TheChek News Flow

A foreign driver with an unbelievably high blood alcohol content (BAC), one that would routinely kill most people, decided to drive his car at an unbelievable 220 kilometers per hour (137 miles per hour) on the German autobahn. His passenger had an even higher blood alcohol content.

The man was traveling on the A67 in southern Hesse and was so drunk that most medical literature indicates that he should be dead or in a coma.

One witness followed the car and called the police after noticing the man zigzagging between lanes and driving in an erratic manner late one evening.

Officers stopped the car, which had foreign license plates, and checked both occupants of the vehicle in the parking lot of the A5 autobahn.

Incredibly, the passenger of the vehicle had an even higher blood alcohol level.

“A breathalyzer test carried out on the spot showed a blood alcohol level of 4.16 per mille for the driver and no blood alcohol level for the passenger, as the device switches off when the blood alcohol level reaches 5 per mille!” the police wrote in a statement, including an exclamation mark to indicate their shock.

The driver’s license was immediately confiscated, and he was taken into custody.

German media outlet Welt wrote that “The fact that there were no injuries or deaths was likely due to mere coincidence.”

In Germany, drivers are not permitted to drive with a blood alcohol level over 0.5 per mille or higher. At 1.1 per mille, it is a criminal offense, as the law states there is an “absolute inability to drive.”

“A blood alcohol level above 4 per mille poses an acute risk to life,” the ADAC, the German automobile club, the largest in Europe, writes. “Important protective reflexes are lost, those affected fall into a coma, and can suffer shock with progressive circulatory failure, even complete respiratory and cardiac arrest.”

A blood alcohol concentration of more than 5 per mille almost always leads to death.

The fact that the passenger had a blood alcohol level above 5 and did not die is remarkable in itself.

There is speculation about the identity of the foreigners; however, the German media did not indicate their nationality or what country the vehicle was registered under.

Although the 5 per mille level seems remarkable, it is far from the highest blood alcohol level ever recorded. One poorly sourced claim notes a Polish driver recorded an incredible 1.480 percent level, which was recorded after he crashed his car. Despite the level, he survived, only later to die from the injuries from the crash — and not the alcohol. However, the Guinness Book of World Records claims the highest level was 1.374 percent, also set by a Polish man who survived but suffered serious damage to his internal organs.

Experts reveal how monitoring your blood sugar, in moderation, may help avoid spikes linked to intense hunger, cravings and even poor sleep

Fewer food cravings, better sleep and clearer skin – these can all be a result of stable blood sugar levels.

A growing number of social media users are promoting the benefits of blood sugar monitoring to avoid experiencing blood sugar spikes and their adverse effects.

For people with diabetes, a chronic disease characterised by high blood sugar (glucose) levels, keeping their blood sugar levels near normal helps manage the condition and prevent potentially serious complications.

This is also useful for non-diabetics, some experts say.

Glucose, a simple sugar, is the body’s primary fuel source. It comes mainly from the carbohydrates that our digestive system breaks down in the food and drinks we consume. The glucose is then absorbed into the bloodstream and carried to the body’s cells.

“It’s our currency in energy metabolism,” says German nutritional scientist Dr Brigitte Baeuerlein.

A person’s blood sugar level is the amount of glucose in a specific volume of blood, measured in milligrams per decilitre. It typically rises after eating, when the small intestine releases glucose into the bloodstream. Insulin, a hormone released by the pancreas, then helps cells absorb it for energy or storage.

“It docks onto the cells that need energy at the moment,” Baeuerlein says.

Blood sugar levels return to normal within a couple of hours in healthy individuals. In diabetics, however, insufficient production or function of insulin results in chronically high levels.

Stress can cause spikes in blood sugar, too – for instance, Baeuerlein notes, when you are frantically trying to find a parking space during working hours in a big city centre.

Your liver, in a “fight or flight” response, releases glucose into the bloodstream to provide a quick energy boost to deal with the perceived threat.

There are several ways to test your blood sugar. The conventional method is with a blood glucose meter: you insert a test strip into the meter, prick your fingertip with a lancet, and place a small drop of blood onto the strip. After a few seconds, the meter displays your blood glucose level on a screen.

A test often used by medical professionals to diagnose diabetes is a haemoglobin A1c (HbA1c) test, which provides the average blood glucose level over two to three months.

Another method is a continuous glucose monitor (CGM), a device that measures glucose levels in the interstitial fluid of your subcutaneous tissue, which reflects blood glucose levels with a slight delay.

CGMs use tiny sensor wires inserted under the skin of your upper arm or abdomen with an applicator. Readings are displayed on a receiver or your smartphone.

A blood sugar spike is a sharp, marked rise in the amount of glucose in your blood, usually occurring after eating quickly digested, short-chain carbohydrates – sugars, basically, such as those in white-flour products, sweets and fruit juices.

These spikes can trigger cravings for the very foods that caused them: if too much insulin is released to move glucose out of the bloodstream and into cells, blood sugar can rapidly drop, or “crash”, leading to intense hunger.

But Bernhard Kulzer, head of the research institute of the Diabetes-Academy Bad Mergentheim in Germany, says blood sugar measurements are not really needed by non-diabetics, pointing out that a healthy body regulates blood sugar levels itself.

The pancreas releases insulin if it is elevated, or glucagon – a hormone that stimulates the liver to release stored glucose – if it is low.

Endurance athletes are an exception, Kulzer says. Monitoring their blood sugar levels can help keep them from “hitting the wall”, a sudden loss of energy due to depletion of glycogen – the stored form of glucose – in the liver and skeletal muscles. If they know their level is low, they can fuel up on fast carbs.

According to Baeuerlein, people with a family history of diabetes might also want to check their blood sugar level to see how it reacts to certain foods. “Doing it for a few weeks is enough,” she says.

One plus of measuring blood sugar levels, says Kulzer, is that you get “biofeedback”. In other words, you eat something and then see straight away how it affects your blood sugar. Or you exercise and then see how the level drops. This can help you if you want to lose weight by changing your habits, he says.

However, “the readings are just a diagnostic tool and of no use unless you draw conclusions from them”.

