Turn insights into solutions

CES is where technology comes to announce itself. Eight years ago it was voice assistants. Four years ago it was the metaverse. This year it was full of AI, but the room responded most strongly to a Danish toy company that spent years making a brick that hums when you swing a lightsaber.

The Smart Brick is genuinely impressive engineering — a 4.1mm custom chip packed with accelerometers, magnetic field sensors, light sensors, a synthesiser and a miniature speaker, all inside a standard 2x4 brick that is backwards compatible with every LEGO set made since 1958. Swing a Star Wars lightsaber and it hums. Move the X-wing and the engines roar as lights fire across the hull. Put the policeman in the driver's seat and the chase begins. There is no app, no pairing, no screen. The reactions are generated live from motion, placement and context — not triggered from pre-recorded clips. LEGO describes this as its most significant evolution since the introduction of the Minifigure in 1978.

CES is dominated by companies racing to embed AI into everything — laptops, televisions, refrigerators, cars. The premise of the show is that more intelligence, more connectivity, and more software is always the direction of progress. Against that backdrop, LEGO arrived with a product whose central argument is almost the opposite: that the most valuable thing technology can do is disappear into a physical object and enhance what the hands are already doing. No screen. No app. No account.

Some critics worried the Smart Brick could undermine what was once great about LEGO — that children imagine the sounds and lights themselves. Others who got the sets early found the experience underwhelming: the sounds don't match Star Wars, the battery lasts under an hour. But at the most AI-saturated CES so far, the thing people talked about was a toy disconnected from AI.

Published initially in June 2024 (revised in January 2026), the paper landed in the same quarter that Starbucks reported its worst sales performance in years, replaced its CEO, and launched a turnaround plan called "Back to Starbucks" — an explicit attempt to recover the third-place identity the chain had spent a decade quietly dismantling. Outlet covers had been removed, soft chairs replaced by hard ones, drive-thru and mobile ordering now accounting for over 70% of transactions. The Columbia economists were measuring the social and economic value of something the company was simultaneously deciding it could no longer afford to provide for free.

Ray Oldenburg, who coined the term "third place" in his 1989 book The Great Good Place, was clear that the model depended on informality and low cost of entry — spaces where showing up was enough, and where no transaction was formally required to belong. What Starbucks understood, and what the NBER data confirms, is that the value produced by these spaces is real: denser networks, more ideas in circulation, more firms started. What neither Oldenburg nor Starbucks fully resolved is who absorbs the cost of the people who stay three hours on one coffee. That question has now been answered, at some locations in New York, with a membership fee charged separately from any drink — table access priced explicitly, belonging monetised directly. The third place, in other words, is becoming a second workspace.

In France the same collapse has been slower, older, and structurally different. France had 500,000 cafés in 1900 and fewer than 40,000 today, a decline driven by deindustrialisation, car culture sprawl, and the slow disappearance of the daily rituals — the espresso at the zinc counter, the lunch crowd, the après-work aperitif — things that made the business viable. The French bistro was never a brand strategy; it was civic infrastructure, and a community that lost its last one lost the room where its social life happened. In March 2025, a French lawmaker introduced a bill to ease alcohol licensing rules specifically to allow new bars to open in villages under 3,500 residents. Legislative action to restore something the market had removed. The association representing France's bistro owners has twice applied for UNESCO intangible cultural heritage status, watching French gastronomy accumulate distinctions while the neighbourhood institution that hosted it daily remains unrecognised.

The social cost of all this is not abstract. The 2023 US Surgeon General's advisory on loneliness — a public health document with peer-reviewed foundations — reported that roughly half of American adults were already experiencing loneliness before the pandemic began, with the highest rates among younger adults. The populations most likely to be in cafés with laptops are the ones registering the highest disconnection. The NBER paper shows that when a café opens in a neighbourhood that had none, startups follow as the networks form and the ideas flow. The question the paper does not answer, and that nobody has answered cleanly, is what happens to those networks when the café converts to mobile-only pickup, starts charging for the chair or simply closes.

In February 2025, Duolingo killed Duo — its green owl mascot — in a hit-and-run involving a Tesla truck, generating over a billion organic impressions. It was the most controlled piece of manufactured grief in brand history. Three months later, they published a LinkedIn memo announcing it would become "AI-first," phase out contractors, and only hire where teams couldn't automate further. The top comment on TikTok — "Mama, may I have real people running the company" — got 69,000 likes. Duolingo wiped all its social media content. The brand that had spent years building the internet's most loyal and slightly unhinged fanbase managed, in a single memo, to make itself feel like the villain it had always pretended to be.

The CEO clarified, walked back, and clarified again. Duolingo launched 148 new language courses in under a year using generative AI — a process that would previously have taken decades. Earnings beat estimates. Daily active users grew 40% year-on-year. Stock rose 30% on the news. The backlash, TechCrunch noted drily, "didn't even matter." The brand took a hit, but the product got bigger and faster.

