Square launches AI kiosk, Android taps Gemini for scam defense: AI is embedding into hardware and systems

AI is moving out of standalone chat windows and into checkout counters, text-message inboxes and weather forecasting systems: Square is putting conversational AI inside ordering hardware, Google is using Gemini to intercept scam messages on Android, and Microsoft is using machine learning to correct systematic bias in weather models. At the same time, Netflix is launching its largest round of layoffs in years even as its business remains solid, and indie game Titanium Court is expanding from Steam to consoles and mobile. The spread of technical capability and the reallocation of platform resources are happening at once.

Square puts conversational AI inside a 21.5-inch ordering kiosk

Square has launched new hardware called Square Kiosk, a wide pill-shaped device with a 21.5-inch touchscreen and a conversational AI voice layer. According to Square, this 'agentic, conversational voice layer' fully understands a seller's menu and listens and responds in real time, enabling customers to order the way they'd naturally talk to staff; during the conversation, customers can ask questions and request modifications or substitutions. The device also has presence detection, so it can greet customers as they approach, and it offers both a contactless card reader and a card slot for payment. Square Kiosk is available starting today for businesses with a Square Plus or Premium subscription, at $1,999 for the hardware plus a $99 per month per device software fee. For food-service merchants, this means AI is no longer just a back-office analytics tool but appears directly at the front line where transactions happen; the hardware-plus-subscription pricing model will also test whether small and midsize merchants judge 'conversational ordering' to be worth the cost. ^[1]

Microsoft open-sources a weather forecast bias-correction tool aimed at two-to-six-week outlooks

Microsoft Research is applying machine learning to the 'bias correction' step of weather forecasting rather than training an entirely new weather model from scratch. Researchers say they have found a cheaper way to improve forecast accuracy two to six weeks in advance, with the findings published in the journal Nature and the tool released as open-source software. Traditional physics-based models rely on supercomputers to simulate atmospheric physics, while generative AI models rely on powerful GPUs; Lester Mackey, a senior principal researcher at Microsoft Research New England, said the new tool can run on CPUs alone, making it more accessible to weather agencies with limited resources. Both physics-based and AI models run into similar problems when predicting weather more than two weeks out: they need approximations to fill gaps in data, and those patches can become systematic biases that compound over longer lead times. Microsoft's method is called Probabilistic Bias Correction, which uses machine learning to identify and correct systematic biases in models. Mackey said one use is correcting for the effects of climate change. Mackey also said he is optimistic about improvements in the resource efficiency of weather forecasting and about extending the reliable range of extreme-weather forecasts. ^[4]

Android: Gemini enters the text-message scam-defense chain

Android is adding new scam defenses to Google Messages. Building on the existing Scam Detection feature, a new 'enhanced protection' setting will roll out in the coming weeks and will tap cloud-based Gemini AI models to quickly assess the intentions of a potential scammer in conversations with an unrecognized sender. Google said in its update announcement that scammers often disguise their true motives behind slow-burn conversations, and that this secure cloud analysis augments and extends existing on-device detection to help identify novel and sophisticated attacks. User data is encrypted during this AI analysis and processed in a dedicated secure space that Google and anyone else cannot access. Enhanced protection is opt-in, and it also uses on-device Gemini Nano to check real-time web signals for unknown URLs, determining whether a link is malicious before the user taps it. Google also announced additional security measures: blocking apps from secretly enabling call forwarding and notifying users when someone attempts it, and automatically hiding sensitive information during screen sharing in Chrome and Android WebView. ^[5]

