Today's Wisdom - Cocktails, Prediction Models and Neuralink

Welcome back to A Little Wiser. We hope everyone is having a great week. Before we dive in, make sure to watch out for Friday’s newsletter, we have a brilliant feature on Rolex lined up that you won't want to miss. Today's wisdom explores:

Grab your coffee and let’s dive in.

The cocktail as a category is surprisingly young and for most of human history, people drank wine, beer, and spirits in largely unadorned form. Everything changed with ice. The American entrepreneur Frederic Tudor, who became known as the Ice King, built a global trade in harvested natural ice during the 1820s and 1830s, shipping blocks cut from frozen New England lakes to cities across the United States, the Caribbean, and eventually India. For the first time, bartenders in warm cities had reliable access to something cold, and the American cocktail culture that emerged from that moment was unlike anything that existed elsewhere in the world.

The individual histories of the drinks most people order without thinking about are considerably stranger than the drinks themselves suggest. The Martini's origins are genuinely contested, with claims ranging from a gold rush town in California called Martinez to the Knickerbocker Hotel in New York. What is actually documented is that the drink began its life far sweeter than the ice-cold, bone-dry version that became culturally dominant in the twentieth century, with early recipes calling for sweet vermouth and a dash of maraschino liqueur. The Negroni was born in Florence in 1919 when Count Camillo Negroni walked into Caffe Casoni and asked the bartender Fosco Scarselli to strengthen his Americano by replacing the soda water with gin, a substitution so successful that the drink has carried his name ever since.

The twentieth century produced its own mythology at a pace that older drinks never managed. The Moscow Mule was invented in 1941 in Los Angeles as an explicit commercial collaboration between a vodka importer named John Martin, who had purchased the Smirnoff brand and was struggling to sell Americans on a spirit they had no cultural relationship with, and a bar owner named Jack Morgan who had a surplus of ginger beer and a collection of copper mugs he could not shift. The drink was a marketing exercise that became a genuine classic. The Espresso Martini was created in London in 1983 when a young model sat down at the bar of the Brasserie Soho and told the bartender Dick Bradsell that she wanted something that would, in her words, wake her up and then mess her up. What unites all of these origin stories is the same thing that unites the drinks themselves: they were almost never planned, rarely sophisticated in their conception, and almost always the product of someone wanting something that did not yet exist and being impatient enough to make it.

In 1984, a young researcher named Philip Tetlock began an experiment that would run for twenty years and produce the most comprehensive dataset ever assembled on the accuracy of expert prediction. He recruited 284 professional forecasters and asked them to make nearly 28,000 specific, falsifiable predictions about future events. When he finally analyzed the results in 2005, the average expert performed barely better than random chance. A dart-throwing chimpanzee, as Tetlock famously put it, would have matched most of them. The more eminent the expert, the more media appearances they made, the more confidently they spoke, the worse their calibration tended to be. What Tetlock also found, buried inside the aggregate failure, was a smaller group of forecasters who were genuinely and consistently better than chance. He called them foxes, borrowing from the ancient Greek poet Archilochus who wrote that the fox knows many things while the hedgehog knows one big thing. The poor forecasters were hedgehogs: specialists with a single grand framework they applied to every problem with total conviction. The good forecasters were foxes: comfortable with uncertainty, willing to update their views when new evidence arrived, and ruthless about distinguishing between what they actually knew and what they were merely assuming. The fox's central tool was probabilistic thinking, the habit of expressing beliefs as likelihoods, and then tracking those likelihoods against outcomes over time to measure and improve calibration.

The machinery behind modern forecasting models reflects the same lessons at a technical level. Monte Carlo simulation, the method used in everything from nuclear weapons design to World Cup prediction, works by running a process thousands or tens of thousands of times with randomness built in at each step, generating a probability distribution across possible outcomes rather than a single point prediction. The approach was pioneered at Los Alamos during the Manhattan Project by mathematicians Stanislaw Ulam and John von Neumann, who named it after the Monaco casino because the method is essentially structured gambling: you cannot know what will happen in any single run, but across enough runs the underlying probabilities reveal themselves with precision. The critical insight is that a good Monte Carlo model is not trying to tell you what will happen but the shape of what could happen, and what the realistic range of outcomes looks like given everything you currently know. The domains where prediction remains genuinely hard, where even the best models and the best forecasters struggle, are those where feedback loops are long, where randomness is high relative to signal, and where the act of making a prediction changes the behavior of the system being predicted. Financial markets are the canonical example. Sport sits somewhere in the middle, offering enough signal to reward rigorous modeling and enough irreducible randomness to ensure that even an octopus occasionally is right (context in report below)!