This requires knowledge of how to interpret them, so “without instruction and training, they make little sense”, he says.

If you do use a blood glucose measuring device, Kulzer advises that it be temporary.

“There’s no point in wearing a sensor all of the time, especially considering that it’s not cost-free for healthy people,” he says.

Baeuerlein, for her part, warns that constantly worrying about what you eat could become compulsive.

Generalisations need to be treated with caution, she says, since “everyone, whether they’re healthy or have diabetes, reacts differently to sugar in food”. Their reaction depends in part on cell fitness – on their age, in other words. Musculature and exercise play a role, too.

If you want to lose weight, Kulzer says, you have to realise that your calorie intake is decisive. So to shed a few kilograms, you must consume fewer calories than your body burns through your metabolism.

A balanced diet is always a wise choice, says Baeuerlein. That means plenty of fruit, vegetables and dietary fibre, good oils, protein and not too many empty carbs whose calories come with little nutritional value. This will help prevent glucose spikes and food cravings.

She also advises against consuming isolated, dissolved sugars such as those found in soft drinks, juices and smoothies. It is also important to identify sugar sources in your diet and not add sugar to your coffee, for example. This can help you permanently reprogramme your food preferences.

A final tip is to begin a meal by eating vegetables or a salad.

“Everything that falls onto this foundation will then be absorbed into the bloodstream more slowly than if you start by eating noodles.”

Google is rumoured to have completed its first AI-enabled glasses, but has yet to decide on a contract manufacturer

Taiwan’s Quanta and HTC are seen as possible contract manufacturers for Google’s imminent AI glasses, as such products are expected to emerge as essential mobile devices that will drive rapid expansion across the industry.

Google is rumoured to have completed development of its first artificial intelligence-enabled glasses, which are reportedly to be made in Taiwan. But the US search giant has yet to decide on a contract manufacturer. Quanta was in charge of prototype design and is a strong contender, but market rumours suggest HTC may produce the devices.

HTC recently launched its first AI glasses model, the Vive Eagle, which it said were 100 per cent manufactured in Taiwan. The company also recently transferred some R&D personnel with skills in augmented reality and virtual reality (AR and VR), as well as non-exclusive licensed patents, to Google for US$250 million, leaving room for future collaboration between the two companies.

HTC previously sold its Powered by HTC smartphone division to Google for US$1.1 billion in a one-time deal, and some industry sources said it would be less than ideal for HTC if the recent US$250 million transaction is also a one-off.

Although contract manufacturing is no longer HTC’s core business, the company’s production of its own AI glasses shows that it has the capability. If HTC did manage to secure Google’s manufacturing contract, it would be a major boon to the company in improving capacity utilisation at its plant in Taoyuan, near Taipei, and expanding operations.

HTC’s Vive Eagle is the first set of AI glasses to support traditional Chinese voice control and is the world’s first branded wearable device to adopt an open AI architecture, allowing users to personalise their AI assistant experience by choosing from among large language models such as Google Gemini and OpenAI’s GPT.

On August 21, HTC senior vice-president Charles Huang said that the company would continue to expand AI functions and application scenarios through software updates.

The rapid development of generative AI, as well as mature AR and VR, cloud computing, and edge-computing technologies, had brought diverse options and innovations to human-computer interaction, Huang said.

Global smart glasses shipments grew by 210 per cent in 2024 from a year earlier, data showed, with a further increase of 60 per cent projected for 2025 and a compound annual growth rate above 60 per cent through 2029, Huang said.

Industry findings and trending online comments help explain how Chinese brands are steadily outpacing foreign firms, as their store closures make headlines

The world’s largest jewellery company by volume is facing a wave of store closures across China – a move that serves to illustrate a broader retreat by foreign brands as consumer preferences shift towards practical products that offer more value for the money.

Danish jewellery brand Pandora, known for its fast-paced product launches and global sales of more than 100 million pieces a year, announced in its second-quarter earnings report on August 15 that it would expand its original plan to close 50 stores in China this year to 100 stores.

According to Pandora’s annual financial reports, its sales in China peaked in 2019 at 1.97 billion Danish kroner (US$305.73 million). But since the global pandemic, sales in China have steadily declined, dropping to 1.126 billion kroner in 2021, 737 million kroner in 2022, and 564.2 million kroner in 2023. Last year, Pandora’s revenue in China plummeted to just 416 million kroner, less than a quarter of the 2019 peak.

“Pandora is more about aesthetics and versatility for me; it’s not something to buy for value retention,” said Vicky Wang, a Pandora enthusiast from Jiangsu province.

Over the past three months, she spent 7,500 yuan (US$1,044) on a half-dozen charm bracelets but is now trying to sell what she considers “impulse purchases” on China’s second-hand trading platform Xianyu.

The closures by Pandora, which did not immediately respond to an emailed request for comment, are not isolated incidents among foreign retailers struggling in China.

According to the 2025 China Shopper Report released in June by Bain & Company and Worldpanel, Chinese domestic brands have steadily eroded the market share of foreign brands, with local brands accounting for 76 per cent of the market by 2024. That was up from 75 per cent in 2023, 74 per cent in 2022, and 73 per cent in 2021. Back in 2012, it was 66 per cent.

Foreign firms closing stores in China include fast-fashion clothing brands such as GU, owned by Japanese multinational Fast Retailing Group, and Spain-based Zara, as well as beauty and skincare brands such as Australia’s Aesop and Japan’s Decorte. Retail chains such as US-based Walmart and Japan-based Aeon have also announced withdrawals from certain regional markets.

Analysts say these types of retreats can be partly explained by subdued consumer spending amid deflationary pressure, coupled with uncertainties surrounding China’s economic outlook.

“Over the next decade or beyond, we will be in an era of low-price retail; this is a major trend,” said Chen Liping, a professor at Capital University of Economics and Business, in comments posted earlier this month to the social media account of a Chinese retail industry platform. “For some companies, it’s no longer just about operational challenges, but about whether they can survive.”

China’s consumer price index in July was flat from a year earlier, and producer deflation continued, indicating persistent weak demand in the world’s second-largest economy amid trade disruptions.

