Packed Thoughts

24

August 5, 2026

The minister woke up one morning and said 24.

Or he saw it on a licence plate. 24. Even. Round. Easy to say twice and remember.

Or he asked the consultant. The one who knows everything. The one who was behind a shawarma grill last year and sits on four committees this year.

None of that happened. And that is the point.

Last month the Ministry of Electricity sent the usual message. Feeders failed. Two areas lost power. Set the air conditioner to 24 or higher.

They have been saying 24 every summer since 2019. And in 2024 they stopped asking and started cutting, two hours a day, across forty areas. So it is not a slogan.

Two weeks earlier, New York asked the same thing. Set your air conditioner to 78.

78 Fahrenheit is 25.6 Celsius. Kuwait asked for the colder room.

Kuwait says 24 because that is what the buildings were built for.

When an engineer designs cooling for a building here, he assumes 45 degrees outside, 55 on the worst day, and 24 inside. That last number decides everything. The size of the ducts. The power of the unit. How much cooling the building can produce when it matters.

So the ministry is not asking you to suffer. It is asking you to use the machine at the setting it was designed around. Below 24, you are asking the building a question it was never built to answer.

New York says 78 because that is what the research says is comfortable.

The number started in 1979 as an emergency order during the oil crisis. It was cancelled two years later. But it stayed, because the science caught up with it.

ASHRAE, the American engineering body, spent decades studying how people actually experience temperature. Their comfort range for summer runs from about 22.5 to 26. Both 24 and 78 sit inside it.

And every degree you raise the thermostat saves roughly one to three percent of your cooling bill. That was measured, not guessed.

Two countries. Two completely different reasons. One and a half degrees apart.

Does it work?

Kuwait’s campaign took 30 megawatts off peak demand in 2025. Real. Also about 0.2 percent of a peak near 17,600 megawatts.

New York asked and still had outages.

Not nothing. Not enough.

Because 24 is the cheapest thing a government can do and the only thing it can do this afternoon. The real answers are power stations, insulation, better compressors, and houses that do not leak. Those take five years and billions. 24 is what you do while you wait.

Now the part that matters.

Ask Kuwait where 24 came from and there is a building code. Ask New York where 78 came from and there is a comfort standard and an energy calculation.

You can disagree with both. You can say 24 is too generous for a country heading toward 19,000 megawatts. You can say 78 is an old American number the rest of the world has no reason to inherit.

But you can argue, because there is something to argue with.

That is all evidence-based decision-making means. Not being right. Just being able to point at where the decision came from.

Now count how many decisions we live under can do that.

The requirement that changed. The circular that reversed last year’s circular. Ask for the study and you get a name. Ask for the name and you get a title. Ask for the title and you get a committee. Ask about the committee and you find the man.

Last year he was selling shawarma. Or he was a حارس عمارة. Then someone knew someone, and now he is a consultant. The salary is 180. The allowances are ten times that, because the salary is for the job and the allowances are for the meetings, and there is always another meeting. He has never opened a building code. He does not need to. He has an opinion, a seat, and a signature down in the corner as two ugly tiny initials.

His number becomes policy exactly the way 24 did. The difference is that when you ask where it came from, there is nothing at the other end. Just him.

A number from research can be wrong and still be honest, because it tells you how it got there. A number from a chair is right until the chair changes.

24 is not correct because it is 24. It is correct because someone can show you where it came from.

Ask a number where it came from. The good ones answer.

Fun fact. Setting the air conditioner to 16 does not cool the room faster. The machine removes heat at one fixed rate. No setting makes it work harder. All 16 does is keep it running past the point where it would have stopped.

Elsewhere.

Where Number
Kuwait 24°C
UAE, Saudi, Qatar 24°C
United States 78°F / 25.6°C
India 24–25°C
China 26°C
Japan 28°C
France 26°C minimum
Spain 27°C
United Kingdom Nothing

Write in the logbook
The Silence

August 4, 2026

Everything I do for a living traces back to two men.

