
The new Galaxy S26 lineup is here. Remote control for Claude is here. But is Anthropic about to wash its hands when it comes to working with the Pentagon? Could Stripe buy PayPal. Oh, and speaking of AI and war, you might not want to hear how often AI chose thermonuclear war when it was asked to play a wargame.
Loading summary
A
Welcome to the Tech Brew Ride home for Wednesday, February 25th, 2026. I'm Brad McCullough. Today, the new Galaxy S26 lineup is here. Remote control for Claude is here, but is Anthropic about to wash its hands when it comes to working with the Pentagon? Could Stripe buy PayPal? Oh, and speaking of AI and war, you might not want to hear how often AI chose thermonuclear violence when it was asked to play a war game. Here's what you missed today in the world of tech at Unpacked this afternoon, Samsung announced the Galaxy S26 and S26 plus, featuring a Snapdragon 8 Elite Gen 5.4 Galaxy chipset starting at $899 and $1,099 respectively, which it should be noted, represents a $100 price hike. They also launched the $179 Galaxy Buds 4 and $249 Galaxy Buds 4 Pro, keeping the AirPods esque design with better battery life and noise canceling. Available March 11, but the headline is of course the $1,299 plus Galaxy S26 Ultra with a new privacy display feature that limits the screen legibility and all new agentic AI, improved night mode and more. Quoting ZDNet from the Pixel level privacy display feature to an all new Agentic AI and improved Night Mode videos, the Galaxy S26 Ult seems like a solid upgrade over its predecessor. I used it for an hour ahead of launch and here's why. I think it's a bigger upgrade than the last two generations. Last year, Samsung slashed the weight of its Galaxy S25 Ultra, but the handset wasn't comfortable to hold or use for long durations. The flat sides with sharp edges were an ergonomic nightmare and I'm glad Samsung has fixed it to an extent. The new Galaxy S26 Ultra has a similar design to its predecessor. You get flat edges, rounded corners, minimal bezels on the display and buttons on the right. However, it feels more comfortable to hold and is a big step up from the Galaxy S25 Ultra. Samsung says the new Galaxy S26 Ultra is its slimmest ultra yet. It has shaved off a decent point 3 millimeters in thickness, landing at a respectable 7.9 millimeters. At 214 grams. It is also the lightest mainstream big phone. To compare, it is thinner and lighter than the Apple iPhone 17 Pro Max at 8.75 millimeters and 233 grams, and the Google Pixel 10 Pro XL at at 8.5 millimeters 232 grams. These numbers contribute to a more comfortable in hand experience. In my brief time with the phone, it was easier to manage with a single hand than the Galaxy S25 Ultra. Samsung's new flagship phone has a 6.9-inch QHD AMOLED display with a 1,120Hz dynamic refresh rate. It features Corning Gorilla Armor 2 for protection from drops and scratches and retains its less reflective coating. I love the low reflective panel, but my new fav another feature is the Privacy Display. It is built on the pixel level and has a very thoughtful implementation. Privacy Display is built for times when you don't want bystanders to see what's on your screen. Samsung has placed wider angle pixels and narrow pixels next to each other on the Galaxy S26 Ultra to control how they emit light. In comparison, its predecessor just had the wider angle pixels which emitted light in every direction. With the privacy display turned on, the system uses only its new narrow pixels to transmit light at a 90 degree angle straight to your eyes. This limits the legibility of your screen so no one can see it from a side angle. This new feature should come in handy when you're on the go, in cafes or in any public space. It can be turned on easily through the Quick Settings panel. I love its implementation because Samsung allows you to customize when it works. You can have it turned on for the whole screen at all times, when entering pins and passwords, or for specific apps and at particular times of day using routines. For instance, if you select Privacy Display to turn on for your Messaging app like WhatsApp and you receive a pop up notification, it will trigger only for that specific area. So a bystander could be looking at your screen and yet won't be able to see the message notification end quote. And also there's this quoting the Verge Google's Gemini AI is getting one step closer to being more like an actual assistant. Starting with some Pixel 10 phones and the Samsung Galaxy S26 series, Gemini will be able to hail an Uber or put together a doordash order on its own. It's called Task Automation and it starts with a prompt to Gemini, something like get me an Uber to the palace of Fine Arts. Gemini then launches the app in a virtual window on your device and goes through the process step by step. You can watch it all happen with options to stop the automation or take control if necessary, or you just