Remarkable 15
Take The American Gut Project, a 2018 endeavour that combined the efforts of dozens of researchers and more than 10,000 “citizen scientists”. This bit of research, conducted across the US, UK and Australia, found that those who eat upwards of 30 different plants each week are far more likely to have a gut populated with the most beneficial bacteria than those who eat just 10. It isn’t just fruit and veg – anything that grows counts towards your 30 per week, from nuts and seeds to lentils, pulses, herbs and spices. Even tea and coffee, popcorn and dark chocolate will top up your total.
How to eat 30 plants a week and transform your gut health
For years, the Tesla versus Waymo debate has been framed like a referendum on ideology. Tesla is supposed to be the scrappy scale play: millions of cars, camera-only, end-to-end neural networks, and a system that is not “boxed in” by geofences. Waymo, by contrast, is often portrayed as the cautious boutique: expensive sensors, mapped cities, limited domains, and a business that looks like a science project. That framing is comforting because it turns autonomy into a future-tense argument. In 2026, it is getting harder to keep it in the future tense. Waymo is already selling the thing Tesla is still selling as a promise: rides where nobody is responsible for driving, because nobody is driving. Tesla’s most common “FSD advantage” arguments keep circling around scale, simplicity, and generality. But when you look at what autonomy actually is, and what it takes to scale it safely, those arguments start collapsing one by one.
Waymo is Destroying Tesla’s Self-Driving Dreams
The impact of the Citrini scenario has startled some commentators, including experts who say AI tools are not yet capable of enacting it. Stephen Innes, a managing partner at SPI Asset Management, says AI thought pieces have become market movers. “We have watched this market absorb wars, sticky inflation, banking tremors and tariff theatrics with a shrug, yet a widely circulated Substack thought piece is enough to knock it sideways,” he said.
‘A feedback loop with no brake’: how an AI doomsday report shook US markets
What if artificial intelligence actually meets or exceeds the hype around it, but that turns out to be very bearish for stocks? That’s the scenario posed by a new piece of commentary doing the rounds over the weekend, which has caused quite a flutter online. The piece, co-authored by Citrini Research and guest Alap Shah, managing partner at Lotus Technology Management, is written as a lookback from June 2028, when the unemployment rate has just risen to 10.2% and the S&P 500 is down 38% from its Oct. 2026 highs around 8,000. The first step on this dystopian journey will be familiar to us in 2026: the dislocation caused by agentic AI. Software-as-a -service companies will lose sales as client demand falls. “The company that sold workflow automation was being disrupted by better workflow automation, and its response was to cut headcount and use the savings to fund the very technology disrupting it,” projects the report.
The Citrini post argued that investors and workers hurt by AI will cut their spending, which they will. But if AI delivers big productivity gains, it will reduce prices and raise real income in sectors that aren’t displaced, causing other Americans to spend more. There’s no reason to believe that disrupting part of the economy will reduce overall demand. The only way I can see that AI could be a recessionary force would be if the firms and/or workers who lose from the technology were highly leveraged — that is, were carrying a lot of debt — and so were forced to cut their spending much more than those gaining from AI increased their spending. But there’s no evidence for that.
When Extraterrestrials Attacked the Stock Market
Bitwarden, LastPass, and Dashlane are less secure than you might expect, at least if you go by the findings of security researchers at ETH Zurich and the Università della Svizzera italiana (USI) in Lugano. They’ve allegedly discovered serious security vulnerabilities in these popular password managers. “In tests, they were able to view and even change stored passwords,” writes the editor (machine translated).
Your password manager isn’t as safe as you think
Hundreds of millions of people now use chatbots every day. And yet the large language models that drive them are so complicated that nobody really understands what they are, how they work, or exactly what they can and can’t do—not even the people who build them. Weird, right? It’s also a problem. Without a clear idea of what’s going on under the hood, it’s hard to get a grip on the technology’s limitations, figure out exactly why models hallucinate, or set guardrails to keep them in check. But last year we got the best sense yet of how LLMs function, as researchers at top AI companies began developing new ways to probe these models’ inner workings and started to piece together parts of the puzzle.
10 Breakthrough Technologies 2026
Unitree Robotics is planning a major production ramp-up, aiming to ship up to 20,000 humanoid robots in 2026. The target marks a sharp rise from the roughly 5,500 units the company shipped last year. The ambitious target follows a widely watched appearance at China’s annual Spring Festival Gala, broadcast by state media giant China Central Television (CCTV). The event, known for drawing massive domestic and international audiences, has increasingly become a showcase for the country’s robotics progress. Speaking to Chinese media outlet 36Kr after the show, Unitree CEO Wang Xingxing said global humanoid robot shipments could reach “tens of thousands” this year, with his company accounting for around 10,000 to 20,000 units, according to SCMP..
