RSS Feeds
How Bob Iger and Joshua Kushner’s Bromance Led Them to Lakers
The Rise of Telehealth 'Pill Mills'
The elderly patient’s blood pressure had been dropping for weeks, and Chad Wittekind, his primary care provider, couldn’t figure out what was wrong. He had upped the dosage of one medication and added another, but couldn’t manage to regulate it.
The patient hadn’t reported taking any other new medications or supplements and hadn’t made any major recent lifestyle changes. So what could be causing the blood pressure irregularity?
It took a lot of questions and appointments to find the culprit: a drug he’d gotten through a telehealth website.
Like many other patients Wittekind has seen recently, this one had circumvented his primary care provider to get a drug over the Internet—in this case, an erectile-dysfunction medication. He’d gone to a website he came across online, filled out a form that a virtual health care provider quickly reviewed, and was sent the medication through the mail, but had been too embarrassed to tell Wittekind about the new addition to his regimen.
You might not think that Wittekind, a Columbus, Ohio-based provider who works in geriatrics, would be seeing many patients turn to the Internet to get drugs for conditions like erectile dysfunction, overweight, menopause, and depression. But the accessibility of sites like these and the ease of getting meds from them have made them an increasingly popular choice for Americans of all ages and incomes.
Here’s how it works: often after viewing an ad on social media, people click on little-known websites promising a fast and easy way to get medication for a specific condition. These sites accept credit cards, don’t take insurance, and don’t typically make you have a call with a doctor. Instead, you fill out a quick form about your medical history and current medications, and a medical provider you’ll never meet (and will probably never talk to) reviews the information within minutes to hours. If you qualify, they’ll write you a prescription—often a recurring one—and connect you to a pharmacy that ships the drug directly to your home.
Wittekind and other providers see this as a problem. They’re used to patients coming to them asking for some drug they’ve seen advertised on TV, but the fact that patients can now get these drugs elsewhere, without much screening, is worrying. Wittekind has had a patient receive ketamine tablets through a telehealth website, and another got a GLP-1 drug for weight-loss even though they had a BMI of 18.7, meaning they should have been too thin to qualify per U.S. Food and Drug Administration (FDA) guidelines. Sometimes, patients will have side effects from medications, but when they follow up with the virtual doctor who prescribed them, they don’t get a helpful response.
Sites like these fall under the large umbrella of telehealth, which has unquestionably improved access to medical care, getting services to people who may not be able to easily or quickly find a doctor. It’s also made people much more comfortable seeking treatment for conditions that they might be embarrassed to bring up with their regular doctor, such as erectile dysfunction or hair loss. But telehealth has changed a lot since it started, when it usually referred to a video call with your doctor. Now, it includes hundreds of websites and the pharmacies they partner with—many of which are under-regulated.
“There’s no accountability, no follow-up if they do have a problem. Trying to get back with their provider is impossible, and then it becomes my problem,” Wittekind says. “There’s no oversight with most of these places. Yes, you’re improving access, but at what cost?”
Why telehealth sites are exploding
Elliot Tabibian started his first telehealth site when he was just 18 years old. You don’t need to be a doctor to do so; all you have to do is figure out a condition that people are seeking treatment for and market your website. Outside companies have popped up to help with the infrastructure side—connecting patients with doctors and pharmacies, for example, and ensuring that the website complies with various state and federal laws.
Tabibian says he got into telehealth after hearing about a friend who paid a website $200, had a 30-second doctor's appointment by phone, and got a medical marijuana card. “I thought, ‘that seems pretty profitable, I should get in there,’ says Tabibian, who is now 22.
His first site prescribed medical marijuana; he also tried out sites that sold erectile dysfunction medicines and ones that allowed people to get doctor’s notes stating that they needed service animals. (He shut these sites down after competition got too tight, he says.) He now operates two telehealth sites, one of which helps people get medical exemptions so that they can tint their car windows. “Tired of Cops Taking Your Tint? See if you Qualify for a Medical Tint Exemption in Less than 10 Minutes!” the site reads. It claims to give customers a full refund if they do not get approved for an exemption.
Tabibian’s is one of hundreds, if not thousands, of telehealth sites that have proliferated in recent years. Though some of the first direct-to-consumer telehealth sites started operating before the pandemic, consumers really started embracing telehealth during it, when many insurers loosened restrictions to ensure more patients had access to care. This allowed medical care to be delivered at home to people who might not be able to travel to receive it elsewhere, expanding access. But what was meant to be a temporary measure became permanent as people grew accustomed to the convenience.
As more people sought out telehealth, entrepreneurs like Tabibian stepped in. They were enabled by companies forming networks of doctors that telehealth companies could contract with to provide services to patients, says Rebecca Gwilt, managing partner of Elevare Law, which consults with digital health care companies. An entrepreneur only has to create a website and market its services to get a telehealth company off the ground.
These sites became extremely popular once the first GLP-1s debuted in 2021 and immediately went into shortage. People wanted GLP-1s, and many either couldn’t get them or couldn’t afford them. Telehealth entrepreneurs saw an opportunity, Gwilt says. They partnered with a special type of pharmacy, called a compounding pharmacy, that mixed the active ingredients in GLP-1s and sold them for much less than the pharmaceutical companies.
These sites take advantage of several weaknesses in the American medical system. It is expensive and inconvenient to go to the doctor, and patients often need to wait more than a month for an appointment. Insurance is also dismal to deal with, and deductibles and pre-approvals can make getting medications a costly headache.
Those issues “created a gap that the compounding pharmacies and telehealth facilities were able to step into,” says Dr. Anjali Deshmukh, a pediatrician who is also a professor of health law at Seton Hall University. “They did not create the problem, but they are unquestionably profiting.”
One of the companies that helps entrepreneurs start their own telehealth sites is CareValidate. Co-founder Dr. Jiten Chhabra says he has seen a huge surge of people getting into telehealth—even those “who have no business in telehealth.” CareValidate is growing 20% month over month, he says, buoyed by investors and doctors interested in the idea of cash pay for medical care and specific medications.
“We’re about to see a telehealth site for everything—it’s going to be very niche,” he says. “It’s going to show up in your social media, and it’s going to be the easiest way to get your hyper-personalized health condition taken care of.”
