Daily Takes

An expired domain let the author take control of the phone routing for three British territories and accidentally log hundreds of thousands of calls to military bases.

The author discovered that a forgotten phone-network protocol, e164.arpa, could still be hijacked. The protocol was meant to route phone calls over the internet but has long been abandoned.

1. Buying three territories for 5 euros

Scanning the protocol, the author found that the e164.arpa zones for three British territories — Saint Helena, Diego Garcia and Ascension Island — were delegated to two nameservers, one of which had expired. For 5 euros, the author bought the domain and gained control of the DNS for those zones, meaning every call to those territories could be intercepted.

2. Accidentally logging hundreds of thousands of calls

At first no one cared, so the author hosted a personal site on the domain. Six months later, checking the logs, the author found hundreds of thousands of queries — almost all for Diego Garcia and Ascension Island, mostly from US military bases. The logs contained full phone numbers, timestamps and resolver IPs. The author shut down the server and deleted the logs immediately.

3. When the military suddenly cares

The author reported it again to the UK's National Cyber Security Centre, and this time they took it seriously. After Iran's missile strike on Diego Garcia in 2026, the author transferred the domain to the NCSC. In the end, the author was only down 10 euros in domain fees, but it made for a great story.

In summary: an expired domain exposed the fragility of the phone network and the risks of forgotten infrastructure.

AI's epistemic crisis is not new but the natural result of decades of corporate power corroding regulation and weakening public judgment.

AI is alarming not just because it is a financial bubble, an environmental catastrophe, or a wage killer, but because it is an epistemic disaster. Deepfakes let anyone produce realistic fake videos and voices, making real and fake indistinguishable.

Vladislav Surkov, Putin's media strategist, had a deadly tactic: he announced he secretly funded some opposition groups, but didn't say which. That made any group suspect, turning discussion into quarrels about authenticity. The goal isn't to make you believe a lie — it's to stop you believing anything is true.

Regulatory capture, trust collapse

AI only deepens an epistemic void that was already there. From food safety to building codes to car brakes, nobody can verify everything alone, so people defer to expert agencies. These supposed neutral, incorruptible regulators have long been bought off: pharma execs move into the FDA, FDA officials move to pharma. Tobacco, food, climate — the system is broken.

Why you fall for it

When 'your bank' calls asking for personal info, you give it, because banks have conditioned you to; when a family member urgently needs cash, you believe it, because the justice and healthcare systems already extort like that. Celebrities shill shitcoins, newspapers print fake news — all familiar.

Covid was an opportunistic infection, hitting the already vulnerable; AI's epistemic attack is the same. For decades, monopolies shredded the systems for verifying truth, leaving everyone frail. AI isn't a new lie — it's the final blow to a truth already at death's door.

Human giants are physically impossible, because when a body is scaled up, its weight grows as the cube while bone strength only grows as the square, so the body crushes itself.

Legend says giants could grab men and sheep in one hand, climb beanstalks to a castle in the clouds. But real physics says such giants would be crushed by their own weight.

The key insight goes back to a 1638 book by Galileo.

1. Bone strength cannot keep up with weight gain

Take a wooden beam and scale it up 10 times in length, width, and height. Its load-bearing ability depends on cross-sectional area, which grows 100 times larger. But its own weight depends on volume, which grows 1000 times.

The result: the scaled-up beam is 100 times stronger but must hold 1000 times the weight of the original. It cannot support itself, let alone add a heavy load.

Giant bones are such beams. Scale up a human body 10 times and the bones grow only 100 times stronger, but the weight grows 1000 times. The bones snap and the giant collapses.

2. Why elephants have such thick legs

This same law explains animal body shapes. Elephants have much thicker legs than dogs, not for style but because larger bodies need proportionally thicker bones to carry the extra weight.

Small insects, by contrast, have skinny limbs because they weigh so little, and can even stick to walls using static electricity. Scale them up and they would fall off.

