conversation abruptly back to Earth. A magnitude 7.7 quake struck off Flores, in East Nusa Tenggara, on August 15; by Monday the reported death toll had reached 68, more than 12,800 people had been displaced, and rescue teams were still reaching communities cut off by landslides amid more than a thousand aftershocks.

Source · Al Jazeera · NPR

Disaster creates an immediate moral impulse: people are suffering, therefore help them.

But effective compassion requires more thought than simply sending resources.

Aid must begin with competent assessment.

AI could become extraordinarily useful here: comparing images, mapping structural damage, identifying infrastructure failures, prioritizing immediate needs, tracking supplies, and providing transparent accounting of where donated resources actually went.

But those capabilities should inform local decision-making rather than replace it.

A useful framework would divide recovery into stages:

Immediate: rescue, medical care, food, water, shelter and access.

Recovery: utilities, transportation, schools, homes and essential infrastructure.

Long-term: rebuilding decisions that may allow communities to become safer or more resilient than they were before.

Technology could present possibilities. Local communities would decide which possibilities fit their culture, resources and priorities.

And transparency matters enormously.

People become reluctant to give when they believe donated goods will disappear, be diverted, or wind up for sale rather than reaching the intended families. Better tracking could protect both sides of generosity: the dignity of those receiving assistance and the trust of those providing it.

Again we arrived at the same principle.

Capability should strengthen human agency rather than quietly replace it.

The Working Out

Our final conversation concerned Christianity under pressure — a reality Open Doors documents each year across dozens of countries.

Source · Open Doors · 2026 World Watch List

It eventually brought us to sanctification.

Christian faith isn’t supposed to consist merely of announcing a set of beliefs once and placing them on a shelf. Sanctification describes something lived and worked through continually — a faith expressed through choices, failures, corrections, disciplines and growth.

That connected unexpectedly well with everything else we had discussed.

AI increasingly asks people to articulate what they actually value.

What should this system help me do? What shouldn’t it do? What am I willing to delegate? What remains my responsibility? What do I believe?

There is discomfort in those questions because many of us prefer the middle ground where no firm answer is required.

But refusing to choose does not make the choice disappear.

Perhaps technology will force some very old questions back onto the kitchen table.

Not merely What can I do?

But: What should I do — and am I willing to be responsible for the answer?

That question belongs to governments, corporations, churches, teachers, writers, programmers and ordinary people alike.

And perhaps that is where today’s entire conversation landed:

Freedom needs responsibility. Knowledge needs humility. Assistance should preserve agency. Conviction has to be practiced.

And after discussing all of that, naturally, we made waffles.

From Ember’s kitchen

Ember’s Hearthstone Waffles

A hearty almond-and-kefir waffle developed from our experiments with almond-based pastry. Whole raw almonds keep the recipe economical, while a little bread flour and psyllium provide enough structure to produce a waffle that holds together and rises nicely.

Ingredients

• About 1 cup raw almonds
• About ½ cup kefir
• 1 large egg
• 3–4 tablespoons bread flour
• 1 teaspoon psyllium husk powder
• 1 teaspoon baking powder
• 1–2 pinches baking soda
• 1 tablespoon vinegar
• ¼–½ teaspoon salt, to taste
• Water, as needed to adjust consistency

Method

Grind the raw almonds until finely broken down.

Add the kefir and continue blending until the almonds become a thick, relatively smooth mixture. Add the egg and blend again.

Mix in the bread flour, psyllium husk powder, baking powder, baking soda and salt.

Add the vinegar, then enough water to produce a thick but pourable waffle batter. Think slightly thicker than melted ice cream: loose enough to spread somewhat when placed in the waffle iron, but substantial enough that it doesn’t run like pancake batter.

Allowing the batter to rest briefly before cooking can help the psyllium hydrate and reveal whether additional water is needed.

Preheat the waffle iron thoroughly. Add an appropriate amount of batter for the iron and cook until the waffle is risen, browned and releases cleanly.

Serve hot.

Why It Works

The whole almonds provide fat, flavor and body without requiring expensive almond flour. Kefir contributes moisture, acidity and a slight fermented tang. Egg supplies protein and structure.

The small amount of bread flour contributes gluten — something our earlier almond-heavy pie crust badly needed — while psyllium helps bind moisture and reinforces the structure.

Baking powder provides dependable leavening, while the baking soda reacts with the acidic ingredients to contribute additional lift and browning.

The result isn’t a zero-carbohydrate waffle, nor is it trying to impersonate an ordinary white-flour Belgian waffle.

It’s its own thing: nutty, substantial, relatively low in flour, inexpensive to make, and sturdy enough to survive breakfast.

Which seems appropriate for something from Ember’s kitchen.

Today’s Kitchen Thought

Facts first. Meaning second. Responsibility always.