One explained how words work. The other taught me my first ones.

The first man.

The field I work in, computational linguistics, was rebuilt from the ground up by Noam Chomsky. Before him, linguistics was mostly cataloguing. Recording what people said. Sorting sounds into boxes. In 1957 he argued that language is not a habit we pick up but a structure we are born with. That a child in Kuwait and a child in Chile are running the same machine with different settings.

That idea did not stay in linguistics. It built computational linguistics. It built natural language processing. It built the theoretical floor underneath every large language model working today, from Claude to ChatGPT.

And he did it twice. The same man rebuilt how people talk about power. Manufacturing Consent, written with Edward Herman in 1988, argued that in open societies control does not come from censorship. It comes from shaping what feels reasonable to say, so the boundaries feel natural instead of imposed. You do not need to ban an opinion if you can make it unthinkable.

Two fields. One Chomsky.

In June 2023 he had a massive stroke.

He is 97 now. He lives in Brazil. The right side of his body does not work. His assistant of twenty-four years said publicly that he is not communicating much, if at all.

His wife said something that has not left me since I read it. He still watches the news. And when images of Gaza come on the screen, he raises his left arm. The one that still works. A gesture of anger from a man who spent seventy years explaining exactly this, and now has one arm and no sentence.

Look at what arrived while he was quiet.

AI. The thing his own work made possible. He had already called it high-tech plagiarism. Whether he was right is the largest argument in my field right now, and it is being conducted in his absence. Both sides quote him. He cannot correct either one.

Trump and Mamdani. A president promising borders and a mayor promising free buses, and most commentary treats them as opposites. Chomsky’s framework says they came out of the same hole: an economic order that stopped delivering, and anger that had to go somewhere. Upward at corporations and landlords, or sideways and down at immigrants and whoever has less power than you. He said this out loud years ago about Trump and Sanders. Nobody can ask him about this pair.

Gaza. He described the siege for decades. Now he raises an arm at it.

And the Epstein files. Documents released over the past year showed a friendship far closer than his readers understood. He called Epstein a highly valued friend. Months before the arrest he was advising him on how to handle the press. His wife issued a statement admitting serious errors in judgment. She had to write it because he cannot. A man who spent his life mapping how elite networks protect each other cannot explain why he did not see one that walked into his own living room.

The one question that would actually require his voice is the only one he cannot answer.

The second man.

My father did not write books. He built things.

He turned an area that was sand into an area that is green. Trees that are older than my children now. He worked on the mosque. Not from an office. On site. He built things for neighbours who never asked him to, because he saw something broken and could not leave it. Half the people on that street have something in their home that my father made or fixed or paid for quietly.

And he raised me. Which is the accomplishment I use every single day, in every post on this blog, in every argument I win and most of the ones I lose.

He has dementia and Alzheimer’s.

He does not speak anymore. What is left is emotion. Laughter when something reaches him. Sadness when something else does. And silence most of the time.

He cannot say my little one’s name. He has never said it. My son will grow up knowing his grandfather as a man in a chair who sometimes laughs, and he will never hear his own name in that voice.

He cannot see what his grandchildren are becoming. He is in the room while they become it.

And he watches the news. That is the part that undoes me. When Gaza comes on the screen, something changes in his face. He cannot name the place. He cannot say what happened or who did it or what he thinks should be done. But he knows. The anger arrives without the vocabulary. He makes a sound. He turns away. Sometimes he lifts a hand.

Two men, seven thousand miles apart, in two different rooms, watching the same broadcast, raising the same hand at the same thing, and neither one able to say a single word about it.

Two silences. Different causes. Same result. The thoughts are still in there and there is no door.

And here is the irony sitting underneath all of it.

Both silences are described using the first man’s framework.