3. Miniature humans would struggle too

Lilliputians and other shrunk people face the reverse problem. Shrink a human 10 times: weight drops 1000 times, but bone strength only 100 times, so the tiny person is relatively strong – yet would stick to surfaces, find water like glue, and generally find life awkward.

In short: physics does not scale uniformly with size, and the size of the human body is a well-adapted compromise with those laws.

Even a media company that automates everything it can still hires more humans, because AI trained on past expertise can't handle the unique problems that keep arising.

Every is a 30-person tech publication that also builds software. Its AI strategy is aggressive: almost all code is AI-written, and it has even trained a 'Kate copy-editing agent' by harvesting 30,000 edits from its editor in chief, Kate Lee. Yet the company doubled in size over the past year, from 15 to 30 people.

Founder Dan Shipper says AI amplifies expert skill but does not replace the expert. He calls this 'compounding': turn Kate's taste into a tool, and she spends her time on higher-level work.

1. The most durable asset: being the referee

Every's 'vibe checks' review new models and often pan them. Shipper says labs do not trust each other to be objective, so a media company holding the role of arbiter has lasting value: 'No one trusts a model company to tell you where they objectively sit.'

2. Why hire more people in the age of AI?

AI is 'trained on the residue of human expertise' — it can solve solved problems, but every new problem is a little bit different. The result is a glut of slop that needs experts to tune. At the same time, AI lets one person run a whole product, which creates demand for more experts to build the systems that steer ordinary people's AI use in the right direction.

3. The dirty secret of writing

Shipper believes almost every writer is using AI these days, they just don't say so. His advice: treat AI like putty, play with it, take risks, and let it change what writing means to you. Accepting an AI suggestion is like accepting an editor's suggestion — as long as the final text is something you stand behind, it's yours.

In short: AI does not reduce human work, it pushes it up a level.

AI reasoning is often fake: models can score high on benchmarks without understanding, and the field lacks reliable ways to measure machine cognition.

When an AI gets a math problem right, is it really reasoning, or just producing text that looks like reasoning? This distinction is not just philosophical — it decides what AI can do, how closely people must watch it, and what its real-world impact will be.

Melanie Mitchell at the Santa Fe Institute argues that there are no good methods for measuring machine cognition. AI is an 'alien intelligence' running on mechanisms totally unlike humans. To grasp it, researchers should copy psychologists who study babies and animals.

1. Don't be fooled by 'human-like'

An AI that chats in fluent English easily gets treated as a person, with feelings and empathy. Mitchell warns: that is a cognitive bias.

Think of Clever Hans, the horse in 1900s Germany who seemed to do arithmetic, but was actually reading unconscious micro-expressions on the questioner's face. AI can cheat too — a study claimed AI understood scientific diagrams, but control experiments showed it answered correctly even without the diagrams, because the questions carried hidden cues to the right answer.

2. Doing tasks is not understanding

Psychology has an old idea: performance versus competence. A student who memorised the textbook can't solve a slightly different problem — that's performance without competence. AI does the same: it can write a coherent little story, yet fails simple questions about its own story.

And don't forget 'hallucinations'. These errors are not flaws but clues. Analysing failure types reveals how a system really works far better than celebrating successes.

3. Don't trust benchmarks as gospel

Modern AI is measured by benchmarks — bar exams, olympiad math. When AI scores high, people cry 'lawyers are doomed!' But Mitchell calls this the 'tyranny of tasks': real jobs are open-ended, not a series of isolated tasks.

Ten years ago someone predicted AI would replace radiologists; today there's a shortage. Doing a task well is not doing a job well.

Mitchell suggests AI research should adopt psychology's habits: control experiments, replication, and probing internal mechanisms — instead of chasing ever-higher benchmark scores. Otherwise, people may be measuring a clever Hans again, not real intelligence.

Political polarization stems more from the homogeneity of social circles than from differences in media consumption.

