Sapper Jäger 🇺🇸 🇩🇪 🇺🇦 🥷 🔥

@TheRecord_Media "Section 702 only permits the targeting of non-United States persons who are reasonably believed to be located outside the United States. United States persons and anyone in the United States may not be targeted under Section 702." intel.gov/foreign-intell…

Top intel officials urged lawmakers to renew for 18 months a spy law known as Section 702 of the Foreign Intelligence Surveillance Act (FISA), which enables broad electronic surveillance of the communications of overseas security threats therecord.media/us-intel-chief…

@USARPAC @25thID @4thInfDiv @EighthArmyKorea @7thID @USARJ_PAO @I_Corps @11thAirborneDiv @311thSignalCmd @2INFDIV How are we doing this?

Welcome to the USARPAC Unit Logo Madness! In honor of #MarchMadness, let's see who will win the crown of best logo!

@alifcoder @AnthropicAI Is this true? from you?

🚨 Anthropic just shared the exact file their Claude Code team uses internally. It’s called CLAUDE.md. Drop it into your project and Claude will: → Plan before coding → Learn from corrections → Verify before shipping → Fix bugs end-to-end This changes how you build with AI.

It's kind of weird that Meta didn't announce this. Seems like it could be a big deal, maybe one that pressures other AI companies to follow suit. wired.com/story/signals-…

WTAF is wrong with some people!?!

@rryssf_ always good to highlight when a paper is still pre-print, and/or not yet peer reviewed

Your AI has been quietly forgetting everything you told it. Not randomly. Not loudly. Systematically. Starting with the decisions that matter most. > The constraint you set three months ago "never use Redis, the client vetoed it after a production incident." Gone. The GDPR deployment region restriction. Gone. The retry limit you tested empirically after the cascade failure. Gone. > The model never told you. It just started using defaults. > This is called context rot. And Cambridge and Independent researchers just quantified exactly how bad it is. > Every production AI system that runs long enough will eventually compress its context to make room for new information. That compression is catastrophically lossy. They tested it directly: 2,000 facts compressed at 36.7× left 60% of the knowledge base permanently irrecoverable. Not hallucinated. Not wrong. Just gone. The model honestly reported it didn't have the information anymore. > Then they tested something worse. They embedded 20 real project constraints into an 88-turn conversation the kind of constraints that emerge naturally in any long-running project then applied cascading compression exactly like production systems do. After one round: 91% preserved. After two rounds: 62%. After three rounds: 46%. > The model kept working with full confidence the entire time. Generating outputs that violated the forgotten constraints. No error signal. No warning. Just silent reversion to reasonable defaults that happened to be wrong for your specific situation. > They tested this across four frontier models. Claude Sonnet 4.5, Claude Sonnet 4.6, Opus, GPT-5.4. Every single one collapsed under compression. This isn't a model problem. It's architectural. → 60% of facts permanently lost after single compression pass → 54% of project constraints gone after three rounds of cascading compression → GPT-5.4 dropped to 0% accuracy at just 2× compression → Even Opus retained only 5% of facts at 20× compression → In-context memory costs $14,201/year at 7,000 facts vs $56/year for the alternative The AI labs know this. Their solution is bigger context windows. A 10M-token window is a larger bucket. It's still a bucket. Compaction is inevitable for any long-running system. The window size only determines when the forgetting starts not whether it happens.

@thetect0nic @TheChiefNerd wonder if this is a case of choosing to "resign" with apparent moral-high-ground, before getting fired and taking a huge credibility loss

Kent was under investigation before he resigned. The White House knew. They cut him from intelligence briefings months ago. They asked Gabbard to fire him. She didn't. Then he resigned publicly on official letterhead and 70 million people read it. The investigation doesn't change what Kent said. It changes the question of why the White House let him stay in the role while he was already under investigation and what that says about how the decision to go public was calculated.

TUCKER: “Was Iran on the verge of getting a nuclear weapon?” JOE KENT: “No. They weren't three weeks ago when this started, and they weren't in June either.”