Aphasia, the loss of language after brain damage, is understood today through the model of grammar Chomsky built. When a neurologist maps which part of a sentence a patient can still produce and which part is gone, they are working inside his system. The language loss in dementia is measured the same way. He gave medicine the vocabulary for exactly this.

So the doctors examining Chomsky are using Chomsky’s tools to measure Chomsky’s silence.

And the doctors who examined my father used those same tools to tell me what he was losing, and in what order, and roughly when the next thing would go.

The man who explained how humans produce sentences also gave us the instrument for measuring their disappearance. He built it. Now it is pointed at both of them.

I sit with my father and watch him react to something on the screen and I know he understood it. I just cannot get to it and he cannot get out. Somewhere in São Paulo, a woman sits in the same room with the same problem, watching a man lift one arm at a news broadcast, knowing what he means and unable to ask him anything else.

One of them built the modern study of how humans make sentences.

The other one taught me my first.

Neither of them can make one now.

And the cruellest part is what the first man taught us.

Manufactured consent rests entirely on who gets to speak and who gets spoken for. He spent seventy years warning about people who are described rather than heard.

Now he is one. His wife speaks for him. His admirers explain what he would say. His critics explain what he really meant. And he sits in a room with the news on, raising one arm.

My father is one too. I speak for him at the hospital. I answer questions about his day that he is sitting right there for.

He explained the machine to us.

Then his own stopped. And so did my father’s.

Write in the logbook
The Blueprint: DHAKA

August 3, 2026

Developing High-schoolers’ Abilities, Knowledge And AI. DHAKA. In Arabic, ذكاء. Intelligence. The plan is named after what it builds.

Something is coming to Kuwait’s high school. In the last weeks the newspapers have carried it: after twenty years, the ministry is reported to be bringing back نظام المقررات, the credit system, replacing the year-group model that has run since 2005.

I say reported, and I mean it. There is no decision number, no gazetted circular, no minister naming it on the record. What exists is one detailed newspaper account and the discussion around it. So everything that follows is built on what was described, not on a law. Hold that.

But from what was described, one thing is clear enough to talk about: it looks like Saudi Arabia’s system.

I care about this because I have a son who will sit in whatever system Kuwait builds next. And because I have spent this whole blog asking for systems that see the person. If this reform is real, it is one.

What Saudi Arabia did.

Saudi Arabia rebuilt its high school around three ideas. A shared first year that every student takes together. Then a split into tracks the student chooses. Not the old science-versus-arts split. Five real paths: general, health and life, computer science and engineering, business administration, and Sharia. And a credit system across three semesters a year. You pass a course, not a whole grade. A failed course is retaken alone. Not the whole year.

The point is choice matched to a person. A student leans toward health, or engineering, or business, and the last years of school start preparing them for that. Whether their next stop is university or work.

The version reported for Kuwait is thinner. Two tracks instead of five. But the bones are right: flexibility, choice, a student treated as someone with a direction rather than a seat in a row.

But here is what I could not find in it.

Nowhere in the reported plan is it clear how artificial intelligence fits. There are fragments. A “future sciences” track. A phrase about “future skills.” An AI-and-cybersecurity reform that Kuwait already ran for the younger grades. But at the stage that matters most, the last stage before university or a job, AI appears as a slogan. Not a spine.

That is the wrong century to be vague in.

The problem.

Kuwait is redesigning its high school and has no clear place in it for the one skill that will define the next forty years of work.

The evidence.

The evidence is in the silence. The reported plan names health, engineering, business, and humanities. It says “future skills” without saying what they are. Kuwait put an AI curriculum into grades one through nine last year. Then, at the secondary stage, the reform goes quiet on it.

Meanwhile Kuwait sits near the bottom of the Gulf on government AI readiness. And the ministry has a recent record of failing to deliver a digital reform, not for lack of money but for lack of preparation. During COVID, distance learning collapsed. Not because the screens were not there. Because the teachers were not trained to teach through them.

The appetite is there. The plan is not.

Who else faced this.