But the AI-first pivot obscured a more important question. Duolingo's business depends on people wanting to learn languages. Apple AirPods with live translation, Meta glasses, and Google Translate accessible anywhere are quietly making fluency unnecessary for the practical reasons most people start learning in the first place. And the current version of ChatGPT can teach them some basic Spanish. This is what investors were factoring in when the stock fell 80% from its May 2025 highs to February 2026.

The owl is the visible surface of something more substantial. Grammarly faces the same AI threat and has none of it — pure utility, fully exposed. What makes Duolingo harder to replace is that the owl embodies a method that actually works for habit formation in a way that an open-ended conversation with ChatGPT doesn't. ChatGPT is infinitely capable and infinitely unstructured. It will teach you Spanish if you know how to ask, how to progress, how to test yourself, how to stay motivated. Most people don't — which is exactly why Duolingo exists. Build something people don't want to lose, not because of what it does, but because of what it is and how it makes them feel about themselves while doing it.

iRobot's founder Colin Angle has a clear explanation for what went wrong: the Chinese fast follower. Companies like Roborock, Dreame, and Ecovacs entered the market after 2018 with a protected domestic market to cut their teeth on, iterated faster than iRobot could respond, and arrived in Western markets with better products at lower prices. That framing is accurate as far as it goes. It undersells the more specific story of what "better" actually meant — and why consumers chose it.

The first dimension is functional. iRobot stayed focused on vacuuming. Chinese competitors moved earlier and faster into mopping — a feature that sounds incremental but represents a fundamentally different product proposition. A robot that vacuums saves you one task. A robot that vacuums and mops replaces a cleaning session. Roborock, Dreame, and Ecovacs were building combination machines with self-emptying docks, auto-washing mop pads, and hot water cleaning cycles while Roomba was still refining its bump-and-turn navigation. iRobot adopted LiDAR navigation only in its 2025 lineup — years after Chinese brands had made it standard across mid-range models. By then, Chinese competitors had turned a premium feature into a floor expectation: budget models under $250 now ship with LiDAR navigation and self-emptying docks that were exclusive to $1,000+ devices in 2023.

The second dimension is privacy — and here the competitive logic is less obvious but more durable. iRobot's camera-based navigation created a persistent anxiety. In 2022, MIT Technology Review revealed that development Roombas had captured intimate images inside testers' homes which subsequently appeared on social media. The incident involved test units, not consumer products. The damage to trust was not contained to test units. The proposed Amazon acquisition amplified the concern: a company already accused of building surveillance infrastructure around its customers was about to own a robot that mapped the interior of your home room by room. European regulators opened an inquiry. The acquisition was blocked — partly on antitrust grounds, but the data anxiety it crystallised did not disappear with the deal. Chinese brands, whatever their own data practices, arrived without that specific story attached. Roborock's navigation relies primarily on LiDAR — a laser-based spatial sensing system that maps distance and geometry without capturing images of the people living in the home. The consumer proposition is meaningfully different: the robot knows where the sofa is without knowing what you look like on your sofa.

iRobot created a category and then held its position rather than developed it, while a cohort of competitors used a protected home market to fund rapid iteration and arrived in the West having already solved the problems Roomba hadn't. The bankruptcy is being read primarily as a geopolitical story — unfair competition, protected markets, the Amazon deal blocked at the wrong moment. It is also a product story: a pioneer that confused inventing a category with owning it, while the companies that came after kept asking what the product should actually do next.

Labubu is the most compelling cultural export China has produced in decades. It didn't need a narrative or government support. It just needed a blind box and a waiting list. Pop Mart's genius is in the mechanic — you don't buy a product, you buy anticipation, the small suspense of not knowing what's inside. And that feeling crosses borders effortlessly. It's on Lisa's handbag, in queues outside Ohio malls, on TikTok feeds across languages and time zones.

Soft power has always worked best when desire arrives before questions do — Levi's making Soviet-era teenagers dream of another world, anime giving a generation a new visual language, K-pop turning fans into participants. China spent decades as the world's factory floor, useful but not charismatic. What's shifted is that the brands arriving now use price to get in the room, and product to stay in it.

TikTok is the bigger story. With over a billion users worldwide, it is the platform through which an entire generation of young Westerners encountered a Chinese brand, without ever thinking of it that way. That's the most efficient soft power machine ever built: one that doesn't announce itself, doesn't ask for permission, and runs on dance videos and thirty-second recipes.

Luckin Coffee doesn't just do discounts — the app floods you with coupons so relentlessly that paying full price feels like a mistake. By the time you notice you've built a habit, you are ‘lucked’-in. In Brazil, BYD took a different route: it bought Ford's shuttered factory and is now building one of Latin America's largest EV plants on the site. Chinese expansion isn't just commercial. It's physical. They're occupying the infrastructure others left behind.

Young Americans and Europeans tend to view China more favourably than their parents. They think of Chinese products simply as apps. That's what soft power looks like when it's working: nebulous and unidentifiable. Whether it lasts is a different question. Retention rates tell a more sober story — Temu holds onto 60% of its users against Amazon's 93%, and Pop Mart's model lives and dies by the next drop. The brands that endure won't be the ones that arrived with the most hype. They'll be the ones that quietly became habits.