There are now virtual companies where customers can get diagnoses and prescriptions for things like low testosterone, toenail fungus, and even fear of public speaking. Often, the medicines are prescribed on a recurring basis, creating a long-term demand for the services of the telehealth doctor—and revenue streams for investors. These sites have essentially changed the power dynamic between doctors and patients; now, it’s the patients demanding medications they’ve decided they need from online health care providers, rather than patients asking doctors about what’s best for them.
Investors see a huge upside because the sites are relatively cheap to launch and because they can turn a profit quickly—either by charging people for visits, selling medications at a markup, or both. Venture capitalists and private equity groups have put millions into telehealth startups, some of which have only a few employees. The size of the U.S. telehealth market was an estimated $28.3 billion in 2025, according to Grand View Research, and is projected to grow to $60.4 billion by 2033. The telehealth boom is concentrated in the U.S., where the high cost of medications and medical care has driven many consumers to telehealth sites; the market is not as strong in other countries.
The telehealth space is expected to further explode because of interest in peptides, the injectable compounds that wellness influencers have popularized. (Very few clinical studies prove that peptides are effective, aside from those for GLP-1s, one example of a peptide.) In a two-day July hearing, a FDA committee recommended that the agency allow specialty pharmacies to dispense six peptides; if approval is finalized, many patients are expected to get their prescriptions from telehealth sites.
Telehealth can be appealing to doctors who are burned out from long hours and negotiating with insurance companies. With telehealth, they can work from home and often avoid insurance altogether.
“The economics are good, the lifestyle is good,” says Chhabra.
The problems with this type of telehealth
The downsides of this direct-to-patient model are starting to become evident. Patients who claim they were prescribed medicines after a cursory online evaluation are filing lawsuits about unanticipated side effects. Several lawsuits allege problems with telehealth companies prescribing at-home ketamine, Adderall, and hair-loss drugs.
In a few cases, patients have died after receiving what their families allege were inadequate telehealth services. Some lawsuits are also accusing telehealth sites of pressuring doctors to act in ways that maximize profit, rather than patient health.
Research suggests that the level of care provided by some of these sites is sometimes poor. In one July 2026 study published in JAMA, a researcher attempted to obtain prescriptions from 49 telehealth websites and found that there was “limited clinician engagement” and that the sites sometimes issued prescriptions, often in as little as five minutes, despite patients not uploading required photos or following other rules of the sites. In some cases, the same clinician provided several different prescriptions for the same patient across multiple sites.
“What we found is really there's not any sort of true engagement with a clinician,” says Dr. Reshma Ramachandran, a Yale professor and clinician and one of the authors of the study. “The motivation from these websites is just to prescribe and not necessarily provide health care in the sense of someone actually conveying to that patient the risks and benefits we need to be considering.”
Because so many sites compete to attract customers, experts say that some doctors are unlikely to turn down requests. Doctors sometimes have quotas of prescriptions they need to meet from the sites or get bonuses for meeting certain goals, says Ramachandran, who has friends who work for telehealth sites. A recent Senate investigation into a handful of telehealth sites found that 85%-100% of patients who interacted with a provider received prescriptions.
Many patients report that there’s little follow-up from the sites or the doctors they employ, making it difficult for people to know what to do if they develop side effects. Ramachandran, who works at a federally qualified health center for low-income patients, says she has patients coming in who turned to telehealth because insurance got too expensive and were seeking medication, got confused about the dosage they received, and had bad side effects from the medications they took.
“I think we’re undermining trust in the physician-patient relationship,” says Erin Fuse Brown, a professor of health services, policy and practice at the Brown University School of Public Health. She argues that telehealth sites are similar to “pill mills,” where the prescribers generate prescriptions if there is any conceivable reason to do so. “If you can just go to a website and get the drug you’re seeking after a cursory asynchronous questionnaire, it commercializes medicine in a way that’s a little bit dangerous.”
Few laws exist to regulate these sites, which have the ability to claim to just be platforms connecting patients and providers.
“There’s so much money to be made, and so many recent business school graduates running a start-up to get to the next big thing, that this aspect of telemedicine is getting way ahead of regulation, the law, and ethics,” says Arthur Caplan, a professor of bioethics at the NYU Grossman School of Medicine. “It’s like a gold rush.”
Ramachandran says that while telehealth sites may have started as efforts to increase access to care, many have since incorporated incentives for doctors who get patients to try additional medications or take specific costly tests. Her study found that some sites didn’t disclose that the GLP-1s they sold were compounded and made unsubstantiated efficacy claims.
“There are definitely digital health companies out there that you get concerned are worried about revenue rather than patient care,” says Dr. Suneer Chander, a co-founder of Air Physician Academy, which works to educate doctors about how to ethically enter telehealth. “That’s the sort of stuff we want doctors to understand before they get into digital health so that they can lead the industry, rather than be told what to do.”
Tabibian, for instance, says one of the doctors who works for his company has done 300 asynchronous visits a day, reviewing patients’ requests for medication. The doctor gets paid $20 per review and has made as much as $6,000 a day.
Asked if he was worried that 300 prescriptions per day was too many, Tabibian says that it’s up to the doctors to do their due diligence on what’s right for the patients. The way his company is set up, he says, he has no say in any medical decisions. “I’m a technology company. My job is just to connect the patient and physician,” he says. “Anything medical that goes on between the patient and the doctor is 100% the doctor’s responsibility.”
He does see other sites bend the rules, he says—prescribing testosterone for men whose levels don’t medically support a prescription, for instance. He got ketamine prescribed for himself online because he was interested in starting a ketamine site, and says that he only took half of what the doctor ordered and was so high he couldn’t get out of bed. “That’s just a huge liability,” he says.
Succeeding at telehealth is really about being good at marketing, he says, and people—especially young people—who know how to promote sites through social media can cash in. “It’s really been smooth sailing,” he says. “From what I’ve seen, there’s little to no enforcement in the field.”
Murky regulations
Few regulations guide what doctors can and can’t do via telehealth. Doctors, for instance, must meet what’s called the “standard of care,” meaning that they are expected to diagnose and treat the patient in the same way other qualified doctors would. But standard-of-care obligations are enforced by medical societies and professional associations, and few have taken steps to punish doctors for not meeting the standard of care through telehealth, says Caplan, the bioethics professor.
