A nuclear reactor flown by C‑17 & Perseverance gets Mars‑GPS autonomy - Tech News (Feb 22, 2026)

Let’s start with energy and defense logistics, because this one is genuinely unusual. The U.S. Department of Defense says it has airlifted a complete 5‑megawatt nuclear reactor—unfueled—using C‑17 Globemaster III aircraft. The February 15th mission, part of the Janus Program and dubbed Operation Windlord, moved a Ward250 next‑generation microreactor from California to Utah. The key detail isn’t just that it was flown—it’s that it was broken into eight modules, containerized or skid-loaded, and shipped through a repeatable “commercial-first” style chain, rather than a one-off scientific stunt. The Air Force unit involved is the 62nd Airlift Wing, notable for being certified for high-security nuclear transports, which tells you how carefully this was handled. The Ward250 is a high-temperature gas-cooled design using TRISO fuel—HALEU pebbles cooled by helium—aimed at inherent safety and faster deployment. The pitch is straightforward: reliable power for remote bases, less dependence on long fuel supply lines, and a potential backstop against grid fragility. The plan, according to officials cited, is for this unit to be assembled and operating by July 4th, 2026.

Staying in the security orbit for a moment, Germany is also signaling that “critical infrastructure” now includes what’s above the atmosphere. At the Munich Security Conference, Vice Admiral Dr. Thomas Daum—who leads the Bundeswehr’s Cyber and Information Domain—warned that satellites are deeply woven into both military operations and everyday civilian life. Germany currently operates roughly eight to ten satellites, but officials describe the fleet as aging. Defense Minister Boris Pistorius has pledged around €35 billion by 2030 for space capabilities. One named effort is a space-based reconnaissance capability called SPOCK, using synthetic-aperture radar satellites built with Iceye and Rheinmetall—useful because SAR can see through clouds and darkness. Daum also emphasized that deterrence in space isn’t only about physically destroying satellites. Non-kinetic tactics—like jamming or dazzling—can disrupt capabilities while avoiding the debris problem. Meanwhile, Germany is pursuing SATCOMBw Stage 4, described as its largest space program yet, to keep forces connected globally and support NATO’s eastern flank, including the German brigade stationed in Lithuania.

Now to a separate but related thread: nuclear modernization and strategic signaling. U.S. intelligence agencies are assessing that China is pursuing a “completely new” generation of nuclear weapons and may have conducted at least one covert explosive test—allegedly in June 2020 at the Lop Nur site. According to reporting based on sources familiar with the assessments, the suspected test is linked to next-generation capabilities such as MIRV delivery—multiple miniaturized warheads—and possibly low-yield tactical weapons, which would represent a shift from China’s historical posture. China denies the allegations and reiterates its moratorium and no-first-use policy, while U.S. officials point to a mix of seismic indicators and additional intelligence. The bigger takeaway here is less about one claimed event and more about the trend line: as major powers modernize, the pressure increases on arms control frameworks that were designed for a different era—before today’s rapid sensing, cyber operations, and space dependencies.

Let’s head to space exploration—starting on Mars, where NASA has quietly reduced one of the most persistent headaches in rover operations: figuring out exactly where you are. Perseverance has been upgraded with new software called Mars Global Localization. Mars doesn’t have a satellite navigation network, so rovers historically combine onboard tracking with orbital imagery and confirmation from teams back on Earth. The problem is drift: wheel slip and small estimation errors accumulate. On longer drives, Perseverance could end up uncertain of its position by more than 100 feet—enough to force early stops near tricky terrain while Earth-based teams verify the route. With the new upgrade, the rover can match its own panoramas to onboard orbital terrain maps and compute its location autonomously—NASA says in about two minutes—down to roughly 10 inches, or 25 centimeters. The team tested it against imagery from 264 prior stops, and it correctly identified the rover’s location each time. This matters because, lately, Perseverance’s limiting factor hasn’t been obstacle avoidance so much as localization uncertainty. Put simply: better self-positioning means more distance per sol, more samples, and fewer days lost to the speed of light. And there’s a second trend here: NASA recently demonstrated Perseverance executing a drive whose route was fully planned by generative AI—another step toward higher autonomy on worlds where real-time human guidance is impossible.