From notifications to calls: scam defenses expand to more devices

Google also announced more Android scam-defense and privacy-protection expansion plans. With Scam Detection enabled, Google's most advanced Gemini models evaluate sender intent and warn when a message may come from a bad actor; in Google's example, a sender first pretends to want to show a photo with a pet dog, then, after being told they have the wrong number, tries to continue the conversation and asks for the recipient's name. The analysis takes place in Private AI Compute, where messages are encrypted and processed in a secure space inaccessible to others; Gemini Nano also analyzes links in Messages on-device. The feature will come to Google Pixel 10 and newer devices, Samsung Galaxy S26 series, Galaxy Z Flip 8 and Galaxy Fold 8, and some other devices. Scam Detection had previously rolled out in chat notifications on Pixel devices; now Google is working with Honor, Motorola, OPPO, OnePlus, realme, vivo, Xiaomi and other manufacturers, with availability expected on those devices later this year. Android 15 and newer versions will get automatic screen-share privacy protection this month, hiding passwords, banking information and credit card security codes by default; in early December, call forwarding protection will arrive on Pixel and some other devices, blocking apps from secretly enabling call forwarding in the background and warning when scammers trick users into entering MMI forwarding codes. Google also said it has partnered with Brazilian banks to fight bank impersonation; Android verifies calls and hangs up and warns the user if verification fails; since launching in June, it has blocked more than 3 million fake calls, though whether it will come to the US is unclear. ^[6]

Netflix cuts about 1,000 jobs: reallocating resources around growth targets

Netflix plans to lay off around 1,000 employees, roughly 6 percent of its workforce. According to The Wall Street Journal, Netflix had around 16,000 employees worldwide last year; this is the company's largest round of cuts since 2022, when around 450 employees were let go. The move comes after Netflix backed away earlier this year from its bid to acquire Warner Bros., a deal ultimately won by Paramount Skydance. In a company-wide email, Netflix co-CEOs Ted Sarandos and Greg Peters wrote that the company's mission to entertain the world is ambitious, and that delivering on it requires discipline about where to prioritize people and resources; they also stressed that the business is strong and still growing, that full-year guidance is unchanged, and that the company will continue to invest aggressively in high-impact areas. By the numbers, Netflix's 2025 revenue was around $45.2 billion, and it had previously raised consumer prices, so this does not look like a company in crisis. In recent years Netflix has also expanded beyond movies and TV into games, podcasts and live sports. The layoffs therefore look more like a reallocation of resources around growth targets: concentrating people in businesses judged to be high-impact. ^[2]

Titanium Court goes from Steam to consoles and mobile

Titanium Court, one of 2026's standout indie games, will no longer be confined to Steam. Publisher Fellow Traveller announced in an October 14 tweet that the game will arrive on Nintendo Switch, Switch 2, PlayStation 5, Android and iOS on October 29. Titanium Court debuted on Steam in April this year and won the Seumas McNally Grand Prize at the Independent Games Festival Awards. Developer AP Thomson also won the same award in 2025 as part of the team behind Consume Me. The game mixes roguelite, visual novel and RPG elements: a run is called a 'war' and is made up of multiple battles, each with two phases; the first phase is essentially a match-three game in which players move tiles representing wheat, rivers, forests and enemy castles, lining up three matching tiles to gain resources or destroy enemy strongholds while also setting up the second phase; in the second phase, enemies try to attack the player's base, which is also a square on the grid, and players can use resources to recruit soldiers and other workers to fortify their defense before the fighting begins. The game also limits how much players can trade in shops and may even block players from executing strategies they believe are sure wins; figuring out those limits is part of the fun. There is also a meta-game: the player is a new queen who has just arrived in this strange kingdom knowing nothing about it, and on the way home encounters baseball, Jenga and various pop-culture references. More than 21,000 games have been released on Steam this year, and Titanium Court's move to consoles and mobile means more players will have a chance to try this distinctive title. ^[3]

Where AI adoption and platform shifts meet

Putting these threads together, AI's next phase looks more like 'embedding' than 'replacing': it is entering merchants' existing ordering flows, Android's messaging and call security chain, and scientific computing processes such as weather forecasting. The common technical approach is a division of labor between on-device and cloud, layering capabilities onto existing systems, and making resource efficiency a selling point. For resource-limited users such as merchants and weather agencies, low-cost access may matter more than model size in determining how fast adoption happens. On the platform side, there is another line: Netflix is trimming staff while revenue grows, concentrating resources on high-impact areas; Titanium Court's expansion from a single platform to consoles and mobile also shows content players looking for a wider reach. What is worth watching next: whether Square Kiosk's hardware-plus-subscription model is accepted by food-service merchants, how widely Microsoft's open-source bias-correction tool is adopted by weather agencies, what the real interception results and privacy boundaries of Gemini scam defense look like as it expands to more devices, and where Netflix will direct new investment after the layoffs. ^{[1][2][3][4][5][6]}