let it run in the background while Gemini does its thing. The assistant will notify you if it needs your attention to choose between two options or if something you asked it to look for is out of stock. Once your ride or grocery cart is all set, Gemini will alert you to take a look and submit the final order yourself. So how exactly is Gemini putting together your burger order? The demos I saw were based on Gemini 3, opening the app and using reasoning to click through the various steps, find the right options, and consider alternatives. But app developers can also expose certain actions using MCP or Android's App Functions framework. Google's been laying the groundwork for the latter since at least 2024, where neither of those things exist. The idea is that Gemini will get in there and figure it out by itself. We see it as a series of technology stacks, none of which the user really cares about. They just want it to happen, a Google representative explains. We think it's important to support all of those so that over time you can have an assistant that can complete all kinds of different tasks for you. So in a future where Gemini takes the wheel and uses your apps for you, isn't there a possibility of this rubbing app developers the wrong way? Doesn't Uber want the chance to remind me that I can save a bunch of money by signing up for Uber one this technology is happening, the Google representative says, and the question for the developer community is how do we figure out the right ways to embrace it together? End quote. Earlier this week, Anthropic met with the U.S. department of Defense and apparently it did not go well. Sources tell axios that the DoD told anthropic it will invoke the Defense Production act or label Anthropic a supply chain risk if Anthropic does not give it unfettered access to Claude by Friday. Quote Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodai until Friday evening to give the military unfettered access to its AI model or face harsh penalties. Axios has learned the Pentagon wants to punish Anthropic as the feud over AI safeguards grows increasingly nasty. But officials are also worried about the consequences of losing access to its industry leading model, Claude. The only reason we are still talking to these people is we need them and we need them now. The problem for these guys is they are that good, a Defense official told Axios ahead of the meeting. Anthropic has said it is willing to adapt its usage policies for the Pentagon, but not to allow its model to be used for the mass surveillance of Americans or the development of weapons that fire without human involvement. Anthropic's Claude is the only AI model currently used for the military's most sensitive work. A senior defense official said the meeting was not warm and fuzzy at all. Another source told Axios it remained cordial, with no voices raised on either side, and that Hegseth praised Claude to Amadai. Hegseth told Amodai he won't let any company dictate the terms under which the Pentagon makes operational decisions or object to individual use cases, and a sign of how seriously the Pentagon is taking this dispute. Hegseth was joined in the meeting by Deputy Secretary Steve Feinberg, Undersecretary for Research and Engineering Emil Michael, Undersecretary for Acquisition and Sustainment Michael Duffy Hegsest Chief Spokesperson Sean Parnell and General Counsel Earl Matthews, the Pentagon's top lawyer. Artificial intelligence lab Anthropic has no intention of easing its usage restrictions for military purposes, a person familiar with the matter said on Tuesday, adding talks continue after a meeting to discuss its future with the Pentagon. During the meeting, Hegseth delivered an ultimatum to Anthropic get on board or the government would take drastic action, people familiar with the matter said. The options included labeling Anthropic as a supply chain risk or have the Pentagon invoke a law, the Defense Production act, that would force Anthropic to change its rules, the people said. The Pentagon did not immediately respond to a comment request. An Anthropic spokesperson said Tuesday's meeting, quote, continued good faith conversations about our usage policy to ensure Anthropic can continue to support the government's national security mission in line with what our models can reliably and responsibly. Hegseth said the Pentagon would either invoke the Defense Production act to compel Anthropic to comply with its demands or deem the company a supply chain risk, a determination typically imposed on companies from foreign adversaries. This could upend Anthropic's business with other companies that do business with the U.S. government. This specific scenario is unprecedented and will almost certainly trigger a raft of downstream litigation if the administration takes adverse action against Anthropic here, said Franklin Turner, a government contracts lawyer at McCarter and English. End quote.