Wang said she was initially drawn to Pandora’s charm bracelets due to their marketing concept of “one bead, one story”, allowing her to create personalised memories. However, her attention, like that of many young consumers, has been shifting to safe-haven investments.

“I started to lose interest because, compared with gold, Pandora doesn’t hold its value [in terms of resale],” Wang explained. “Many people around me also didn’t understand why I spent so much on silver jewellery.”

Meanwhile, popular Japanese lifestyle brand Muji, which focuses on selling a minimalist lifestyle to the middle class, has made headlines following its string of store closures in China. The Shanghai-based China Business Network recently reported that multiple Muji locations, including in the cities of Beijing, Shanghai, Suzhou and Changsha, as well as in Zhejiang province, have shut down in recent months.

On Rednote, a popular Chinese social media platform, the hashtag “Muji pitfalls” has garnered 1,800 posts and 334,100 views in the past couple of years. The platform’s users have criticised the brand’s products for failing to meet their quality expectations and for offering what some contend are overpriced items that lack practical value for daily use.

One comment from May, liked more than 1,000 times, lamented the 68 yuan (US$9.50) cost of a cardholder, saying the price would have been more fitting for a backpack. “With that kind of money, I’d rather buy two cat bowls - they’d be far more useful,” the person said.

Muji did not immediately reply to an emailed request for comment, but it was quoted by the China Business Network earlier this month as saying that selective store closures were routine adjustments to improve operational efficiency.

Washington’s strict blacklist rule amid national security concerns has forced Beijing to become self-sufficient with ‘incredible’ results

When Sama Bilbao y León, director general of the World Nuclear Association – a global industry body based in London – toured China’s nuclear facilities this summer, she was left speechless.

“I couldn’t close my mouth in amazement,” she admitted, stunned by China’s advanced capabilities and “incredible” industrial scale.

This awe-inspiring progress exists despite stringent US sanctions – including the 2019 blacklisting of China General Nuclear Power Group (CGN) and the recent suspension of equipment licences – imposed over national security concerns.

Ironically, these very restrictions have backfired spectacularly. Instead of crippling China’s nuclear ambitions, the sanctions have forced Beijing to develop a fully self-sufficient nuclear ecosystem, achieving near-total domestic equipment production and rapid reactor deployment.

Now, as the US and Europe struggle with costly delays and atrophied supply chains in their own nuclear expansions, China’s sanctioned industry has become an unattainable benchmark of efficiency – and its exclusion threatens to stall the West’s own atomic energy revival.

“The current geopolitical situation at all levels is incredibly complex. But our position has always been that the nuclear industry is global, and we would like to see countries working together and sharing best practices and lessons learned,” said Bilbao y León.

The industrial leader, originally from Spain, has been calling on member countries to strike a balance between global cooperation and local participation to ramp up the supply chain.

The world is in a heated race to expand nuclear power capacity, with major countries – particularly the US and China – announcing ambitious goals recently, driven by concerns about climate change mitigation and energy security.

“Many countries are looking to triple or even quadruple their nuclear capacity, but achieving these goals will require rebuilding and reintegrating supply chains and industrial capabilities,” Bilbao y León told the South China Morning Post in an exclusive interview at the end of July.

“Everybody is taking note of what is happening in China,” she added.

History shows that supply chain bottlenecks have already stifled mega nuclear projects in developed countries. However, as geopolitical tensions intensify, Chinese technologies and suppliers are increasingly being scrutinised or even blocked, by the West.

Kevin Tu, a non-resident fellow at the Centre on Global Energy Policy at Columbia University, said that, ideally, other countries could benefit from collaborating with China across multiple areas – such as plant design, construction, equipment procurement and talent training – as Beijing had undertaken one of the fastest nuclear power build-outs in history in recent years.

But the reality is far more complicated. Given the sector’s highly sensitive nature, it “will inevitably become politicised”, making international cooperation with China difficult, Tu noted.

Philip Andrews-Speed, senior research fellow at the Oxford Institute for Energy Studies, wrote in an email response to the Post that, “I think it is unlikely that any Western country – Canada, the US, Japan, South Korea, the UK or most EU member states – will choose a Chinese company as a vendor for a nuclear reactor for as long as distrust remains at current or higher levels.”

For example, last May, Hungary and China signed a nuclear energy cooperation agreement, encouraging capable Chinese enterprises to invest in Hungary. But Andrews-Speed pointed out that Hungary’s nuclear sector contracts have actually gone to Russia, their long-term nuclear technology supplier.

In 2022, China’s state-owned group CGN – which held a 20 per cent stake in the Sizewell C nuclear power plant project in Suffolk, eastern England – was forced to withdraw by the British government amid national security concerns.

Despite breaking ground on its first reactor almost 40 years after the US pioneered the tech, China is now building far more fission power plants than any other country. Around 10 new reactors have been approved by the Chinese government each year since 2022, putting China on track to overtake the United States and become the world’s largest generator of nuclear power by 2030.

According to data from the World Nuclear Association, seven of the 10 most recent projects to begin construction since last July belong to China, while the other three are from Russia, South Korea and Pakistan.

Thanks to the rapid development over the past three to four decades, “the entire industrial chain has been well-established”, said Zhao Chengkun, former executive deputy director of the expert committee of the China Nuclear Energy Association, with around 90 per cent of equipment domestically produced.

Meanwhile, China has established a robust system for cultivating a stable pool of skilled personnel, with a significant demand for professionals in the industry and related subjects in higher education, Zhao added.

Andrews-Speed, from the Oxford Institute for Energy Studies, echoed Zhao.

“China has built an almost complete supply chain for nuclear power, along with the skilled labour force needed to build and operate the world’s fastest-growing fleet of plants,” he said.

However, the story is different when it comes to countries like the US and France – which once dominated nuclear technology development. Due to historical reasons such as public opposition, rising costs and regulatory challenges, construction of large-scale nuclear power plants in these countries has stagnated for decades.

“US and most European countries have not consistently built nuclear power plants, particularly gigawatt-scale pressurised water reactors, for many years. As a result, their industrial supply chains and talent bases have weakened considerably over time,” Tu said.

