Air Force flies a microreactor & U.S.-Hungary civilian nuclear cooperation - Tech News (Feb 17, 2026)

Please support this podcast by checking out our sponsors: - Invest Like the Pros with StockMVP - https://www.stock-mvp.com/?via=ron - Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad - Consensus: AI for Research. Get a free month - https://get.consensus.app/automated_daily Support The Automated Daily directly: Buy me a coffee: https://buymeacoffee.com/theautomateddaily Today's topics: Air Force flies a microreactor - Operation Windlord saw a U.S. Air Force C-17 airlift a microreactor system—eight Valar Atomics Ward250 modules—for DOE testing in Utah. Keywords: microreactor, C-17, Ward250, TRISO fuel, energy resilience. U.S.-Hungary civilian nuclear cooperation - The U.S. and Hungary signed a civilian nuclear cooperation deal on Feb. 16, aiming to shift Hungary toward U.S. fuel and small modular reactor partnerships. Keywords: Rubio, Orbán, SMRs, spent-fuel storage, Russian energy influence. Pentagon’s voice-controlled drone swarms - SpaceX and xAI were picked for a secretive Pentagon prize challenge to turn spoken commands into autonomous multi-drone swarm actions, raising human-in-the-loop concerns. Keywords: DIU, DAWG, autonomous weapons, voice control, Grok. AI boom triggers memory chip crunch - AI data-center buildouts are soaking up DRAM and HBM capacity, pushing prices higher and pressuring phones, cars, and consoles. Keywords: DRAM shortage, HBM, Nvidia racks, Micron, price spikes. AI job fears and policy gaps - Tech leaders’ automation timelines—like Mustafa Suleyman’s 12–18 month claims—are intensifying worker anxiety, while critics argue safety nets and ‘trigger’ policies lag behind. Keywords: white-collar automation, UBI, training, labor market, AI hype. Claude Code opacity sparks backlash - Anthropic changed Claude Code’s default logs to hide file names unless expanded, prompting developers to demand better auditability and security visibility. Keywords: agent transparency, file paths, terminal logs, verbose mode, developer trust. Gemini misuse and AI phishing - Google says threat actors tried to use Gemini for phishing, malware support, and influence ops—and warns AI makes phishing more polished and personalized. Keywords: GTIG, APT actors, AI phishing, social engineering, model distillation. Apple pushes deeper into video podcasts - Apple Podcasts will unify audio and video feeds this spring, add picture-in-picture, offline video downloads, and HLS-based dynamic video ad insertion. Keywords: Apple Podcasts, video podcasts, HLS, dynamic ads, creator distribution. ByteDance Seedance faces Disney threat - ByteDance says it will add safeguards to Seedance after Disney threatened legal action over alleged copyrighted character generation and viral clips. Keywords: Seedance 2.0, copyright, Disney, Marvel, Star Wars. Brain tech: BCIs and spinal repair - Two reality-checks in neurotech: BCIs remain bandwidth-limited and medical-first, while Northwestern’s mini spinal cord organoids show promising regeneration signals with ‘dancing molecules.’ Keywords: BCI limits, neural privacy, spinal cord organoids, microglia, regenerative nanomedicine. Episode Transcript Air Force flies a microreactor Let’s start with nuclear tech meeting logistics. The U.S. Air Force has carried out what’s being described as a first: a C-17 Globemaster III airlifting a nuclear reactor system. The mission—Operation Windlord—moved a microreactor package called the Ward250, built by Valar Atomics. In total, eight reactor modules are being transferred from March Air Reserve Base in Southern California to Hill Air Force Base in Utah, using three C-17s. From there, the components head to the Utah San Rafael Energy Lab in Orangeville for extended testing under the Department of Energy’s Nuclear Reactor Pilot Program—set up after President Trump’s Executive Order 14301 last year. The pitch is straightforward: microreactors could keep critical defense sites running even if local grids fail, and could power remote installations where grid access is weak or nonexistent. Technically, the Ward250 is framed as a next-generation design—helium coolant, graphite moderators, and TRISO fuel, which is basically uranium kernels wrapped in protective ceramic layers. Earlier reporting pegs the target around 100 kilowatts thermal, and Valar’s founder says the DOE selection is tied to a goal of reaching criticality on U.S. soil by July 4, 2026. One open question: why fly it instead of trucking it. The story hints security and established nuclear-transport certifications may have been the deciding factors. U.S.-Hungary civilian nuclear cooperation Staying with nuclear power, but shifting to geopolitics: the U.S. and Hungary signed a civilian nuclear cooperation agreement on February 16th. Secretary of State Marco Rubio went to Budapest, met Prime Minister Viktor Orbán, and signed a deal meant to kick off what the State Department called decades of cooperation—covering small modular reactors and spent-fuel storage. The strategic subtext is hard to miss. Hungary’s current nuclear fleet is deeply tied to Russian technology and fuel, and the U.S. is clearly trying to pull Central Europe’s energy orbit away from Moscow—and, by extension, Beijing. Under this agreement, Hungary is expected to buy nuclear fuel from American suppliers for the first time, and Holtec is slated to help with spent fuel management. It also lands two months before Hungary’s April