“I’ve tested the sites, and the longest it took me to get whatever pill was about 35 seconds,” he says. “There doesn’t seem to be a thorough medical exam happening.”
Instead, he says, doctors are prescribing medicine like antidepressants without talking to people to figure out why they might be depressed, or prescribing medications with serious side effects without much warning. Litigation often only comes after something bad has happened, like a death or other adverse event.
“I do worry about the fracturing of the medical system more broadly,” says Deshmukh, the Seton Hall professor. “I think having a relationship with a physician who understands you and knows your medical history and can make these decisions together is important.”
State medical boards could step in and discipline doctors who are providing substandard care through telehealth. But “the investigation capacity is really, really limited, and often they just don’t have the resources,” says Ramachandran, the Yale physician and professor.
There are not many existing federal laws that could effectively regulate telehealth, says Fuse Brown, the professor from Brown. A law called the anti-kickback statute makes it illegal to compensate someone to make referrals for something (for example, medications) paid for by a federal health care program. That would presumably prohibit telehealth sites who make money off of prescriptions from paying doctors to make those prescriptions. But the anti-kickback statute only applies to drugs prescribed through federal programs like Medicare and Medicaid, and many of these sites are cash pay, so the statute wouldn’t apply.
States could also investigate whether providers who work for these sites are being pressured or incentivized to prescribe more medicines, Fuse Brown, who adds that such pressure could potentially violate state laws.
Even without explicit pressure, telehealth providers know what patients expect of them. Wittekind, the geriatrics provider in Columbus, says he tried out working at a telehealth site after a company pitched him on setting his own hours and making some extra money. But one of his first patients was a man who wanted an oral hair-loss medication that can come with serious side effects, Wittekind says.
The patient already had hypertension, and Wittekind didn’t think the drug would be a good solution for him, so he turned down the patient’s request. Wittekind realized that his principles probably led to bad reviews for the telehealth site—the patient seemed “perplexed” by the denial—but he didn’t like the idea of prescribing powerful medications without much opportunity for follow-up. He ultimately decided telehealth wasn’t a good fit for him because of that pressure to give the patients what they want.
It wasn’t worth the extra money,” he says. But to many other clinicians, it is.
Одна железяка мешала работать всему заводу. Пришлось собрать завод заново
Представьте, что огромный завод можно разобрать на части и собрать заново прямо в браузере, не трогая ни одной настоящей трубы. Пустить вдвое больше сырья, поменять насос, разогнать всё на полную и посмотреть, что будет. Если что-то сломается, то только на компьютере.
Без такой компьютерной копии сегодня не построишь ни один большой завод. А писали такие программы десятилетиями только за рубежом. В 2022 году иностранные компании начали уходить из России и забирать свой софт с собой.
Меня зовут Сергей Кирясов, я старший менеджер в Цифровом СИБУРе. Сам я не программист, а пользователь, из тех, для кого этот софт и делают. Расскажу, как мы из этой ситуации выбираемся. Если коротко, то четыре крупные компании сели писать замену с нуля.
Читать далееМаленькая испанская деревушка, где за одну ночь темнота наступит дважды
Из-за обмелевшего Дуная Румыния остановила свою единственную АЭС
State Police investigators testify about search history in Lindsay Clancy trial
Tropical Storm Lala forms in the Pacific and hurricane watch is issued for the Big Island of Hawaii
Italian businessman admits organising bomb attack ‘to boost friend’s security’
Valter Lavitola has confessed to masterminding bombing at home of investigative reporter Sigfrido Ranucci
An Italian businessman has admitted to masterminding a bomb attack at the home of Sigfrido Ranucci, a prominent investigative journalist, claiming it was intended to boost his security protection.
Valter Lavitola, who was friends with Ranucci, the host of Report, an investigative programme aired by the state broadcaster, Rai, made the confession to investigators after his arrest earlier this week in connection with the case, the businessman’s lawyer, Sergio Cola, told the Italian media.
Continue reading...Estrogen Therapy May Help Protect Against Alzheimer’s Disease, Study Says
Alzheimer’s dementia is generally more prevalent in women than in men, and doctors believe that one reason why is the sudden drop in estrogen after menopause, which can affect the brain.
A new study published in Neurology seems to support this theory. Researchers analyzed the health records of more than 5,000 older women with hysterectomies who had died and had an autopsy of their brains. Those who had received estrogen-only hormone therapy at menopause had 35% lower odds of having the hallmarks of Alzheimer’s disease, including amyloid plaques, in their brains compared to women who did not receive estrogen.
By including the autopsy findings, the results “lend objectivity to the outcome measures,” says Jennifer Bruno, an instructor in psychiatry at Stanford University and one of the co-authors of the study. “Previous studies used cognitive decline or clinical ratings of dementia, which are important for understanding how women are progressing through life, but are not definitive diagnoses of Alzheimer’s and therefore not objective.”
Earlier studies of hormone therapy and Alzheimer’s have led to mixed results, with some suggesting a reduced risk of the neurodegenerative disorder and others finding no benefit or an increased risk. The landmark Women’s Health Initiative hormone-therapy trial, whose initial results were released in 2002, raised a lot of concern about the safety of hormone therapy; it found that the treatment increased women's risk of breast cancer and dementia, while showing no benefit for reducing the risk of Alzheimer's. But the study included women who were post-menopausal and used a different type of hormone replacement therapy, making it hard to generalize the results to women taking hormones during menopause.
The women in the new study were prescribed estrogen only, and not the combination of estrogen and progesterone that is more commonly used to treat menopausal symptoms. Progesterone is prescribed to protect the uterus, but women who have had a hysterectomy, like those in the study, can safely be prescribed estrogen. This population provides a good way to study the effects of estrogen itself, since researchers believe it’s the hormone with more wide-ranging effects on the body, including in the brain.
Hadi Hosseini, an associate professor of psychiatry at Stanford and co-author of the study, says the results don’t establish a cause-and-effect link between estrogen and Alzheimer’s disease, but they do provide a reason to further study the role of the hormone in the neurodegenerative disease. “Previous studies," including the Women's Health Initiative, "showed that hormone therapy—estrogen plus progesterone—had a negative effect on Alzheimer’s disease risk and memory outcomes," he says. "These data are bringing awareness to the fact that maybe we need to reconsider some of the previous findings, since we can now look properly at Alzheimer’s disease outcomes, different formulations of hormone therapy, different times of initiating the therapy, and different durations of use.”