From a pragmatic standpoint, developing a robust supply chain takes time and necessitates international cooperation. Partnering with Chinese manufacturers represented a viable option, Tu said, “if political trust is rebuilt and geopolitical tensions are eased”.

The progress made by China and the West in developing third-generation nuclear power technology reflects precisely the impact that supply chain maturity can have on project construction.

The Flamanville Unit 3 reactor in northern France is owned and operated by the French company EDF. The project began in 2007, with commercial operations initially scheduled to start in 2012.

However, it has been repeatedly delayed due to various factors. In 2023, for example, work was halted for several months due to issues with welds in the main secondary system. The reactor was finally connected to the grid by the end of 2024, and the project has cost at least five times more than originally anticipated.

The Plant Vogtle Unit 4 in the state of Georgia – one of the first new nuclear units to come online in the US in over 30 years – is based on Westinghouse’s third-generation plus AP1000 reactor technology. Yet the project also faced an overall delay of around seven years and cost overruns, partly due to manufacturing defects and delays in delivering critical components, such as the main pump.

In contrast, the Fuqing 5 unit in southeastern China’s Fujian province – the first Hualong One model to enter commercial operation globally – set a record for construction time. It began construction in 2015 and entered commercial operation in January 2021 at a cost 20 to 30 per cent lower than that of its foreign competitors.

Although the Hualong One design was modelled on foreign technology, this milestone project has greatly boosted China’s self-reliance in terms of its supply chain and workforce.

According to a 2021 report by Communist Party mouthpiece People’s Daily, the project involved over 5,300 suppliers nationwide, with all core equipment produced domestically. The project also helped China to train professionals in the core technologies and construction of third-generation nuclear power plants.

China has exported at least 15 nuclear power units and research facilities worldwide. The K2 unit at the Karachi Nuclear Power Plant in Pakistan, which started operating in May 2021, is the first overseas nuclear power unit to use the Hualong One model.

In an article published in the prestigious journal Nature on July 28, several researchers from US institutes, including Harvard Kennedy School, examined how China had succeeded in avoiding the “cost escalation curse” faced by the nuclear industry through regulations and supply chain efforts.

China has been researching and developing its own reactor design based on French technology for two decades and has been producing simple components domestically for an even longer time.

Some Chinese-made nuclear components, such as tubes, ring cranes and charging pumps, cost half as much as their imported equivalents, according to the study.

In some cases, however, importing nuclear technologies is a more strategic choice, the author noted, citing the example of France’s deviation from the tried-and-tested US reactor design in favour of a national version in the 1980s, which hindered standardisation and caused construction costs to rise.

“As countries rush to expand nuclear capacities, they must combine affordability with safety, scalability, investor confidence and public trust. Without this, nuclear power will remain an expensive bet the world can ill afford,”the Nature article concluded.

Nuclear power is undergoing a global resurgence, with more than 30 countries pledging to triple global capacity by mid-century. Technology giants, including Amazon, Google and Microsoft, are also investing in nuclear power to fuel their energy-intensive AI data centres and reduce carbon emissions.

In May, the new German government dropped its long-held opposition to nuclear power. Denmark is also considering lifting its 40-year ban on atomic power to enhance its energy security amid the Europe-wide shift.

Also in May, US President Donald Trump signed executive orders to accelerate the construction of additional nuclear capacity with the goal of quadrupling the country’s installed capacity to 400 gigawatts (GW) by 2050.

The importance of nuclear power was also demonstrated in early July when Trump’s signature budget bill, officially known as the One Big Beautiful Bill, marked a striking shift towards fossil fuels and nuclear power.

The latest data from the World Nuclear Association shows that China currently has 58 operable reactors with a capacity of around 57GW. But according to a June report by the China Nuclear Energy Association, the country is set to build dozens of new reactors to raise its installed capacity to 200GW by the end of the next decade.

“Many countries have realised that there is a dire need for abundant, affordable, clean energy and electricity that is available 24/7,” said Bilbao y León.

For many countries, it is simply not possible to meet carbon emission goals in a cost-effective, timely manner without doubling down on nuclear power, and the quest for energy security and independence is increasingly urgent, she explained.

On top of that, as the world rapidly becomes more electrified, affordable new energy sources are needed to power industrial development and hi-tech sectors such as AI.

Andrews-Speed believes the issue for Western nuclear development is the ability to deliver projects on time and on budget – a management problem that will not easily be solved with China’s help only in the form of equipment and workforce. “But most Western governments would not allow a Chinese company to build one of their own reactor designs for them,” he said.

Nevertheless, he added that once Western companies get into the practice of building several reactors of the same design, they should become more efficient.

Les enquêteurs soupçonnent ce jeune migrant sans abri d’avoir tué quatre hommes en seize jours en raison de leur orientation sexuelle. Mais l’identité de deux victimes, des compagnons d’infortune du suspect, jette le trouble sur le déclenchement de cette folie meurtrière.

Quatre corps, un suspect, une hypothèse tenace de crimes homophobes, mais un mobile encore incertain. Onze jours après la découverte de quatre cadavres dans la Seine à Choisy-le-Roi (Val-de-Marne), le 13 août, un sans-abri d’une vingtaine d’années a été mis en examen, dimanche 24 août, pour « meurtres en concours » au préjudice des quatre victimes. Cette infraction rare, qui désigne les crimes sériels, ouvre une perspective vertigineuse : celle d’un tueur en série qui aurait commis quatre meurtres en seulement seize jours dans le même secteur.

L’identité de ce migrant désocialisé et précaire, tout comme ses motivations, demeurent confuses. En garde à vue, il s’est présenté comme un Algérien de 24 ans, né à Oran et prénommé « Ahmed », installé irrégulièrement depuis trois ans en France, où il travaillerait occasionnellement sur les marchés et dans le bâtiment. Mais les enquêteurs ont trouvé dans son squat un document en arabe portant sa photo laissant supposer qu’il s’appellerait en réalité Monji H. et serait de nationalité tunisienne, une identité qui a été confirmée par un de ses proches, explique au Monde une source proche de l’enquête.