China’s Unitree Aims to Ship 20,000 Humanoid Robots in 2026
But the era of the “pocket book” is drawing to a close. ReaderLink, the biggest book distributor in the US, announced recently that it would stop distributing mass-market paperbacks. The decision follows years of plummeting sales, from 131m units in 2004 to 21m in 2024, and marks the end of a format that once democratised reading for the working class.
Bodo/Glimt, who were in Norway’s second-tier as recently as 2017, are the northernmost team to ever play in the Champions League. Based in the small town of Bodo, a 16-hour drive north of Oslo and inside the edge of the Arctic Circle, its whole population of 55,000 could have travelled to Inter’s iconic San Siro and there still would have been plenty of empty seats. Bodo/Glimt also operate on a budget that is a fraction of the size of Europe’s biggest clubs. Remarkably, Knutsen’s team are also in their off-season; the Norwegian top-flight, the Eliteserien, ended on 30 November 2025 due to the winter and will only resume when spring arrives in mid-March. But, in that time, Bodo/Glimt have beaten Manchester City, Atletico Madrid and knocked out Italian giants Inter, the runaway Serie A leaders, by winning home and away... Bodo/Glimt are also the first team outside of England, Spain, Germany Italy or France to win four consecutive games against teams from Europe’s big five leagues since Johan Cruyff’s Ajax in 1971-72. Ajax went on to win the European Cup that season, too.
Buenos Aires City recorded the lowest homicide rate in 31 years in 2025, with all crimes falling by 30%, according to official data city authorities presented on Tuesday. Mayor Jorge Macri announced the findings during the presentation of the 2025 Crime Map, a management tool for processing and analyzing crime statistics the city has been presenting since 2017. Last year’s homicide rate of 2.5 per 100,000 is 3% lower than 2024. The figure, lower than the rate for the country (3.7) and Buenos Aires province (4.34) means the city ranks as “the second safest capital in the Americas, behind Ottawa, Canada,” the official statement said.
Buenos Aires City records lowest homicide rate in 31 years - Buenos Aires Herald
Firefighters rushing into burning buildings may soon have backup that never tires, never panics, and never needs to catch its breath. Hyundai Motor Group says it donated four unmanned firefighting robots to South Korea’s National Fire Agency, designed for remote fire suppression and search work in “high-risk environments.” If they perform as intended, the robots could help keep more firefighters alive by taking on early-stage tasks when smoke, heat, and structural uncertainty are at their worst.
Hyundai’s Robot Firefighters Head to South Korea
In 2021, the Australian company Cortical Labs used its neuron-powered computer chips to play Pong. The chips consisted of clumps of more than 800,000 living brain cells grown on top of microelectrode arrays that can both send and receive electrical signals. Researchers had to carefully train the chips to control the paddles on either side of the screen. Now, Cortical Labs has developed an interface that makes it easier to program these chips using the popular programming language Python. An independent developer, Sean Cole, then used Python to teach the chips to play Doom, which he did in around a week. “Unlike the Pong work that we did a few years ago, which represented years of painstaking scientific effort, this demonstration has been done in a matter of days by someone who previously had relatively little expertise working directly with biology,” says Brett Kagan of Cortical Labs. “It’s this accessibility and this flexibility that makes it truly exciting.”
Human brain cells on a chip learned to play Doom in a week
We were the last generation raised on the principle that discomfort wasn’t the enemy. It was information. Hungry because you forgot your lunch? You’d remember tomorrow. Cold because you left your coat at home? You’d check the weather next time. These weren’t traumas requiring therapy; they were life’s little tutorials in cause and effect. Research published in Behavioral Sciences found that parental overprotection is negatively associated with offspring’s emotional intelligence and physical health in adulthood, with emotional intelligence mediating this relationship. In other words, all that cushioning doesn’t protect kids. It handicaps them.
Every AI model you’ve ever used writes the same way: one word at a time, left to right, like a typewriter. If it drifts off course early, tough luck. It keeps typing. Well, Inception Labs just launched Mercury 2, which is an AI that works completely differently: Instead of predicting one word after another, it starts with a rough sketch of the entire answer. Then, it refines everything at once, like an editor revising a full draft in parallel. The technical term is a “diffusion LLM” (dLLM), which uses the same core approach behind AI image generators like Midjourney, but applied to text and reasoning. Let us tell you, the speed is real. Independent testing from Artificial Analysis clocked Mercury 2 at 1,196 tokens per second, over 3x faster than the next fastest model in its price class. That’s a very big deal if you need speed. For context, Claude 4.5 Haiku hits ~89 tokens/sec and GPT-5 Mini ~73. RIP.