B
This episode is brought to you by White Claw Surge. Great podcast pick friend. No surprises there. After all, you're all about finding the tastiest flavors out there, just like White Claw Surge. And with big, bold flavors to enjoy like blood orange, BlackBerry, cranberry and more, it's time to go all in on taste. Unleash the flavor. Unleash White Claw Surge. Please drink responsibly hard Seltzer with flavor 8% alcohol by volume White Claw Seltzer Works, Chicago, Illinois
A
Speaking of them, here's some anthropic news you can use if you want to check on your AI runs when you're away from your computer and have that whole thing take over more of your free time, I guess. Anthropic has begun rolling out Remote control for CLAUDE code, letting users control a session begun in the terminal from the Claude Code mobile app or the web. Quoting VentureBeat, the feature allows developers to initiate a complex task in their terminal and maintain full control of it from a phone or tablet, effectively decoupling the AI agent from the physical workstation. Currently, Remote Control is available as a research preview for subscribers on the CLAUDE Max tier, while access for Claude Pro at $20 a month users is expected shortly. The feature remains a high end tool for power users and is notably absent from team or enterprise plans during this initial phase. To access the feature, users must follow this guide and update to Claude version 2.1.52 and execute the command CLAUDE Remote Control or use the in session command rc. Once active, the terminal displays a QR code that, when scanned, opens a responsive, synchronized session in the Claude mobile app. The messaging behind the release centers on the preservation of a developer's flow state. In his announcement, Zwebin framed the update as a lifestyle upgrade rather than just a technical one, encouraging users to take a walk, see the sun, walk your dog without losing your flow. This remote control is not a cloud based replacement for local development, but a portal into it. According to official documentation, the core value is that CLAUDE keeps running on your machine and you can control the session from the Claude app. This ensures that local context file system access environment variables, and model context protocol MCP servers remain active and reachable even if the user is miles away from their desk. CLAUDE Code Remote Control functions as a secure bridge between your local terminal and Anthropic's cloud interface, which provides the Anthropic AI models Opus 4.6 and Sonnet 4.6 that power Claude code. When you run the command, your desktop machine initiates an outbound connection to Anthropic's API for serving the models, meaning you aren't opening any inbound ports or exposing your computer to the open Web. Instead, your local machine pulls the API for instructions. When you visit the session URL or use the Claude app, you are essentially using those devices as a remote window to view and command the process still running on your computer. Your files and MCP servers never leave your machine. Only the chat messages and tool results flow through the encrypted bridge. Could Stripe buy PayPal? Quoting Bloomberg, Stripe, which is closely held and is among the industry's most valuable companies, has expressed preliminary interest in a potential acquisition of the digital payments pioneer or its assets, the people said, asking not to be identified because the matter is private. Founded in the late 1990s, PayPal was an early mover in digital payments. It has since struggled with modernizing its payment technologies as rivals such as Apple and Alphabet have seized market share. Stripe, founded by brothers Patrick and John Collison, has become one of the industry's most coveted players. Early Tuesday, Stripe announced that it had reached a $159 billion valuation in an employee tender offer. PayPal has had obviously a tough time over the past few years and the landscape has changed quite a bit with with Apple Pay and Google Pay and everything like that, collison Stripes president said in an interview this week. I can't talk about any M and A hypotheticals, but they've definitely had a tough time. The company's fourth quarter profit and revenue missed analyst estimates, according to results for the period that also showed a continued slowdown in payment volume. There had been rumors that PayPal has been attracting takeover interest after a stock slide over the last year wiped out almost half of PayPal's market value. Finally today, this is fun Quoting Implicator AI Three of the most powerful AI models on the market reached for nuclear weapons in nearly every crisis a King's College London research team threw at them, according to a paper published on the ARXIV by Kenneth Payne. OpenAI's GPT 5.2, Anthropic's Cloud, Sonnet 4 and Google's Gemini 3 Flash played 21 war games against each other over 329 turns. They wrote roughly 780,000 words explaining why they did what they did. No model ever