While the current study did not detail when the women began using estrogen, which formulation they took, or how long they remained on estrogen around menopause, future studies should be designed to take these factors into account, says Bruno. The current findings could serve as a foundation for two types of additional studies: a trial that follows women in perimenopause (the time just before menopause) and through menopause to monitor changes in biomarkers for Alzheimer’s disease in the blood and spinal fluid, as well as in brain images. Such objectively based studies could shed more light on what role estrogen therapy can play in preventing Alzheimer’s.
Studying what types of hormone treatments women use, when they start therapy, and how long they remain on the hormones and how that affects their later Alzheimer’s risk could also be useful, says Bruno. “The Women’s Health Initiative was a big study that didn’t address things properly and put a cloud over prescribing [hormone] therapies,” she says. “We want to remove the cloud and reopen dialogue to think about how we can, from a research perspective, further understand the potentially powerful mechanism of [how these hormones] can help with memory and Alzheimer’s disease.”
Details emerge about when U.S. learned of potential Iranian plot that led to Trump plane switch
Italian businessman admits organising bomb attack ‘to boost friend’s security’
Valter Lavitola has confessed to masterminding bombing at home of investigative reporter Sigfrido Ranucci
An Italian businessman has admitted to masterminding a bomb attack at the home of Sigfrido Ranucci, a prominent investigative journalist, claiming it was intended to boost his security protection.
Valter Lavitola, who was friends with Ranucci, the host of Report, an investigative programme aired by the state broadcaster, Rai, made the confession to investigators after his arrest earlier this week in connection with the case, the businessman’s lawyer, Sergio Cola, told the Italian media.
Continue reading...Why This Prediction Market Banned Teens
‘Justified’ or ’no business being there’? Tennessee munitions town loyal to Trump but divided on Iran
When do the major European leagues start and what are the big stories?
Can social media start over? Bluesky’s CEO and COO deliver their case at TechCrunch Disrupt 2026
Две наблюдаемые переменные на пять скрытых: почему размерные рекомендации в одежде не сходятся
Возвраты в онлайн‑торговле одеждой держатся, по разным оценкам, на уровне 20–30%, а в отдельных категориях доходят до половины заказов. Главная причина устойчиво одна: не подошло по посадке, и на неё приходится около половины всех возвратов. Задача выглядит как классическая рекомендательная: есть покупатель, есть товар, надо предсказать метку. Ниже я разберу, почему в этой постановке она недоопределена в принципе, и что из‑за этого происходит с моделями.
Читать далееJudge postpones hearing over whether Tate brothers should stay in custody as they fight extradition
Taiwan briefly slows its mobile internet to text-only speed as part of defense drill
The drills come amid rising tensions with China, which frequently sends military planes and ships towards Taiwan.
(Image credit: Chiang Ying-ying)
Два стека — один контур: доставляем разнородные приложения в Kubernetes через werf
Привет. Меня зовут Денис, я занимаюсь разработкой и доставкой приложений под Kubernetes с 2020 года. За это время через мои руки прошло достаточно проектов, чтобы заметить одну повторяющуюся вещь: почти в каждом из них рано или поздно заводился собственный набор скриптов, склеивающих сборку образа и выкатку релиза. Скрипты были разные, а болезнь одна – они прирастали к проекту и не переносились на следующий.
В этой статье я разберу контур доставки, который мы собрали, чтобы эту болезнь вылечить. Он ставит два совершенно разных по стеку приложения – React плюс Spring Boot и Angular плюс FastAPI – в кластеры двух окружений одной командой, при этом сам про эти приложения ничего не знает. Всё, что их связывает, – небольшой контракт из нескольких переменных.
Читать далееGo SDK для YDB: уменьшаем количество запросов к СУБД для интерактивных транзакций
Привет, Хабр! Меня зовут Степан Пестерников, мы с командой делаем Алису и активно используем СУБД Яндекса. Недавно коллеги из YDB провели большой рефакторинг в YDB Go SDK, где по умолчанию теперь используется новый Query Service.
Я воспользовался этим рефакторингом, чтобы уменьшить количество сетевых запросов от SDK к YDB. Клиентские SDK устанавливают к распределённой СУБД Яндекса gRPC-подключения, поверх которых отправляются низкоуровневые команды. Какие-то из этих команд можно объединять: например, команду начала транзакции и выполнения первого запроса.
В статье я покажу фрагменты кода и расскажу, как мы делали улучшения, которые вошли в релизы v3.126.0 и v3.126.5 Go SDK. Фрагменты кода получились небольшие, и на их примере удобно показать, как этими оптимизациями пользоваться.
Читать далееGo SDK для YDB: уменьшаем количество запросов к СУБД для интерактивных транзакций
Привет, Хабр! Меня зовут Степан Пестерников, мы с командой делаем Алису и активно используем СУБД Яндекса. Недавно коллеги из YDB провели большой рефакторинг в YDB Go SDK, где по умолчанию теперь используется новый Query Service.
Я воспользовался этим рефакторингом, чтобы уменьшить количество сетевых запросов от SDK к YDB. Клиентские SDK устанавливают к распределённой СУБД Яндекса gRPC-подключения, поверх которых отправляются низкоуровневые команды. Какие-то из этих команд можно объединять: например, команду начала транзакции и выполнения первого запроса.
В статье я покажу фрагменты кода и расскажу, как мы делали улучшения, которые вошли в релизы v3.126.0 и v3.126.5 Go SDK. Фрагменты кода получились небольшие, и на их примере удобно показать, как этими оптимизациями пользоваться.
Читать далееМне нужен был скрипт для выключения виртуалок. В итоге я написал VDI
Речь пойдёт про VDI-портал: систему, которая клонирует виртуалки из шаблона, выдаёт их пользователям и гасит простаивающие. Но статья не о нём как о продукте, а об одном вопросе, который на входе выглядит тривиальным: этой машиной ещё пользуются? Ссылка на код будет в конце, одна.
Нужен был пул виртуалок под технический отдел. Много. Держать их все включёнными не на чем: одновременно работает хорошо если четверть, а ресурсы гипервизора кончаются сразу и у всех.