Concernant les faits qui lui sont reprochés et son possible mobile, le suspect ne s’est pas montré plus disert. Durant ses quatre-vingt-seize heures de garde à vue, réalisée avec l’aide d’un interprète en raison de son niveau de français rudimentaire, il a nié les faits ou gardé le silence face aux éléments à charge accablants qui lui étaient présentés. Dans son dernier communiqué, publié dimanche, le parquet de Créteil n’avance d’ailleurs aucune hypothèse permettant d’expliquer cette folie meurtrière et ne retient pas, à ce stade, la circonstance aggravante qu’elle ait pu être déclenchée en raison de l’orientation sexuelle des victimes.

Deux corps partiellement dénudés

C’est pourtant bien aujourd’hui une des principales pistes suivies par les enquêteurs de la brigade criminelle de Paris. « Un mobile sexuel ne peut être exclu à ce stade », explique une source. Les quatre corps ont en effet été repêchés à proximité d’un endroit connu pour être un lieu de rencontres homosexuelles, et le suspect squattait dans un ancien local technique désaffecté servant à la gestion des eaux de la Seine, situé à quelques dizaines de mètres de là.

Mais ce sont surtout les premières constatations réalisées sur les victimes, portées disparues entre le 26 juillet et le 11 août, qui ont donné de l’épaisseur à ce mobile. Leur identification a été rendue particulièrement complexe en raison de la dégradation de certains corps. La victime la plus récente a été la première identifiée : il s’agit d’un Français de 48 ans domicilié à Créteil, dont les investigations ont établi qu’il était homosexuel et avait fréquenté ce lieu de rencontres le soir de sa disparition, le 11 août. Il a été repêché deux jours plus tard vêtu d’un tee-shirt et le bas de son corps « dénudé », précise le parquet.

Selon les informations du Monde, son ADN a été retrouvé mélangé avec celui du suspect sur le pantalon que portait ce dernier le jour du meurtre, au niveau du pubis et des cuisses, laissant entrevoir la possibilité qu’un rapport sexuel ait eu lieu entre le tueur et sa victime avant le crime. Des traces de strangulation, un mode opératoire susceptible de traduire une pulsion sexuelle, ont en outre été découvertes sur son corps, ainsi que sur une, et peut-être deux autres victimes. Un autre cadavre, dont le parquet ne précise pas l’identité, a quant à lui été découvert le pantalon « baissé au niveau des chevilles ».

« Pardon, pardon, Monsieur »

L’avant-dernière victime à avoir été tuée est un Algérien de 21 ans habitant à Choisy-le-Roi. Le 4 août, tandis qu’il se trouvait sur les bords de Seine à proximité du squat et discutait au téléphone avec un collègue, peu avant 22 heures, ce dernier l’avait entendu dire en arabe « pardon, pardon, Monsieur » avant que la conversation ne soit brusquement interrompue, explique une source au Monde, confirmant une information du Parisien. Sa disparition avait été signalée par sa femme, avec laquelle il ne vivait plus, trois jours plus tard.

A priori, le suspect ne connaissait aucune de ces deux victimes, les deux dernières à avoir été tuées. En raison des constatations réalisées sur les corps et de la proximité de leur découverte avec un lieu de rencontres gay, les enquêteurs le soupçonnent de les avoir tuées en raison de leur orientation sexuelle, réelle ou supposée, et l’ont abondamment questionné sur ses propres pratiques.

Aurait-il agi par pure homophobie, ou plutôt en raison de son homosexualité « honteuse », accompagnée d’un sentiment de culpabilité ? L’hypothèse d’un conflit intérieur entre son orientation sexuelle et sa pratique de l’islam fait en tout cas partie des pistes de travail. Elle a notamment été nourrie par les déclarations d’un de ses proches, placé en garde à vue le 21 août avant d’être libéré sans poursuite samedi soir, selon lesquelles le suspect s’était récemment découvert une nouvelle ferveur religieuse et reprochait à ses amis de ne pas prier suffisamment.

Un mobile trouble

Mais là où le mobile de ce possible tueur en série au profil inhabituel se trouble, c’est que ses deux premières victimes faisaient partie de ses « compagnons de galère », explique une source proche de l’enquête. Il s’agit de deux migrants précaires comme lui, qui fréquentaient régulièrement son squat. L’un, Abdellah M., un Algérien de 21 ans, n’avait plus donné signe de vie depuis le 26 juillet ; le second, Amir B., un Tunisien de 26 ans, depuis le 31 juillet.

A ce stade des investigations, la question de leur orientation sexuelle n’a pu être tranchée de manière certaine. Pourquoi le suspect aurait-il commencé par tuer ses deux compagnons d’infortune pour ensuite s’attaquer à deux inconnus ? A-t-il eu des rapports sexuels avec ses quatre victimes ? A-t-il tué ses deux amis pour des raisons crapuleuses, avant d’être pris d’une frénésie criminelle ?

Une chose est sûre : les éléments matériels qui le relient à chacune des quatre victimes sont accablants. L’ADN du Français de 48 ans a été retrouvée sur son pantalon, et le suspect a utilisé deux cartes SIM et une carte bancaire appartenant aux trois autres victimes juste après leur disparition. Le 1er août, la vidéosurveillance d’une boucherie a ainsi permis d’établir qu’il avait utilisé la carte bancaire de son ami Amir B., disparu la veille. Rarement, dans les annales judiciaires, aura-t-on eu affaire à un tueur en série aussi peu précautionneux.

Ce sont d’ailleurs ces erreurs grossières qui ont permis aux enquêteurs de l’identifier. Plusieurs éléments de preuve étaient même en leur possession avant la découverte des corps. Le 5 août, une semaine avant que les cadavres ne soient repêchés dans la Seine, le suspect avait en effet déjà été interpellé après la découverte, lors d’une opération de contrôle dans son squat, de cartes bancaires, de téléphones et de cartes SIM ne lui appartenant pas.

Recel de vol

Déjà connu de la justice pour un vol avec dégradation dans un véhicule commis en janvier, pour lequel il devait être jugé en septembre, le suspect avait été laissé libre à l’issue de ce contrôle de police du 5 août, et s’était vu remettre une convocation pour une ordonnance pénale pour recel de vol. A cette date, les policiers ignoraient que les deux téléphones retrouvés dans son squat appartenaient à la première victime, son ami Abdallah M., et à la troisième, le jeune Algérien de 21 ans tué la veille, dont la disparition n’avait pas été signalée.