Here honesty matters more than enthusiasm. Because the countries that rushed AI into schools are a warning, not a model.

South Korea ordered mandatory AI textbooks and spent about a billion dollars. Within two years it downgraded them to optional. Its own auditor found the decision had been made in seven internal meetings with no outside consultation. A billion dollars decided by seven meetings.

Peru put a laptop in every child’s hands across 531 schools. For a decade. And measured no learning gain at all. Zero. Because the teachers were never trained to use them. The laptops sat on the desks. The teaching stayed the same. The technology was there. The preparation was not.

China is the sharpest warning. A study of twenty-six thousand students found that AI raised their homework scores and cut their exam scores by twenty percent. The damage was hidden until the second year. The students who asked the AI for answers were harmed. The students who used it as a thinking partner were fine. Same tool. Two uses. One helps. One destroys.

Two countries did it well. And both did it slowly.

The UAE trained over a thousand teachers in a preparation year before a single mandatory AI lesson reached a classroom. Teachers first. Content second.

Estonia built its own AI tool that refuses to give answers. Instead of solving the problem for the student, it asks guiding questions. It forces the student to think. Because Estonia’s designers understood something the rushed countries missed: an AI that hands over answers destroys the struggle that learning is made of. And the struggle is the learning.

The lesson across all of them fits in one sentence: AI helps only when it sits on top of trained teachers and real foundations. It harms when it replaces them.

What Kuwait should do.

Six things.

AI as a thread, not a track. Not a single computer path that a few students pick. A strand woven through every track. Health, engineering, business, humanities. Because a nurse, an accountant, and a mechanic will each use AI. Differently. One owner for that thread across the whole secondary stage.

Teachers before tools. No AI content reaches a classroom before its teacher is trained to teach it. The UAE’s rule. Estonia’s rule. Kuwait’s own COVID collapse is the proof of what happens when this step is skipped.

Teach the tutor, ban the vending machine. There are two ways to use AI. As a tutor that asks you questions and makes you think. Or as a vending machine that gives you the answer and lets you stop thinking. Adopt or build an AI that asks questions, like Estonia did. And teach students the difference explicitly. Because the Chinese evidence says that difference is the whole game.

Foundations first. AI literacy sits on top of reading, numbers, and clear thinking. For any student not yet solid there, those come first. Do not paper over a weak foundation with a shiny tool.

Serve the ones not going to university. The majority. Build the vocational tracks that Saudi Arabia has and Kuwait’s thin version lacks. A skilled worker who also commands AI is the most employable person Kuwait can graduate.

The public number. Teachers certified to teach AI. Published each year. Classroom rollout never allowed to run ahead of it. The number that keeps the plan honest.

The point.

The reform Kuwait is reported to be planning is, in its bones, a good one. If it mirrors Saudi Arabia’s, it is better than what we have. But a high school rebuilt for the future with AI left as a fragment is a house wired for yesterday.

The evidence is unusually clear for once, and it cuts against the hype: AI rushed in harms learning. AI built on trained teachers and real foundations lifts it.

Kuwait has the appetite and, right now, a rare opening. A whole secondary system being redrawn at the exact moment the technology matters most.

Put AI at the centre. Carefully. Teachers first. Do not let this chance pass as a slogan in a track nobody was prepared to teach.

The plan is named after what it builds.

ذكاء

Write in the logbook
Prejudice

August 2, 2026

I watched a clip the other night. Norman Finkelstein and a Twitch streamer arguing about whether Candace Owens is antisemitic.

I do not care about that argument, or its result, and this post is not about her.

What interested me was the vocabulary. Finkelstein called it prejudice. The streamer called it unconscious bias. Both were describing something linguists have studied for fifty years, and neither used the word for it.

The word is categorisation.

How a child learns grammar

Nobody hands a child a grammar book. She works it out from the sentences she hears.

She hears walked, played, jumped, wanted. At first these are four separate words. Then she notices the shape underneath them.