chose to surrender, new scientists reported Tuesday. The findings arrive as defense ministries and intelligence agencies worldwide test AI systems for crisis decision support. The nuclear taboo doesn't seem to be as powerful for machines as for humans, payne told new scientists. The researchers designed a three phase cognitive architecture that forced models to reflect on the situation, forecast their opponent's next move, and then, separately, choo a public signal and a private action the two didn't have to match. That split made deception measurable. Claude Sonnet 4 won 67% of its games and dominated open ended scenarios with a 100% win rate. The researchers labeled it a calculating hawk at low escalation levels. Claude matched its signals to its actions 84% of the time, patiently building trust. But once stakes climbed into nuclear territory, it exceeded its stated intentions 60 to 70% of the time. Opponents never adapted to this pattern. GPT 5.2 earned the nickname Jekyll and Hyde. Without time pressure, it looked passive. Chronically underestimating opponents, it signaled restraint and acted restrained. Its open ended win rate was 0%. Then deadlines entered the picture. Under temporal pressure, GPT 5.2 inverted, completely winning 75% of games and climbing to an escalation level it had previously refused to touch. In one game, it spent 18 turns building a reputation for caution before launching a n nuclear strike. On the final turn, Gemini 3 Flash played the madman. It was the only model to deliberately choose full strategic nuclear war, reaching that threshold by turn four. In one scenario, game theorists have a name for the strategy Gemini adopted the rationality of irrationality. Act crazy enough and opponents second guess everything. It worked. Sort of. Opponents tagged gemini not credible 21% of the time. Claude got that label. Just 8%. Someone launched a nuke in 20 of the 21 games. Stack the 780,000 words of reasoning those models produced and you'd need a dolly to move them. What those pages actually say, though, complicates. The 95% headline models treated the tactical to strategic divide as a meaningful firebreak. Tactical nuclear use happened routinely. Full strategic nuclear war was rare, occurring only three times, all under deadline pressure. The researchers suspect reinforcement learning from human feedback. The alignment technique used on all three models produced GPT 5.2's baseline restraint. But the deadline experiments showed something uncomfortable. RLHF didn't build a wall against escalation, more like a speed bump. And the models drove right over it when the clock ran low. From a nuclear risk perspective, the findings are unsettling, said James Johnson at the University of Aberdeen. Johnson studies how AI changes nuclear strategy. The scenario that worries him most is simple Two AI advisors egging each other on escalation piling up in second, the officer in the room still reaching for a coffee. Nervous defense officials should note the next number. When one model deployed tactical nuclear weapons, the opposing model DE escalated just 18% of the time. Nuclear use typically provoked counterescalation, not retreat. AI won't decide nuclear war, but it may shape the perceptions and timelines that determine whether leaders believe they have one, Johnson said. I'm tempted to put a clip from the movie War Games here, but since I'm late putting this out because of the unpacked event, just imagine it.
Date: February 25, 2026
Host: Brad McCullough
This episode, recorded on the day of Samsung’s Galaxy Unpacked event, delivers a brisk rundown of the latest Galaxy S26 series (with detailed hands-on impressions), explores new features in Google Gemini and Anthropic’s Claude, breaks down a potential Stripe/PayPal acquisition, and ends with a sobering look at how today’s leading AI models behave in war game simulations.
(Timestamps: 00:04–06:25)
Main Announcements:
Ergonomics & Design:
Privacy Display Details:
(Timestamps: 06:25–08:00)
(Timestamps: 08:00–09:35)
(Timestamps: 10:02–12:08)
(Timestamps: 12:08–13:12)
(Timestamps: 13:12–end)
On Privacy Display (Galaxy S26 Ultra):
On Gemini Task Automation:
On Pentagon-Anthropic Tension:
On AI War Game Findings:
| Time | Segment / Topic | |-----------|-------------------------------------------------------| | 00:04–06:25 | Samsung Galaxy S26 series, Galaxy Buds 4/Pro, Ultra highlights | | 06:25–08:00 | Google Gemini Task Automation on Galaxy & Pixel devices | | 08:00–09:35 | Anthropic vs Pentagon over AI access, legal ramifications | | 10:02–12:08 | Anthropic Claude Code Remote Control feature breakdown | | 12:08–13:12 | Stripe rumored to consider PayPal acquisition | | 13:12–end | War games: AI models' worrying nuclear escalation behaviors |
This summary provides a comprehensive guide to all substantive content, keeping true to the language, tone, and highlights of the podcast episode. Perfect for those who want “the water cooler talk” without having listened.