elections, so the timing is politically charged on both sides. Pentagon’s voice-controlled drone swarms Now to defense tech, where AI and autonomy keep creeping closer to operational decision-making. According to people familiar with the process, SpaceX—and its fully owned AI subsidiary, xAI—have been selected to compete in a Pentagon prize challenge for voice-controlled autonomous drone swarms. The contest is reportedly a roughly $100 million prize pool with only a small number of participants. The objective: take spoken commands and translate them into digital instructions that coordinate multi-drone swarms across domains—air and sea are explicitly mentioned. The program is being run through the Defense Innovation Unit and a newer group under U.S. Special Operations Command. It’s organized in phases, starting with software and working toward real-world testing—eventually described in terms like “launch to termination.” What’s notable here is the scope. Some entries appear limited to a mission-control layer, but SpaceX and xAI are expected to participate across the full project. That’s raising eyebrows, particularly around how much generative AI should influence lethal systems, and how strong the human-in-the-loop safeguards really are. AI boom triggers memory chip crunch Let’s talk compute—specifically, the kind you can’t get enough of right now. A global memory-chip squeeze is starting to look less like a blip and more like a structural bottleneck. DRAM supply is tightening as AI data-center expansion hoovers up capacity, especially for high-bandwidth memory used alongside accelerators. The ripple effects are already landing in earnings calls and product planning. Apple has warned the shortage could compress iPhone margins. Tesla has said constraints may limit output—while Elon Musk floated the idea of building a memory fab, framing the choice as basically: hit the chip wall or make your own wall. Prices are moving fast—one DRAM category reportedly jumped about 75% from December to January—and some sellers are repricing daily. Meanwhile, memory makers are shifting lines toward HBM, which helps AI racks but leaves less “plain” DRAM for phones, PCs, cars, and consoles. Even entertainment hardware is getting dragged in—reports suggest Sony has contemplated pushing its next PlayStation deeper into the late 2020s, and Nintendo has weighed pricing moves. If you want the simplest mental model: AI accelerators don’t just need chips, they need oceans of memory beside them. One modern high-end rack can consume enough RAM to equal the memory in roughly a thousand premium smartphones. And new fabs take years—not quarters—to come online. AI job fears and policy gaps Against that backdrop, it’s no surprise the investing narrative is back to “AI infrastructure is on fire.” One market take making the rounds argues a handful of major players could spend on the order of hundreds of billions this year on AI data centers, with trillion-dollar annual infrastructure spending floated as a 2030 possibility. The specific winners people keep circling are predictable: Nvidia for GPUs and the CUDA ecosystem; Broadcom for custom AI ASIC work and networking; Micron because memory demand is spiking—especially HBM; and TSMC because it’s the manufacturing choke point for advanced nodes whether the world buys GPUs or custom chips. You don’t have to buy the hype to accept the near-term reality: hardware supply chains are becoming strategic terrain. Claude Code opacity sparks backlash Now, the human side of the AI transition—because the technical story keeps colliding with labor, trust, and plain old anxiety. Microsoft AI chief Mustafa Suleyman is the latest executive to put an aggressive timeline on white-collar automation, saying AI could reach human-level performance on many professional tasks and replace a lot of computer-based work within 12 to 18 months. Lawyers, accountants, project managers, marketing roles—he’s describing a broad blast radius. That rhetoric is landing at the same time more writers and engineers are saying, publicly, that the “vibes” around AI have changed. A recurring critique is that if industry leaders truly believe their own timelines, we should already see serious policy proposals—trigger-based training funding, wage insurance, automatic stabilizers, or some UBI-style contingency planning. Instead, what many people see is a mismatch: loud claims about disruption, and quiet follow-through on societal guardrails. On the career front, investors and operators are also describing a fast-diverging landscape: smaller teams doing bigger work with AI leverage, and a kind of “K-curve” where early movers toward frontier problems compound quickly. The common thread is that judgment—what to build, what to trust, what to ignore—is becoming the scarce skill, not the ability to type code quickly. Gemini misuse and AI phishing Speaking of code: developers are pushing back on a very specific, very practical problem—opacity in agent tools. Anthropic updated Claude Code so that, by default, it no longer shows which files the agent is reading or editing in the main progress output. Instead, it collapses the detail into summaries like “Read 3 files,” and you have to expand to see paths. That set off a strong reaction because file visibility isn’t just a convenience; it’s part of the security posture and the audit trail. If an agent starts pulling context from the wrong directory—or touching secrets—you want to catch it instantly, not