Les éléments issus de cette première procédure, ainsi que l’exploitation de la vidéosurveillance et de la téléphonie ont conduit les enquêteurs à interpeller le suspect, le 20 août, une semaine après la découverte des corps, dans un centre de rétention administrative, où il avait entre-temps été placé en vue de son éloignement du territoire.

Confronté en garde à vue aux photos des quatre victimes, il a assuré n’en connaître aucune. Il a pourtant été établi qu’il se rendait régulièrement au centre Emmaüs de Choisy-le-Roi en compagnie de deux d’entre elles : Amir B., qui se serait installé dans son squat en février, et Abdellah M., avec lequel il apparaît sur des photos et des vidéos retrouvées dans le téléphone de ce dernier.

A l’issue de sa mise en examen, le suspect doit être présenté devant le juge des libertés et de la détention en vue de son placement en détention provisoire, requis par le parquet. Sollicité par Le Monde, son avocat, Antoine Ory, a simplement fait savoir que son client avait « fait usage de son droit au silence lors de sa garde à vue et devant le magistrat instructeur » et qu’il « s’expliquera ultérieurement lors de l’information judiciaire ».

The progress of advanced AI models seems to be decelerating after a period of rapid advancement and high expectations.

Many companies are hesitant to fully adopt AI due to concerns about data security and the reliability of AI-driven decisions.

Slower AI advancement might benefit companies by providing more time to integrate AI and adapt it to specific business needs.

The advance of cutting-edge AI is showing signs of slowing. For many companies looking to harness the technology, that wouldn’t be a terrible thing.

Excitement about AI reached feverish levels at the end of 2022 with the release of OpenAI’s ChatGPT and has stayed red-hot since. A regular cadence of more impressive large language models from startups and big tech companies alike has kept the party going, lifting stocks—including shares of AI-chip giant Nvidia NVDA 1.72%increase; green up pointing triangle—to lofty heights.

Nearly three years later, it looks increasingly as though those models are plateauing.

This summer, Meta Platforms META 2.12%increase; green up pointing triangle delayed the rollout of the next iteration of its flagship AI model, called Llama 4 Behemoth, because engineers were struggling to significantly improve it.

The latest model from OpenAI, called GPT-5, was delayed, and when it did come out it didn’t measure up to the hype. Normally ebullient Chief Executive Sam Altman has sounded more like an AI realist recently, saying at a media dinner that he thought investors had become overexcited about the technology.

But if the leading AI tools indeed are losing steam, that wouldn’t be a huge problem for many of the companies trying to integrate AI into how they work. It might even be welcome.

Generative AI is already powerful and useful in business—for summarizing large texts, for example, and helping employees code or write emails. Other more mundane forms of AI that predated the language-model explosion have also become increasingly useful, for tasks such as processing invoices or giving recommendations on how to manage a fleet of vehicles.

Yet most businesses have barely scratched the surface with AI in its current form, let alone what it could become if it gets a lot better.

While some have been quick to deploy AI, many more have been slow. Corporate tech leaders worry about sensitive data being leaked through chatbot conversations. They are wary of trusting AI with critical decisions that affect finances, employees and customers.

The tendency of even the best AI models to occasionally hallucinate wrong answers widens the trust gap.

A recent MIT study found that companies were already mostly comfortable with off-the-shelf generative AI tools from OpenAI and Microsoft MSFT 0.59%increase; green up pointing triangle. But when it came to building custom AI software to streamline their operations—the kind of things ostensibly most likely to produce real business returns—the failure rate for pilot projects was 95%.

Corporate AI users “were overwhelmingly skeptical of custom or vendor-pitched AI tools, describing them as brittle, overengineered, or misaligned with actual workflows,” the study’s authors said.

AI won’t stand still even if the purveyors of its most advanced models run into a wall: People will look for novel ways to improve it, and they will likely work eventually. Somewhat paradoxically, the mere perception that AI is slowing might give companies more confidence to invest time and money in it, seeing it as less of a moving target.

And the corporate world clearly does need more time to figure it out. The work of adapting large-language models to be useful for everyday tasks remains in its infancy.

“If I want to increase the pace at which I innovate, if I want to reduce the safety stock of inventory I have, if I want to improve the way I connect with millions of consumers, you have to change a lot more than just, ‘Here’s a tool that a few of your workers can use,’” said Michael Chui, a senior fellow at the AI arm of McKinsey & Co. “All of that change is really hard.”

That difficulty—a thorny management challenge as much as a technical one—means corporate AI adoption will be a multidecade effort, Chui says.

That shouldn’t be entirely surprising. The internet transformed how people live and do business, but it took much longer than its enthusiastic early boosters would have predicted in the 1990s. For instance, it took a decade for U.S. home broadband to go from near zero penetration in 2000 to more than 60% of adults’ subscribing to it, according to figures from the Pew Research Center.

The AI boom is different in many ways, but it could follow a similar trajectory: a burst of enthusiasm followed by a leveling-off as it bleeds into society and business, with the true scope of the benefits only clear years later.

In the shorter term, the sense that AI’s rise might be less meteoric than previously thought has sent tech stocks on a bumpy ride. Nvidia, Microsoft, Amazon.com, Meta and other AI leaders sold off last week before Fed Chair Jerome Powell’s comments pointing to an interest-rate cut sparked a Friday rally.

Ironically, though, that it is getting tougher to squeeze out better performance from AI models might result in an extension of the boom for some, especially “pick and shovel” makers such as Nvidia. Altman, Meta Chief Executive Mark Zuckerberg and other big AI spenders likely will put yet more money behind their attempts to overcome recent challenges.

Altman recently suggested the cure to OpenAI’s recent stumbles was to spend trillions more dollars on AI chips. And even the process of adapting models to real-world business tasks will require more incremental computing power.

Eventually, there is reason to suspect that the big winners from today’s AI boom won’t prosper quite as much, should the pace of AI innovation moderate.