So we decided to build our own. For the 2026 World Cup, we took Joachim Klement's Panmure Liberum econometric model, which has correctly predicted the last three World Cup winners using five socioeconomic variables, and extended it to eleven, running 10,000 fully independent Monte Carlo simulations through the official FIFA bracket. The additions were driven by the logic Tetlock identified in his best forecasters: a refusal to ignore variables simply because previous models had not bothered to include them. We replaced the static temperature proxy with an acclimatisation delta measuring the gap between where each squad's players actually train and the specific conditions of their assigned venue, added altitude exposure as a separate variable using a VO2 max impairment model, and introduced a travel burden index calculated from the actual distances between group stage venues divided by rest days between fixtures. The 2026 tournament, spread across three countries and roughly 7,000 kilometers of North American geography, from Mexico City at 2,240 meters above sea level to Miami in July heat that FIFPRO has designated an extreme physiological risk, creates conditions no previous World Cup model has had to account for. The full model and every group stage prediction is available in this month's special report, linked below.

The idea of connecting the human brain directly to a computer sounds like the premise of a science fiction film made in the wrong decade, but it has been the working ambition of serious neuroscientists for longer than most people realize. The underlying concept is called a brain-computer interface, a device that reads electrical signals generated by neurons and translates them into digital commands, effectively allowing the brain to communicate with external technology without the intermediary of muscle movement or speech. The scientific foundations were established in the 1960s and 1970s, when researchers at institutions including UCLA began demonstrating that animals could learn to modulate their own neural activity to control external devices. The field remained largely theoretical for decades, constrained by the difficulty of implanting anything in the human brain without causing serious damage and the limited processing power available to decode the signals that came out.

What changed the pace of the field was money, ambition, and Elon Musk. In 2016, Musk co-founded Neuralink with a team of neuroscientists and engineers and set an explicit goal that the academic field had historically been cautious about stating publicly: to create a device capable of giving ordinary humans cognitive enhancement beyond the restoration of lost function. The company developed a chip the size of a large coin, designed to be inserted into the skull by a surgical robot with a precision that human hands cannot replicate, with threads thinner than a human hair extending into brain tissue to record from thousands of neurons simultaneously. In January 2024, Neuralink implanted its first device in a human patient, a 29-year-old man named Noland Arbaugh who had been paralyzed from the shoulders down following a diving accident. Within weeks, Arbaugh was controlling a computer cursor with his thoughts, playing chess and the video game Civilization online, and describing the experience in a live-streamed demonstration that drew global attention. He told journalists it was like using the Force, a description that captured something the technical specifications could not.

Neuralink is the most visible player in the field but not the only one. BrainGate, a research consortium involving Brown University, Harvard, and several other institutions, has been implanting research devices in paralyzed patients since 2004 and has produced some of the most rigorous published science in the area, including a 2022 study in which a man with ALS was able to type at eighteen words per minute using only his neural signals. Synchron, an Australian-American company, developed a device called the Stentrode that is delivered through blood vessels rather than direct surgical implantation, avoiding the need to cut open the skull entirely, and completed its first American human trials in 2022. The medical applications of this technology are genuinely transformative and relatively uncontroversial: restoring communication to people with locked-in syndrome, returning mobility to the paralyzed, treating treatment-resistant depression through precise neural stimulation. The questions that have no clean answers yet concern what comes after the medical applications, when the technology becomes reliable enough, and cheap enough, to be offered to people who have nothing wrong with them at all. At that point, the conversation stops being about medicine and starts being about what kind of species we want to become.

We hope you enjoyed today’s edition! If you did, feel free to share it on social media or forward this email to friends.

Until next time... A Little Wiser Team