after a scavenger hunt through hidden menus. Anthropic’s team says the change was meant to reduce noise and keep attention on diffs and command outputs. After backlash, they indicated they’re adjusting “verbose mode” to restore file-path visibility for reads and searches, while still suppressing other clutter. But the bigger takeaway is cultural: as agents get longer-running and more autonomous, the demand for clear, scrollable, plain-text accountability is only going up. Apple pushes deeper into video podcasts On the security beat more broadly, Google’s threat intelligence team has published a quarterly report describing how threat actors are already experimenting with AI assistants like Gemini for phishing, malware-related tasks, and influence operations. Google says it saw attempts linked to actors associated with China, North Korea, and Iran—ranging from crafting lures and building personas to troubleshooting malicious tooling. One point worth underlining: the most immediate risk for typical users isn’t some exotic AI super-attack; it’s better phishing. AI smooths out the classic red flags—broken grammar, awkward tone—and enables more personalized bait. Google also calls out “rapport-building phishing,” where attackers run multi-step conversations to build trust before delivering a payload. Meanwhile, on the enterprise side, security firms are racing to formalize best practices for connecting models to tools and data. Wiz, for example, is pushing guidance around securing Model Context Protocol setups—least privilege, locked-down servers, supply-chain hygiene, and human-in-the-loop controls. Even if you don’t care about the acronym, the trend is clear: AI integrations are becoming a new, busy attack surface. ByteDance Seedance faces Disney threat A quick pair of platform updates—one consumer-facing, one a reminder that the cloud is still the cloud. First, Apple says a more integrated video podcast experience is coming to Apple Podcasts this spring. The key change is unifying audio and video versions within the same show feed, with seamless switching, picture-in-picture, and offline downloads. Apple’s also leaning into HLS streaming support, which opens the door to dynamic video ad insertion—free distribution for creators and hosts, but an impression-based fee for ad networks that serve those dynamic ads. Second, Microsoft’s Azure marketing page for “Get to Know Azure” was intermittently unreachable, showing an Azure Front Door 504 gateway timeout. Not exactly a crisis, but it’s a useful reality check: even the biggest cloud platforms have moments where the front door says, ‘not today.’ Brain tech: BCIs and spinal repair Finally, a collision between generative video and old-school IP enforcement. ByteDance says it will tighten safeguards on Seedance, its AI video generator, after Disney threatened legal action over alleged copyright infringement. Seedance 2.0 launched recently and quickly produced viral, highly realistic clips—some appearing to depict Marvel and Star Wars characters. Disney’s claim is blunt: that the tool is being enabled by a pirated library of Disney-owned characters, and it wants it stopped. ByteDance says it respects intellectual property and will implement protections, but it hasn’t shared specifics. And this isn’t staying local—Japan has reportedly opened an investigation after AI-generated anime-character videos spread online. The near-term story here is less about one app and more about the shape of the next fight: provenance, training data disclosure, and whether “safeguards” mean meaningful prevention—or just slower whack-a-mole. Story 11 Before we wrap, one quick look at brain tech—because it’s a space where hype and reality are still miles apart. On brain-computer interfaces, the sober assessment hasn’t changed much: most BCIs still don’t “read thoughts.” They detect noisy electrical signals and infer narrow intent, usually in medical contexts where even slow cursor control can be life-changing. Bandwidth, latency, long-term implant durability, and neural-data privacy remain core bottlenecks. But in regenerative medicine, there’s a bright spot: Northwestern researchers built a lab-grown mini spinal cord organoid that includes microglia—immune cells that make injury inflammation more realistic—and used it to test a therapy they call “dancing molecules.” In the organoids, the treatment reduced scarring and inflammation and boosted neurite outgrowth, echoing earlier animal work. It’s early-stage, but it’s the kind of careful, testable progress that tends to matter. Story 12 That’s the tech landscape for February 17th: nuclear logistics as energy strategy, AI hardware straining memory supply, and a growing insistence that if automation is accelerating, transparency and governance need to accelerate too. Subscribe to edition specific feeds: - Space news * Apple Podcast English * Spotify English * RSS English Spanish French - Top news * Apple Podcast English Spanish French * Spotify English Spanish French * RSS English Spanish French - Tech news * Apple Podcast English Spanish French * Spotify English Spanish Spanish * RSS English Spanish French - Hacker news * Apple Podcast English Spanish French * Spotify English Spanish French * RSS English Spanish French - AI news * Apple Podcast English Spanish French * Spotify English Spanish French * RSS English Spanish French Visit our website at https://theautomateddaily.com/ Send feedback to [email protected] Youtube LinkedIn X (Twitter)