But it isn’t just Big Tech that stands to gain from AI: The payday for all the companies starting to leverage it will come—it might just take longer.

There has been no soft landing for the formerly high-flying Phil Falcone.

Once a hedge-fund superstar worth $2 billion, bad investments have left him having to sell off his prized possessions to pay his bills: Upper East Side townhouses, a Hamptons mansion, a custom made grand piano, all gone.

Now he is in a legal tussle with a pawnbroker over millions of dollars of goods he and movie producer wife Lisa Maria Falcone put up as collateral, according to court records. It’s a long fall from grace from making a name as philanthropists and hosting movie stars.

But Falcone, who hauled in most of his cash betting huge against sub-prime mortgages like in movie “The Big Short,” remains optimistic about a comeback, even as others wager against him.

But first, he has to get his wife’s ring back. The pink-colored 20.45-carat diamond ring came from the elite jeweler Harry Winston and Falcone claims it is worth $6.3 million. It was one of the items pawned.

“I was prepared to buy the ring back. They sold the ring before the note was due,” the blustery Falcone claimed to The Post.

Told that the ring was sold at a foreclosure auction, Falcone claims it was a likely a set up.

“They probably sold it to one of their friends for pennies,” he added. The pawn company maintains he didn’t buy it back in time.

It’s quite a comedown for the cofounder of Harbinger Capital, which managed $26 billion at its peak.

The company’s undoing came amid SEC allegations Falcone improperly used $113 million of investor funds to pay his personal taxes. He denied this to The Post and said the money came from his own capital account.

At any rate, it resulted in an $18 million settlement and Falcone being barred from the securities industry for at least five years.

The investigation was also a harbinger of worse to come. An all-in investment ended badly, contributing to the company’s decimation and leaving Falcone to pawn some of his precious artworks to secure a loan for $92.5 million back in 2013.

Rarefied pieces including Richard Prince’s “Untitled (Cowboy),” Pablo Picasso’s “Deux Nus” and Damien Hirst’s “I love you, love buds” and “A Playful Bubblegum Kiss” were all used as guarantees.

Just one problem: It’s alleged in court papers the artwork was pledged for two different loans to two different lenders.

That’s led to a still-unresolved legal tangle between Falcone and lender BLCE. A judge ruled he should have known he couldn’t put the same asset up two different lenders, Falcone and his lawyer are appealing, contending the value of the art could have been shared by the two parties.

“These guys are professional lenders,” Falcone told The Post. “They entered into this with eyes wide open.”

Falcone insisted he is “not losing sleep over the ring and the art,” but it is the latest chapter in what appears to be an embarrassing third act in an up-from-nothing story that now includes the public liquidation of Falcone and his wife Lisa Maria’s valuable assets.

Last year, the Falcones sold a Sagaponack beachfront mansion for $14 million in an all-cash transaction. Their place in St. Barths went for $57 million. In 2019, he scored $77 million on the sale of a townhouse on 14-16 East 67th Street. Once the home of Bob Guccione, it benefited from a $10 million renovation carried out by the Falcones.

The luxe digs sported a dry-cleaning setup in the basement, a bar in one of Lisa’s walk-in closets and a heated sidewalk to keep snow from accumulating, possibly the best revenge for a guy who grew up in the frigid climes of Minnesota. At the time, it was the highest priced residential sale in New York City history.

Another Upper East Side townhouse of the Falcones went for $27 million, and all the furniture in the place was sold in what The Post dubbed “a fire sale.”

Among the top accoutrements: a custom white lacquer Steinway grand piano with a sterling silver frame (bought for $180,000 and sold for $50,000), a neoclassical artwork by Merry-Joseph Blondel and his partner Joseph Dufour went for $7,500 after being listed at $23,500 and a leather sofa (just $3,750).

In 2022, when he was called out for unloading the stuff, Falcone referenced an upcoming trial from which he hoped for a 10-figure settlement. “It’s a $5 billion lawsuit,” he told a Post reporter. “Do you think that’s worth nothing?”

Now he says, “It’s my money and these things are mine to sell.”

Other complaining creditors over the years have included a limousine company (the Wall Street Journal reported that he owed $34,000), a landlord and the law-firm that represented him in the SEC case, who all accuse Falcone of owing them money.

In 2021, the Journal reported that his one-time attorney won a $14 million judgment against him. In 2021, the attorney said Falcone has made “modest payments … as a result of our endlessly chasing him.”

Upon hearing this, Falcone said, “I’m still dealing with the lawyer, but the others have been settled. It’s been a rollercoaster.”

Once famous for owning a pot-bellied pig named Wilbur, hanging out with Alicia Keys and her husband Swizz Beatz and swanning it up at big-ticket fundraisers, Falcone and his wife hosted lavish parties where Madonna and Edward Norton would regularly attend.

Fast forward to a 2020 court appearance and Falcone was representing himself, claiming to be unable to afford a lawyer. He told the judge, “I’m behind on virtually every one of my bills. Including my kids’ tuition.”

When asked if he’s broke, Falcone shrugged it off and told The Post, “There’s a difference between liquidity and illiquidity.”

Still, headlines like “How to lose $2 billion in 10 years” do him no favors.

However, to lose $2 billion, you need to make it first. On his path to success, Philip Falcone lived the great story of the brainy New York arriviste.

Fresh out of Harvard and one season as a professional hockey player in Sweden, in 1985, he landed his first Wall Street job: trading junk bonds and earning $20,000 per year. In 1992, he met Lisa Marie Velasquez, a model. They fell in love, married in ‘97 and once shared a studio apartment with an air mattress on the floor.

Then he hit big with the his hedge fund, and hit it even bigger with the prescient and ongoing $15 billion wager that subprime mortgages would collapse.

When the sub-primes went to hell, he made money and made headlines. Lisa Marie became a big booster of the Metropolitan Opera and at an event for Friends of the High Line, she announced an apparently spontaneous $10 million donation.

Lisa Maria got involved in producing movies and bragged to W about “not making any money” from the productions. “Zero … Everyone knows I don’t have to work,”

Some of the movies, including “Mud” with Matthew McConaughey and the bio-drama “127 Hours,” starring James Franco, were profitable. Falcone, meanwhile, purchased the Minnesota Wild, an NHL team. That’s since gone the way of the fancy houses.