Verb + ed = something that already happened.

Now she can do something new. She hears “zoom” for the first time and says “zoomed.” Nobody taught her that.

You can tell she built the rule rather than memorised it, because of the mistakes. She starts saying “goed” and “runned.” Words no adult has ever said to her. She had “went” correct at two, and at three she says “goed.”

She has not got worse. She has learned a rule, and the rule is beating the memorised exception.

Michael Tomasello documented this across hundreds of children. Joan Bybee found the same process in adults, and made the point that matters here: this is not a language system. It is how the brain works generally. Language is just where you can watch it clearly.

How you learned the word “bird”

Nobody gave you a definition. You saw sparrows, pigeons, crows, chickens. Similar enough to group together.

And some sit closer to the centre than others. A sparrow feels more bird-like than a penguin, because you saw more sparrows. Whatever you saw most became the standard.

Your category is not a definition. It is a collection of remembered examples.

The brain uses the same system for people.

The same process, applied to a group

In the clip, Finkelstein reaches for one of the oldest stereotypes there is: that Jews are cheap.

Here is how someone ends up holding it.

The examples arrive over years. A joke at a family gathering. A film character careful with money. Shylock in the school syllabus. A relative watching someone haggle and saying “what did you expect.”

Nobody decides that fifteen million people share a spending habit. The examples just accumulate, the same way “walked, played, jumped” accumulated.

Then two things happen.

Repetition makes the idea familiar. Familiar things are easier to recall, and easy recall gets mistaken for evidence.

Variety turns it into a rule. A merchant, a shopper, a film character. Different people, same apparent behaviour. That is what turns separate memories into a general pattern, the same reason a child who hears “-ed” on twenty verbs uses it on a twenty-first.

Now the rule works on new cases. Someone negotiates hard over a price, and the brain reads the new event through the closest stored category.

Same operation that produced “zoomed.”

And it does not arrive as a belief. It arrives as an impression. The person cannot trace where it came from, because the examples stopped being separate examples years ago.

And that is the point I want to make

While I was watching the interview, I formed a prejudice about the streamer.

His eyes never stopped moving. He was reading the chat constantly, mid-sentence, mid-argument. And I decided fairly quickly what I was looking at: a clip farmer waiting for the next viral moment and the paycheck attached to it. Someone whose positions arrive from a scrolling chat window instead of from preparation.

I did not choose that judgment. It assembled itself, from every previous streamer I have watched, exactly as described above.

Then I kept watching. I let the algorithm take me through more of his clips.

Which is the opposite of what my prejudice recommended.

That is the whole argument, and it is the one Finkelstein was making. Forming the pattern is not the offence. Acting on it is. My prejudice should not have been allowed to close the tab.

Because if it is, the world flattens. You are left with a few blunt colours and you lose every shade and texture inside them. Every person becomes the category they arrived in, and nobody gets to be anything else.

If that is how it works, my little boy will have ten words for the rest of his life. Which, knowing him, he may attempt anyway, purely to annoy me.

But seriously, this faculty is what makes us human. It is what lets us love, and love again after being wrong. Hate, and stop hating. See danger at a distance. Hold on to the people near us, and be held by them.

And it breaks down in Alzheimer’s. Which for me is worse than the memory loss, because this was the part that made my father my father. The constant joking. The banter. The laughter that never stopped. All of it ran on exactly this. Patterns.

Write in the logbook
Ministry of Blinking

August 1, 2026

I have been driving for twenty years and I now sit at every junction like a man taking an exam he did not study for.

Write in the logbook

Where	Number
Kuwait	24°C
UAE, Saudi, Qatar	24°C
United States	78°F / 25.6°C
India	24–25°C
China	26°C
Japan	28°C
France	26°C minimum
Spain	27°C
United Kingdom	Nothing

24

The Silence

The Blueprint: DHAKA

Prejudice

Ministry of Blinking