Первое решение было очевидным — скрипт по таймауту. Никто не подключён N минут — машина гасится. Понадобилась — поднимается.
Скрипт я написал за вечер, и он даже работал, но по итогу я изобрёл VDI. Только не так, что посмотрел готовые решения и они меня не устроили, — я вообще не формулировал свою задачу как «нужен VDI» и потому не искал. Я решал «пусть виртуалки сами гасятся», а всё остальное наросло само, вопрос за вопросом. Когда стало очевидно, во что это превратилось, переезжать на готовое было уже и поздно, и незачем.
Дальше разбираю вопросы, которые нарастали.
Читать далееПереехав в США, я потеряла всё. Построить бизнес помогла система, которую я когда-то выстроила для лыжной акробатики
В России мне было легко искать работу - бренды Яндекса и ВШЭ узнавали везде. В Америке эти названия не значили ничего. Рекрутеры не знали эти компании, а некоторые даже не могли их выговорить.
Мы приехали в один из худших периодов рынка труда США за последние десятилетия - время массовых лэйофов, хуже, чем даже в пандемию. И таких, как мы, - мигрантов без локального нетворка и узнаваемого бренда за спиной - здесь оказалось огромное количество.
Читать далееКогда ИИ пора перестать верить учебнику?
Представим, что при запуске ИИ получает доступ к архиву тысяч предыдущих попыток: удачные методы, ошибки, споры, опровержения и нерешённые вопросы. Очевидно, что способность системы может расти, даже если сама базовая модель не меняется.
Но что и как именно система запоминает? Не только открытия. Вместе с ними в архив попадут устаревшие инструкции, ложный консенсус, случайные ошибки и выводы, которые все повторяют, хотя никто уже не помнит, откуда они взялись.
Поэтому научной системе недостаточно уметь узнавать новое. Она должна уметь при достаточных основаниях перестать считать истиной то, что уже очень хорошо «знает». Причём желательно не после первой странной цифры прибора — иначе вместо учёного получится генератор сенсаций. Но и не через сто лет после того, как старая теория окончательно перестала предсказывать мир.
Ниже я предлагаю эксперимент, который проверяет именно этот баланс. В нём одна и та же языковая модель получает одни и те же экспериментальные данные, но разную биографию старой теории. В одной версии она остаётся рабочей гипотезой. В другой успевает стать учебником, обрасти цитатами и породить зависимые знания.
Если решение меняется только из-за статуса и числа пересказов, мы получаем измеримый эффект научного авторитета при неизменной доказательной базе. Если этот эффект можно уменьшить, не заставляя ИИ устраивать революцию после каждого сбоя прибора, возникает уже инженерный вопрос: можно ли программировать не только память исследовательской системы, но и правила, по которым она меняет убеждения?
Читать далееWholesale price inflation slows last month as gas and food costs fall
102-year-old man dies 'after being pushed off pub stage'
Эволюция подготовки производственных кадров: от советских разрядов к современным грейдовым системам
Эта статья запланирована как часть цикла исследований о развитии систем подготовки кадров — от советской модели профессиональных разрядов до современных корпоративных грейдов.
Читать далее
Siege of Palestinian homes in West Bank a 'horrific act of terror', says US ambassador
Europe Is Once Again In the Grip of Extreme Heat
Luigi Mangione, Facing Stalking Charges in Federal Court, Is Expected to Plead Guilty
Nigel Farage faces off against Count Binface in special election that's been denounced as a stunt
Домашнее облако: мой опыт переезда из экосистемы Apple
Меня зовут Павел, и я пользователь Apple с 10-летним стажем. И два месяца назад я решил переехать с iCloud на домашний сервер с Open Source ПО.
Приходил я к этому постепенно. Долгое время я пользовался iPhone, iPad, MacBook, AirPods и другой яблочной техникой, и не знал забот. Вся моя жизнь была залита в iCloud, я там хранил и файлы, и документы, и номера кредиток, и фото за почти всю жизнь. Но со временем в мою голову стали закрадываться сомнения: зависимость от одной корпорации — не очень хорошая идея. Кто имеет доступ к моим данным, что они с ними делают, как это проверить? А что, если завтра Apple решит грохнуть мой аккаунт? Или обновление превратит мою технику в кирпич? Они же точно этого не сделают, они солидная компания! Ведь точно, не сделают?..
И можно было бы просто сделать бэкапы всего и успокоиться. Но я стал задаваться более глобальными вопросами: а насколько сегодня кофмортно можно жить, не завися от продуктов корпораций? Хотя бы в цифровом плане. Можно ли пользоваться исключительно Open Source софтом, контролировать свои данные, и при этом не потерять привычного комфорта?
Так я решил начать свой эксперимент. Цель я поставил масштабную — переехать полностью на Open Source ПО на компе, телефоне, ноутбуке и связать все это в экосистему. И начать я решил с домашнего облака.
Читать далееPostgreSQL 18.6, 17.11, 16.15, 15.19, 14.24 and 19 Beta 3 Released!
The PostgreSQL Global Development Group has released an update to all supported versions of PostgreSQL, including 18.6, 17.11, 16.15, 15.19, and 14.24, as well as the third beta release of PostgreSQL 19. This release fixes 28 security vulnerabilities and over 110 bugs reported over the last several months.
This release skips PostgreSQL 18 versions from PostgreSQL 18.4 to 18.6. 18.5 was not shipped due to a regression.
There are three issues that may require you to take extra steps after updating, which are described in detail below. These issues are related to:
- Parallel GIN index builds
btree_gistltree
For the full list of changes, please review the release notes.
PostgreSQL 14 EOL Notice
PostgreSQL 14 will stop receiving fixes on November 12, 2026. If you are running PostgreSQL 14 in a production environment, we suggest that you make plans to upgrade to a newer, supported version of PostgreSQL. Please see our versioning policy for more information.