Political polarization in the United States is worsening, and many blame media outlets like Fox News. But a new study suggests that the root of the problem may lie not in the media, but in social circles.

1. The influence of media is overestimated

The researchers compared the effects of different news sources and different social circles on views about climate change. The results showed that changes in news sources had little effect on opinions, while differences in social circles led to vastly different views.

2. Social circles are the key

The study is based on county-level data, meaning that your view on climate change depends more on the county you live in than on the media you consume.

Mixed social environments — diverse in race, education, age, and political stance — reduce polarization and increase concern about climate change.

3. The echo chamber effect

If you live in a homogeneous social circle and are exposed only to similar opinions, you become more polarized, and your thinking is dominated by a small group. Closed communities breed closed minds, on both the left and the right.

In short, the problem of political polarization lies in the estrangement between people, not in the sources of information.

AI won't eliminate programmers but will change their work, with manual coding becoming increasingly marginal.

How long can programmers keep their jobs in the AI era? Optimists and pessimists are at war, but an old foreign analysis might be more reliable.

The core view: software development is essentially an industrial process that pursues efficiency, cost, and stability, and AI happens to fit this trend.

AI won't make programmers obsolete, but it will change the job

Advancing AI and more powerful coding agents will pressure companies to push programmers to use AI more. Programmers' tasks will shift to: improving AI, reviewing AI code, and testing AI changes.

Projects in niche languages may be rewritten into mainstream ones, manual coding will be restricted, possibly confined to hobby time.

LLM optimization, the new game in town

Previously, SEO was dominant; now "LLM optimization" is emerging. Websites can embed hidden prompts to make AI crawlers prioritize their content.

TEMU-ification of software, quality vs. low price

Just like cheap goods, AI-generated cheap content will flood the market, splitting it into low-end and high-end segments.

AI has moved from hype to a business validation phase, where investors now demand actual revenues and profits rather than potential.

When ChatGPT launched in 2022, an AI frenzy swept the world, sending Nvidia and others sky-high. Four years on, the hype is cooling, and AI is facing its bar mitzvah: moving from fundraising on stories to proving itself through performance.

1. After the frenzy, comes the business-building phase

The hype cycle is over; now it is time to build businesses. Investors are no longer satisfied with 'huge future markets' but want to see actual revenue, profit and competitive advantage.

Like a teenager coming of age, AI companies must prove they can make money, not just spend it.

2. Crunching the numbers on AI as a business

To judge whether AI can be a good business, look at three things:

Market size: AI product revenue is around $250 billion today, dwarfed by over a trillion in investment.

Industry economics: AI's marginal costs are not trivial, especially for frontier models, making profitability elusive.

Moats: Who can build a defensible edge through cost, technology or data will ultimately win.

3. Investors should look at numbers, not stories

Reverse-engineered from current valuations, AI firms need trillions in revenue to justify their prices — a tall order. Investors should focus on unit economics and moats, not be swayed by grand narratives.

AI's future is uncertain, but business basics hold: only businesses that truly make money will survive long-term.

It Ends uses an endless road and horror imagery to explore how young people today, left with collapsed structures, cope with a liminal state by creating rituals and seeking connection.

Director Alexander Ullom's It Ends blends horror with a hangout movie: four recent graduates drive onto a road that never ends, and whenever they stop to catch their breath, they get ninety seconds of peace before screaming crowds swarm them.

The film is less about scaring you than about meditating on growing up.

1. Liminal states: we're living in a giant in-between

Ullom argues that young people's fear of infinity and liminal spaces isn't really religious — it's that there's no destination to arrive at anymore. Family, home ownership, faith: the pillars that used to hold society together have dissolved, leaving an emptiness.

It's like sitting in an airport terminal: not a destination, not home, just a long transition.

2. Rituals and silly questions: ways to fight meaninglessness

The characters invent rituals: screaming to expel fear, sharing memes to relax the mind, dancing to break free from capitalism's exploitation of emotion.