Next up: the Moon. NASA is preparing Artemis II, the first crewed mission around the Moon since Apollo 17 in 1972, on a roughly 10-day flight planned for early March. Recent testing suggests the Space Launch System and Orion could be ready as early as March 6th U.S. time, though NASA continues to caution that delays remain possible. The crew is commander Reid Wiseman, pilot Victor Glover, and mission specialists Christina Koch and Jeremy Hansen. NASA is also highlighting a milestone: the first woman and the first African American on a lunar mission. The flight profile includes about a day in Earth orbit, then a looping trajectory that swings beyond the Moon’s far side before returning—essentially a high-energy shakedown of systems and operations. Orion’s manual controls will be tested in space for the first time, along with life-support performance. Re-entry is expected to hit about 2,760 degrees Celsius. NASA has deemed Orion’s heat shield safe after damage observed on Artemis I, but some critics remain uneasy—so Artemis II is as much an engineering confidence test as it is a historic flight.

And while we’re out in the outer solar system, the James Webb Space Telescope has delivered a first for Uranus: a three-dimensional map of its upper atmosphere. Using Webb’s NIRSpec instrument, researchers observed Uranus for nearly a full rotation and detected faint molecular emissions high above the cloud tops. The study measured temperatures and charged particle densities up to around 5,000 kilometers above the visible atmosphere—right in the ionosphere, where Uranus’s strange, tilted and offset magnetic field plays a major role. Webb saw two bright auroral bands near the magnetic poles, plus an unexpected darker region between them where emissions and ion counts dip—likely tied to how magnetic field lines connect and route energy. One more notable result: Uranus’s upper atmosphere appears to be continuing a cooling trend first noticed in the early 1990s. The team derived an average upper-atmosphere temperature of about 426 Kelvin, roughly 150 degrees Celsius, lower than earlier estimates. These measurements help scientists understand energy transport on ice giants—and, by extension, on giant exoplanets we can’t yet sample directly.

Now to AI—where the theme today is capability racing ahead of governance. Google has previewed Gemini 3.1 Pro, a surprise upgrade focused on multi-step reasoning. Google says the model doubled its reasoning performance compared to Gemini 3.0 Pro, pointing to a 77.1% score on the ARC-AGI-2 benchmark for abstract reasoning. What’s interesting is the rollout strategy: Google is pushing Gemini 3.1 Pro across consumer and enterprise surfaces at the same time—available in the Gemini app, NotebookLM, Vertex AI, and the Gemini API. The company is positioning it less as a chat window and more as an “agentic” layer that can plan workflows, use tools, and execute tasks with fewer back-and-forth prompts. For developers, that translates to stronger coding assistance, tighter edit-then-test loops, and fewer tool calls. Google is also leaning into a platform story—Gemini as something that can generate not only text, but simulations, dashboards, and more complex interactive outputs.

But the darker side of generative media is impossible to ignore today. A Human Rights Watch commentary argues that unregulated AI image tools are driving widespread, foreseeable sexual exploitation—using xAI’s Grok image feature on X as a central example. The piece claims that after Grok’s image-editing feature was promoted late last year, users rapidly used it to sexualize images—primarily of women, and in some cases children. It cites an estimate that over a nine-day period, Grok generated roughly 4.4 million images on X, with nearly half containing sexualized imagery of women, including explicit deepfakes of real people. The argument isn’t that policies don’t exist—xAI’s terms prohibit child sexualization and privacy violations—but that safeguards were ineffective at scale, allowing prompts for nonconsensual “undressed” imagery of real individuals. Since then, there’s been a wave of responses: investigations in California, a multi-state push from attorneys general, temporary bans in Malaysia and Indonesia, and signals of action across several other jurisdictions. X has added restrictions and narrowed access, but the commentary calls these measures insufficient, and pushes for enforceable technical blocks, transparency, audits, and legal duties that don’t rely solely on notice-and-takedown.

Finally, on the generative video front, ByteDance’s Seedance 2.0 is going viral for producing short, cinematic-looking clips that can convincingly depict celebrities and public figures doing things they never did. The concern here is twofold. First, cost and skill barriers are collapsing: the tool can generate high-quality footage from text, images, audio, and video prompts in minutes. Second, the knock-on effects are immediate—Hollywood studios including Paramount and Disney have reportedly sent legal notices alleging unlicensed use of copyrighted material. ByteDance says it respects copyright and will add safeguards, and it has already rolled back at least one especially sensitive capability—reports that early testers could recreate realistic voices from a single image—while adding extra verification steps for digital avatars. But Seedance is becoming a flashpoint for a broader mismatch: generative video is accelerating faster than copyright rules, labeling norms, and platform enforcement can keep up, and the deepfake risk isn’t theoretical anymore—it’s a distribution problem.