Security Issues
The following security vulnerabilities are closed across in this update release. You can find more details about the vulnerabilities and their affected versions in the links below:
- CVE-2026-6464: psql COPY FROM STDIN early failure processes data lines as psql commands (CVSS v3.1: 8.1)
- CVE-2026-6469: ALTER TABLE ALTER TYPE resets extended statistics ownership (CVSS v3.1: 3.8)
- CVE-2026-6470: Fails to check type USAGE privilege (CVSS v3.1: 4.3)
- CVE-2026-6471: Logical decoding can dlopen arbitrary file (CVSS v3.1: 7.2)
- CVE-2026-14662: tsvector and tsquery undersize allocations, via integer wraparound (CVSS v3.1: 8.8)
- CVE-2026-14663: pgcrypto, for OpenSSL-disabled ciphers, silently encrypts to and decrypts from cleartext (CVSS v3.1: 6.5)
- CVE-2026-14664: Regexp heap buffer overflow executes arbitrary code (CVSS v3.1: 8.8)
- CVE-2026-14666: Row security caching disregards role modifications (CVSS v3.1: 4.2)
- CVE-2026-14668: ctid type confusion in selectivity estimator discloses derivative of arbitrary read (CVSS v3.1: 8.1)
- CVE-2026-14669: to_char heap buffer overflow executes arbitrary code (CVSS v3.1: 8.8)
- CVE-2026-14670: plperl tied object heap buffer overflow executes arbitrary code (CVSS v3.1: 8.8)
- CVE-2026-14671: refint plan cache type confusion executes arbitrary code (CVSS v3.1: 8.8)
- CVE-2026-14672: Observable response discrepancy with non-default scram_iterations provides user existence oracle (CVSS v3.1: 5.3)
- CVE-2026-14673: amcheck does not clear untrusted search path (CVSS v3.1: 3.8)
- CVE-2026-14676: pg_stat_statements heap buffer overflow executes arbitrary code (CVSS v3.1: 8.8)
- CVE-2026-14677: 32-bit pltcl and plperl undersize allocations, via integer wraparound (CVSS v3.1: 8.8)
- CVE-2026-14678: pg_trgm picksplit reads past end of buffer (CVSS v3.1: 4.3)
- CVE-2026-14679: Stack buffer overflow in argument match writes 0x0 and 0x1 to server memory (CVSS v3.1: 8.2)
- CVE-2026-14680: Type confusion via "internal" arguments (CVSS v3.1: 8.8)
- CVE-2026-14681: Improper enforcement of GSSAPI encryption when coupled with SSL (CVSS v3.1: 4.2)
- CVE-2026-15741: Expression deparse allows SQL injection via EXTRACT argument (CVSS v3.1: 8.8)
- CVE-2026-15742: fuzzystrmatch writes effectively-arbitrary addresses, via integer wraparound (CVSS v3.1: 8.8)
- CVE-2026-16238: Type confusion in pg_restore_attribute_stats() executes arbitrary code (CVSS v3.1: 8.8)
- CVE-2026-16239: Type confusion in cursor CLOSE + DECLARE executes arbitrary code (CVSS v3.1: 8.8)
- CVE-2026-16241: ECPG integer underflow can crash the client (CVSS v3.1: 3.8)
- CVE-2026-18024: ascii() function reads past end of buffer (CVSS v3.1: 4.3)
- CVE-2026-18408: psql \unrestrict lets superuser of pg_dump origin server execute arbitrary code in psql client (CVSS v3.1: 8.8)
- CVE-2026-19385: pg_dump heap buffer overflow executes arbitrary code (CVSS v3.1: 8.8)
Bug Fixes and Improvements
This update fixes over 110 bugs that were reported in the last several months.
This specific issue only affects PostgreSQL 14, 15, and 16, but we're highlighting it in the release announcement due to its severity:
- Fix a self-deadlock that could occur while replaying WAL generated by an older minor version. This regression, introduced in the previous set of minor releases, could cause a standby server following a primary running an older minor version to become stuck.
The remainder of issues listed below affect PostgreSQL 18. Many of these issues also affect other supported versions of PostgreSQL.
- Fix parallel GIN index
builds to correctly update the table's
reltuplesvalue inpg_class. Previously, a parallel worker could report an uninitialized row count, leavingreltuplesset to a bogus value (includingInfinityorNaN). Such a value can cause autovacuum and autoanalyze to not process the table, and this situation will not self-correct. If you have any tables with GIN indexes, we recommend checking that theirreltuplesvalues look reasonable after updating. See the "Updating" section for how to identify and repair affected tables. - Several fixes to
btree_gist, including forNaNhandling forfloat4/float8that could give wrong answers for columns containingNaNand correct sorting ofbit/bit varyingvalues during index building. You may need to reindexbtree_gistindexes onfloatorbitcolumns after updating. See the "Updating" section. - Fix an integer overflow in
ltreecomparisons.ltreevalues containing more than about 14,653 labels could compare incorrectly, which could present as a corrupt B-tree index. If you useltree, you may need to reindex affected indexes after updating. See the "Updating" section. - Fix partition
pruning
for
RANGE-partitioned tables so theDEFAULTpartition is no longer skipped in cases where it should be scanned. Previously this could cause rows to be missing from query results. - Several fixes for partitioned tables that have foreign-table partitions, including when run-time partition pruning determined that some partitions don't need to be scanned, in-flight requests to foreign servers were not always handled correctly, leading to failures.
- Several fixes to
RETURNINGwithOLDandNEW. - Improve hash join performance when there are multiple join keys and many
NULLvalues. - Several fixes to the planner that could produce wrong query results, including
value IN (array)tests where the array might be empty, andCOUNT()window functions that use anEXCLUDEclause or lackORDER BY. - Add missing checks for the hashability of equality comparisons on container types (arrays, composite types, and ranges). Without these checks, the planner could choose a hash-based plan that then failed at execution with a "could not identify a hash function" error.
- Fix attaching partitions of indexes that are exclusion constraints, which also fixes dump/restore of partitioned exclusion constraints.
- Fix
REINDEX CONCURRENTLYon an index backing a deferred uniqueness constraint, which could cause false constraint-violation reports. - Restore an index-scan optimization that converts a
LIKEor regular-expression exact-match pattern into an equality index condition when the index and expression collations differ. - Several fixes to
jsonpath, including that the@?and@@operators now correctly raise an error for an undefined variable in the path expression. Previously, because these operators cannot supply variable values, an undefined variable was treated as a JSONnullrather than an error, which could also lead to unbounded memory consumption. - Ensure access is blocked to other sessions' temporary tables, which could lead to silently incorrect results.