@SNAFU_Sara @UPS they once asked me to prove that it was delivered to the wrong house the pic they sent doesn't look anything like my house they said "well, how do we know that's true"

Hey @UPS you delivered a package of mine to the wrong house, and. I try to file a claim it tells me the package number is wrong, it’s literally autofilled from what you gave me. This is time sensitive too. 😡

Before concluding that operational art is in decline, planners should ask whether the problem has outgrown the lens they are using. ow.ly/Cnca30sUq5T

Hannah Lamb article "Angle of Attack: Apache Attack Helicopters in Unmanned Skies"  has been recognized by the Chief of Staff of the Army as a recommended article. In this article, Lamb examines the evolving role of the AH-64 Apache in an era increasingly defined by unmanned systems. Rather than framing drones as a replacement, she highlights the growing importance of manned-unmanned teaming (MUM-T) and the need to integrate legacy platforms with emerging technologies to maintain battlefield relevance. As militaries adapt to contested, drone-saturated environments, this analysis offers a clear-eyed perspective on how traditional capabilities can evolve without becoming obsolete. A well-deserved recognition for an important contribution to the future of warfare. Read the full article here: irregularwarfare.org/articles/angle… #IrregularWarfare #MilitaryInnovation #ArmyAviation #Drones #EmergingTech #NationalSecurity

Claude Code Essentials is now on ExamPro. exampro.co/exp-claudecode… If you support the paid materials on ExamPro you will get a 20% discount to our Claude Certified Architect course that we will have out next week. ExamPro Subscribers get the full access.

@orbatmapper very helpful!

The text-to-ORBAT tool now has a scratchpad area where you can mothball your units. Just drag units there for safekeeping. You can also rearrange them before you export them.

We invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do. Nearly 81,000 people responded in one week—the largest qualitative study of its kind. Read more: anthropic.com/features/81k-i…

@halbritz that was when John Boehner (R-OH) and Paul Ryan (R-WI) controlled the House

"Mullin told the committee that while serving in the House in 2015, he was asked to train with 'a small contingency' and go to a 'certain area.' ... Did Congress assign him the mission, Paul asked. Mullin said yes. Peters asked: Who specifically assigned it? Mullin responded that he could not say. Paul asked which agency classified the mission. Mullin responded that the House classified it." #cmmwca4ra00003b60s2c0bubl" target="_blank" rel="nofollow noopener">cnn.com/politics/live-…

EU unveils coordinated strategy to counter cyber, sabotage and disinformation threats amid rising hybrid attacks Industrial Cyber industrialcyber.co/news/eu-unveil…

People used to pay me on on average to build their apps for $30K USD in less than a month. I was the go-to-guy because I knew how to build exactly what they wanted with little instruction and I always met their insane deadlines. I was known as "the machine", I wasn't the 10x developer I was the 100x developer. But it took its toll on my health and I left that do tech education. Claude Code can now do it. No computer science degree. No dev team. No $30K invoice. But most non-tech people are still stuck on the outside looking in because there is a tech knowledge gap. What if someone showed you all the steps, used plain speak, provided technical strategy, technical pathing, technical guidance, made it practical, and takes you to production. That's what my Claude Code from Zero course would be. But I'll be honest with you, this won't be a Udemy $20 priced course. It will be $300, $500 or $1200 depending on your needs. But do you want me to build it?

Microsoft is a ~$4 trillion company, yet it’s promoting Windows 11 with an AI-generated image that literally shows two Start buttons on the taskbar. It takes less than five seconds to capture a real screenshot of the Windows 11 desktop. Perhaps the marketing team at Microsoft isn’t even using Windows, so they rely on Copilot to generate slop.

@Defence_blog such a sexy beast!

U.S. Air Force upgrades U-2 electronic warfare system defence-blog.com/u-s-air-force-…

@TheHackersNews requires you give up your valid Email to receive it

AI comes with potential risks and vulnerabilities, but you can protect your workers and your organization. One of the best places to start is with a comprehensive AI usage policy. This template provides: ✅ A definition of artificial intelligence ✅ A breakdown of acceptable and prohibited AI use ✅ Customizable guidelines for training, human oversight, accountability, and amendments 🔗 Get your AI employee usage policy template → thn.news/ai-policy-guide

that = 18,550 voters that did NOT vote for him I wonder what will happen to them