“That’s one thing I probably should have held onto,” he said with remorse. “When I sold it, it had a $300 million valuation. Now they’re saying it’s worth $1.6 billion. It was part of my selling assets to pay down debt. But, you know, it doesn’t mean I can’t do it again.”

Still, he remains upbeat and warns he may appear down, but he’s not out.

“People say I’ve fallen on hard times. I don’t even know what the hell that means. It’s not like I am having chemotherapy for God’s sake.

“I’m happy and thriving. I came from nothing. I lost a little bit of money. People say I am sitting on death’s door and I’m not. Not even close.”

WASHINGTON—The Pentagon has for months been blocking Ukraine’s use of long-range missiles to strike inside Russia, U.S. officials said, limiting Kyiv from employing a powerful weapon in its fight against Moscow’s invasion.

A high-level Defense Department approval procedure, which hasn’t been announced, has prevented Ukraine from firing any U.S.-made long-range Army Tactical Missile Systems, or Atacms, against targets in Russia since late spring, the officials said. On at least one occasion, Ukraine sought to use Atacms against a target on Russian territory but was rejected, two officials said.

The U.S. veto of long-range strikes has restricted Ukraine’s military operations as the White House has sought to woo the Kremlin into beginning peace talks.

Elbridge Colby, the Pentagon’s undersecretary for policy, developed the “review mechanism” to decide on Kyiv’s requests to fire long-range U.S.-made weapons as well as those provided to Ukraine by European allies that rely on American intelligence and components.

The review gives Defense Secretary Pete Hegseth final say over whether Ukraine can employ the Atacms, which have a range of nearly 190 miles, to strike Russia.

“President Trump has been very clear that the war in Ukraine needs to end. There has been no change in military posture in Russia-Ukraine at this time,” White House press secretary Karoline Leavitt said in a statement. “Secretary Hegseth is working in lockstep with President Trump.”

The Pentagon and Ukrainian officials didn’t respond to requests for comment.

The Pentagon’s approval requirement has effectively reversed a decision by President Joe Biden in his final year in office to permit Ukraine to strike inside Russia with Atacms.

In a social-media post Thursday, Trump said Ukraine couldn’t defeat Russia unless it could “play offense” in the war, which has lasted longer than three years following Moscow’s invasion.

“It is very hard, if not impossible, to win a war without attacking” the invading country, he wrote. “There is no chance of winning!”

U.S. officials said Trump’s statement didn’t signal a policy change that would dispense with the Pentagon’s review mechanism or encourage Ukraine’s use of Atacms and other long-range Western systems. But a senior White House official said Trump could change his mind about facilitating expanded offensive operations against Russia.

As president-elect, Trump said it was “stupid” of Biden to allow Ukraine to attack inside Russia. “We’re just escalating this war and making it worse,” he said in a December interview with Time.

U.S. Atacms and other long-range weapons, such as Britain’s Storm Shadow cruise missile, haven’t been game-changers for Ukraine. But they have enabled it to threaten Russian command-and-control headquarters and airfields far from the front lines, as Moscow presses its advantages in manpower, weaponry and resources.

The Pentagon review procedure also applies to Ukraine’s use of the Storm Shadow as it relies on U.S. targeting data, according to two U.S. officials and a British official. The U.K. government declined to comment.

Trump has threatened to increase sanctions on the Kremlin and tariffs on its major trading partners unless Moscow agrees to a cease-fire. But a decision on that has been deferred following his summit with Russian President Vladimir Putin, in which the Kremlin chief convinced Trump he was serious about making peace.

Trump vowed in July to provide Kyiv with new weapons as long as Europe agreed to pay for them, a decision that came after a brief pause in U.S. arms shipments while the Pentagon examined whether there were shortfalls in its own stockpiles. Trump told reporters that month that “we’re not looking” to provide longer-range weapons to target Moscow.

While the U.S. hasn’t announced any plans to provide more of the missiles, other types of weapons that European governments are buying from the U.S. could help Ukraine within its own borders. They include air-defense systems and the Guided Multiple Launch Rocket System, which has a 90-mile range.

The administration this past week approved the sale of 3,350 Extended Range Attack Munition air-launched missiles, or ERAMs, which are set to arrive in Ukraine in about six weeks, two U.S. officials said. The $850 million arms package, mostly funded by European nations and which includes other items, was delayed until after Trump’s summits with Putin and Ukrainian President Volodymyr Zelensky.

Several U.S. officials said use of the ERAM, with a range between 150-280 miles, would require Ukraine to seek approval from the Pentagon. The State Department didn’t respond to a request for comment.

The Biden administration provided hundreds of Atacms to Ukraine starting in 2023. The restrictions on Atacms attacks inside Russia initially imposed under Biden were removed in the fall of 2024 after North Korean troops entered the war alongside Russia.

The last of the Atacms authorized to be sent to Ukraine by the Biden administration arrived during the spring, and Kyiv has a small supply left, according to U.S. officials.

Ukraine has developed some of its own long-range weapons to strike targets inside Russia, specifically drones that it has used to strike Russian oil refineries and aircraft.

Zelensky said recently that his country was developing a new cruise missile—the Flamingo—that could be produced in significant quantities by the end of this year or early next year.

“Drones are great for certain things, but they have their own vulnerabilities too,” said James Townsend, who was the top Pentagon official for North Atlantic Treaty Organization issues during the Obama administration, about the restrictions on use of Atacms. “You don’t want to limit the Ukrainian ability to put pressure on the Russians.”

In addition to requiring approval for missile attacks, Colby, the Pentagon’s third-ranking official who has long been concerned about husbanding Pentagon resources to deter China, tightened control over additional munitions for Ukraine. In a June memo he established three categories—green, yellow and red—for assessing whether U.S. stockpiles of weapons were adequate to allow some to be shared with Kyiv.

The green category consisted of systems that the U.S. had in abundance and that could readily be provided to Ukraine, while yellow and red systems are in shorter supply. It gave Hegseth authority to reclaim weapon systems earmarked for Ukraine that fell in the yellow and red categories.