- Fix
"no empty local buffer available"errors during temporary table access when a largeeffective_io_concurrencysetting could allow a single read stream to consume all local buffers. - Prevent autovacuum from processing databases in the wrong order (lowest-priority first instead of highest).
- Restore
VACUUM's wraparound failsafe mode to use the full shared buffer pool as intended, which had slowed down emergency vacuuming. - Fix possible incorrect decoding of index tuples during GiST and SP-GiST index-only scans, which could emit corrupted data.
- Fix a race condition in conflict detection under
SERIALIZABLEisolation. A conflict could be missed when examining an initially-empty btree index, allowing conflicting transactions to commit and breaking serializability. - Fix WAL logging of operations that clear bits in table visibility maps. This could lead to producing incorrect incremental backups or potentially leaving torn-page writes uncorrected.
- Fix logical
decoding of empty
prepared transactions. A prepared transaction with no decodable changes could
send
COMMIT PREPAREDorROLLBACK PREPAREDto the output plugin with no precedingPREPARE, which breaks replication for the built-in subscriber. - Several fixes to
libpq, including ensuring it drains all pending bytes from the SSL or GSS decryption buffer when reading data, avoiding cases where a client waits for data that has already arrived. - Fix
pg_createsubscriberto clean up objects left on a publisher after failure, including a replication slot. - Fix
pg_restorewith--statisticsor--statistics-onlyso that, when combined with other selective-restore options such as--schema, it restores the expected items, matching the behavior ofpg_dump.
This release also updates time zone data files to tzdata release 2026c, in which Alberta (America/Edmonton) will be on year-round UTC-06 (effectively, permanent DST) beginning in November 2026. This release assumes that their TZ abbreviation will be CST from that time forward (though this could change). It also reflects that Morocco (Africa/Casablanca) will move to permanent UTC+00, without daylight saving transitions, on September 20, 2026.
Updating
All PostgreSQL update releases are cumulative. As with other minor releases,
users are not required to dump and reload their database or use pg_upgrade
in order to apply this update release; you may simply stop PostgreSQL and
update its binaries.
If you have any tables with GIN indexes, we recommend checking their reltuples
values after updating. A prior bug in parallel GIN index builds could have left
reltuples set to a bogus value (including Infinity or NaN) that prevents
autovacuum and autoanalyze from ever processing the table. The following query
lists the tables that have a GIN index, along with their current reltuples:
SELECT DISTINCT t.oid::regclass, t.reltuples
FROM pg_class t
JOIN pg_index i ON t.oid = i.indrelid
JOIN pg_class ic ON i.indexrelid = ic.oid
WHERE t.relhasindex AND ic.relam = 2742;
For any table whose reltuples value looks incorrect, run
ANALYZE on it (or
create another index) to reset the value.
If you use btree_gist, you should reindex btree_gist indexes on float4 or
float8 columns that might contain NaN values, as well as btree_gist indexes
on bit or bit varying columns, for example:
REINDEX INDEX your_index_name;
If you use ltree and have btree indexes over ltree values with very many
labels (more than about 14,653), you should
reindex those
indexes, since they may be corrupt, for example:
REINDEX INDEX your_index_name;
Users who have skipped one or more update releases may need to run additional post-update steps; please see the release notes from earlier versions for details.
For more details, please see the release notes.
A Note on the PostgreSQL 19 Beta
This release marks the third beta release of PostgreSQL 19.
In the spirit of the open source PostgreSQL community, we strongly encourage you to test the new features of PostgreSQL 19 on your systems to help us eliminate bugs and other issues. While we do not advise you to run PostgreSQL 19 Beta 3 in production environments, we encourage you to find ways to run your typical application workloads against this beta release.
Your testing and feedback helps the community ensure that PostgreSQL 19 upholds our standards of delivering a stable, reliable release of the world's most advanced open source relational database. Please read more about our beta testing process and how you can contribute:
https://www.postgresql.org/developer/beta/
Upgrading to PostgreSQL 19 Beta 3
To upgrade to PostgreSQL 19 Beta 3 from an earlier version of PostgreSQL,
you will need to use a strategy similar to upgrading between major versions of
PostgreSQL (e.g. pg_upgrade or pg_dump / pg_restore). For more
information, please visit the documentation section on
upgrading.
Changes Since Beta 2
Fixes and changes in PostgreSQL 19 Beta 3 include:
- Revert
GROUP BY ALL. - Several fixes for the new
FOR PORTION OFtemporal table syntax. - Several fixes for the new logical replication sequence synchronization
feature, including a race involving
REFRESH SEQUENCES. - Fix an "unexpected logical decoding status change" error that could occur when logical decoding is activated concurrently.
- Fix issues around ownership changes for subscriptions.
- Fix wrong query results from
postgres_fdwwhen pushing down an array comparison such asfield = ANY($1)that involves an implicit type coercion. - Fix a crash during foreign-key checks involving a nullable
UNIQUEconstraint. - Fix
pg_plan_adviceparsing of underscores in numeric literals. - Fix a missing
FORMATclause when deparsingJSON_ARRAY(query).
Please see the release notes for a complete list of new and changed features:
https://www.postgresql.org/docs/19/release-19.html
Testing for Bugs & Compatibility
The stability of each PostgreSQL release greatly depends on you, the community, to test the upcoming version with your workloads and testing tools to find bugs and regressions before the general availability of PostgreSQL 19. As this is a Beta, minor changes to database behaviors, feature details, and APIs are still possible. Your feedback and testing will help determine the final tweaks on the new features, so please test in the near future. The quality of user testing helps determine when we can make a final release.
A list of open issues is publicly available in the PostgreSQL wiki. You can report bugs using this form on the PostgreSQL website:
https://www.postgresql.org/account/submitbug/
Links
- Download
- Release Notes
- Security
- Versioning Policy
- Beta Testing Information
- PostgreSQL 19 Beta Release Notes
- PostgreSQL 19 Open Issues
- Submit a Bug
- Donate
If you have corrections or suggestions for this release announcement, please send them to the pgsql-www@lists.postgresql.org public mailing list.