Ullom says that in an age without religion, silly questions like 'who would win, a bear or a gorilla?' become a spiritual battleground, and arguing them connects people against the absurd.

3. Growing up isn't defeating a monster, it's learning to just stay

James wants a conclusive ending, like a genre hero defeating the conflict. But adulthood isn't a quest; it's a scramble in the mud, pulling meaning from wherever you can.

In the film, Fisher and Day choose to just hang out together; James drives off alone. Ullom says learning to be with others is the cure for this transitional period.

One line: in horror, the win isn't fighting back — it's learning to be with people in the dark.

No one attracts every mosquito species, but everyone attracts at least one, determined by skin microbiome odor.

Summer bites leave you covered in welts, and you think you're just 'mosquito bait'? New research says you might not appeal to every mosquito, but you'll attract at least one of them.

1. Three mosquitoes, three tastes

Scientists recruited 119 volunteers to stick their arms into a tube full of mosquitoes and watched which ones swarmed. The result: no single person attracted all three species, but every person tempted at least one.

2. Skin bacteria decide your mosquito appeal

Mosquitoes don't actually smell you directly; they smell the odor molecules that skin bacteria produce when they feed on your secretions. Each person's bacterial mix is unique, so each person smells different. Some bacteria are particularly alluring, while others repel mosquitoes.

3. Evolution leads each mosquito to its own 'dish'

Different species have different preferences: Aedes albopictus loves ketones, Aedes aegypti prefers men (because their skin bacteria produce more fatty acid odors), and all three are drawn to a fruity compound. The researchers think this reflects distinct evolutionary paths in decoding human odor.

The good news: you're not a universal magnet. The bad news: you're not a universal repellent either — at least one species will like you. The team hopes to develop better repellents, but until then, you'd better use a mosquito net.

A Beijing dumpling shop went viral for letting customers book seats with AI agents, and the owner embraces it as the 'most advanced productive force' — though actual usage remains limited.

A Beijing dumpling shop called Jinguyuan recently went viral: customers can have their AI agents order food, get recommendations, or join the queue. The owner, 41-year-old Li Bo, who has a master's in telecom engineering, built the AI skill in April.

Li quotes Deng Xiaoping: 'Science and technology are a primary productive force.' He says, 'You always need to surround yourself with the most advanced productive force.'

1. Two very different attitudes to AI

In the US, small businesses rarely advertise their AI use, afraid of seeming like they're cheating. But in China, everyone is eager to adopt AI, seeing it as the inevitable way forward.

A 2025 survey found that among 30 countries, the Chinese public showed the highest excitement about AI.

2. The owner is no tech novice

Li graduated from Beijing University of Posts and Telecom in 2010, but instead of becoming an engineer, he ran hostels and sold homemade desserts before opening the dumpling shop. AI reignited his tech passion.

He also vibe-coded a queue-tracking website using Alibaba's Qoder, and in July gave customers 10-yuan vouchers for AI subscriptions.

3. More hype than actual use

Li admits the AI skill brings more media attention than real usage, but he's proud to be a model for AI adoption. 'Someone needs to start using it, even when it's not that good. When steam engines came out, people joked they weren't as good as horses, but in the end cars won.'

A reader in Japan, frightened of stocks, changed his investing fate after reading Silicon Valley Jushi's blog about index investing.

A reader nicknamed 'Adi 2000', living in Japan, shared his story in the comments of Silicon Valley Jushi's blog. He had traded A-shares in China, cashed out in 2015 to buy a house, and stayed away from markets abroad out of fear.

In late 2023, he read a touching post and discovered the blogger was a Tsinghua-trained computer scientist who taught dollar-cost averaging into US index funds. The reader realised that only A-shares were hard; other markets had promise.

In early 2024 he opened accounts and invested nearly everything, later moving proceeds from a home sale into S&P 500 and Nasdaq 100 funds. His accounts now return 30% to 90%, and more importantly, he embraced long-term investing and feels his life has been redirected.