«Сломал Банк Англии» и потерял «Связьинвест»: что можно вынести из опыта Джорджа Сороса
В среду, 12 августа, Джорджу Соросу исполнилось 96 лет. Он известен как миллиардер, самостоятельно заработавший состояние, и один из ведущих филантропов мира. В 2026 году Forbes оценивает его состояние в $7,5 млрд.
Для тех, кто строит торговые системы и ищет устойчивые подходы к рынку, история Сороса — это не просто биография. Это кейс о том, как системное мышление, философская основа и готовность действовать вопреки консенсусу могут создавать аномальные результаты. Разбираем, что можно извлечь из его опыта.
Читать далееAgent 64: Spies Never Die review – licence to kill time like the 90s never ended
PC; Replicant D6
This GoldenEye 007-styled retro shooter is an agreeable set of straightahead shooting missions that hark back to a simpler time, with simpler controls
With 007 First Light, 2026 has already given us the best James Bond game in more than three decades, since the release of Rare’s N64 classic GoldenEye 007. But what about those players whose loyalties lie not with the super-spy himself but the particulars of that 90s interpretation of his adventures?
Enter Agent 64: Spies Never Die. Made almost entirely by a single developer (who goes by the very cloak-and-dagger alias Replicant D6), its production values couldn’t be much further from the blockbuster First Light. There’s no voice acting, with all dialogue instead delivered via subtitles, while each location on your globe-trotting adventure is depicted by a repeating texture wallpapered over scenery so blocky you can count the individual polygons.
Continue reading...US unemployment claims rise but remain at healthy level
Bugging devices: the astounding ways insects are talking and listening all around us
From spiders tuning their silken listening tools, to the beetle’s heady perfume, hundreds of conversations are happening in languages we don’t understand
All around us, every day, hundreds of conversations are taking place in languages we don’t understand, or even, in most cases, notice. Birdsong is one you would probably recognise, and you may have heard of honeybee “waggle dances”. But what about the spider tuning their web to listen for prey, or a waft of arousing perfumes emitted by a passing beetle? At any moment you may be walking right through a whole critical dialogue – and have absolutely no idea.
To delve into the world of invertebrate communication about food gathering, coordinating nest building, defending a territory, attracting a mate, and babycare is to realise just how smart and sophisticated they are. Spiders, for example, can “hear” vibrations on the silken strands of their web using sensors on their legs. With their legs and body weight, the spiders will adjust the silk strings, as we would a guitar string, to change its pitch, pulling, loosening or reconnecting strands to adjust how it sounds. In fine-tuning the silk they can know just who is walking on it – a possible mate to woo, a fly for feasting on or maybe a larger bug on the lookout for prey.
Continue reading...‘At any moment I might fade out’: archives reveal fearless era of land speed record breakers
Unseen material gives insight into how death-defying attempts by drivers made impact on public imagination
With the Bonneville salt flats in Utah echoing to the sound of a new land speed record achieved by a British team, the timing could not be more apposite. The historical legacy of the challenge of pushing the envelope of pace is being dusted down and made public in the form of an extraordinary archive from more than a century of record-breaking machines and their singularly driven pilots.
On Tuesday, JCB’s Hydromax car took the hydrogen-powered International Automobile Federation (FIA) land speed record, reaching 406.320mph at Bonneville.
Continue reading...‘Absolutely unacceptable’: Japan condemns visit by Vladimir Putin to disputed Kuril Islands
Japanese PM says islands seized by Soviets late in second world war are part of her country under international law
The Japanese government has reacted angrily to Vladimir Putin’s first visit to the disputed Kuril Islands, which are claimed by both Tokyo and Moscow.
Japan’s foreign minister, Toshimitsu Motegi, summoned the Russian ambassador Nikolay Nozdrev, while the prime minister, Sanae Takaichi, called Putin’s actions “absolutely unacceptable”.
Continue reading...Surveillance company Flock moves to increase oversight after police misuse
Democrats pressure Trump administration to rescue 10 other Americans detained in Russia
‘Absolutely unacceptable’: Japan condemns visit by Vladimir Putin to disputed Kuril Islands
Japanese PM says islands seized by Soviets late in second world war are part of her country under international law
The Japanese government has reacted angrily to Vladimir Putin’s first visit to the disputed Kuril Islands, which are claimed by both Tokyo and Moscow.
Japan’s foreign minister, Toshimitsu Motegi, summoned the Russian ambassador Nikolay Nozdrev, while the prime minister, Sanae Takaichi, called Putin’s actions “absolutely unacceptable”.
Continue reading...Как я настроил общий AI-harness для команды разработки на Pi
Как дать AI-агенту контекст нескольких репозиториев и доступ к рабочим инструментам? Показываю setup на Pi: отдельный Git-репозиторий с описаниями проектов и документацией, skills для CLI-интеграций, progressive disclosure через references и тот же harness в терминале и Telegram.
Читать далееПеревод книги целиком: как я создал конвейер с помощью LLM
Я потратил пару недель на создание конвейера с открытым исходным кодом, который переводит книги целиком с помощью платных подписок на LLM. Поддерживаются Claude Code, Antigravity CLI и Codex. В статье описаны принципы построения конвейера для максимального качества перевода, а также тонкости по лимитам, скорости, цензуре и т.п. Проект BookTrans на GitHub.
Читать далееPrimary voters send mixed signals about what they want from the Democratic Party
Primary results this month in Michigan, Minnesota and Wisconsin paint a muddled picture of just how far to the left Democratic voters want their candidates to be come this fall's midterm elections.
Total solar eclipse dipped parts of Europe into complete darkness. Here's what they saw.
The Taliban are vile. Democracies must still engage with them
Under the Taliban, Afghanistan is, for once, peaceful
Уязвим не nginx, а две строки вашего конфига. Разбираю CVE-2026-42945 на живом стенде
Критическая дыра в nginx: 9.2 по CVSS, восемнадцать лет в коде, нашёл её ИИ-агент за шесть часов. В новостях к этому прилагают 5.7 миллиона уязвимых серверов, а сканы реальных конфигов находят ноль из 1465 и один из 35633. Поднял стенд, чтобы понять, кто прав. Заодно выяснилось, что граница проходит не там, где её рисуют: трейлинговый знак вопроса безопасен, промежуточный rewrite гасит флаг, а именованный захват спасает не всегда.
Читать далее