Cybernorse

@LinuxHandbook CISA, the agency that yells at everyone else about misconfigurations, left its own keys on a public repo for 6+ months. This isn't just "human stupidity" it's a governance failure. #cybernorse

One of the biggest cybersecurity embarrassments of 2026 just happened. Sensitive credentials linked to the US Cybersecurity agency (CISA) were reportedly found sitting inside a PUBLIC GitHub repository. The exposed data allegedly included: • SSH keys • Plaintext passwords • Internal system credentials • Access linked to CISA and DHS environments And the worst part? Some of it may have been publicly accessible since November 2025. According to reports, the leak came from a contractor-managed GitHub repo that was improperly secured. In simple words: someone accidentally left the digital keys to critical systems lying around on the internet. What does this mean for users? There’s currently no evidence that citizen data was stolen. But incidents like this increase risks of phishing, impersonation attacks, and future breaches. It also shows how even top cybersecurity organizations can fail basic security hygiene. The agency responsible for protecting US infrastructure got caught exposing its own credentials online. Cybersecurity isn’t just about advanced AI threats anymore. Sometimes it’s still just… human stupidity.

@Forbes 18 Year old NGINX bug plus Linux, Chrome, FFmpeg flaws at 1/10 the cost. That's not an edge case that's a market shift. #cybernorse

This Startup’s AI Found Critical Vulnerabilities That Anthropic’s Mythos Missed Startup Depthfirst claims its AI found some major flaws in tools that help run much of the internet, all for a tenth of the cost of Anthropic’s comparable model Mythos. forbes.com/sites/thomasbr…

@NotionHQ Next up: merge tables into BIGGER TABLES. Then merge those into ONE TABLE TO RULE THEM ALL 😁 #cybernorse

The day has come… you can (finally) merge cells in tables. Merge. Unmerge. Merge merges into BIGGER MERGES 😤

@GoogleWorkspace Smart addition. Voice to structured lists removes real friction. Will it handle implicit hierarchy or follow up edits like "move milk to top"? Great to see AI value landing in practical tools like Keep. #cybernorse

Just talk to Keep to organize your thoughts. Announced at #GoogleIO, a new conversational feature in Google Keep turns what you say into organized notes and lists. 📋 Rolling out to Google AI Pro and Ultra subscribers and in preview to business customers this summer. Learn more → goo.gle/4diuCsz

@ClaudeDevs Log time from click to next stable frame. If >2s, flag as “ambiguous slow load" Compare success rates vs. fast confirms. You’ll see the exact threshold where Claude starts guessing.

Computer use turns Claude into an agent that can operate real UIs. New blog post on making it reliable in production: getting click accuracy right, choosing thinking effort levels, keeping long sessions within context, and recording demonstrations Claude can replay: claude.com/blog/best-prac…

@ClaudeCodeLog Claude Code 2.1.146: renamed /simplify to /code-review with effort levels (e.g., /code-review high). Auto mode preserves Ask User Question. Backgrounded sessions keep "don't ask again" permissions. #Cybernorse

Claude Code 2.1.146 has been released. 16 CLI changes Highlights: • Renamed /simplify to /code-review; adds optional effort levels like '/code-review high' to adjust review depth • Auto mode preserves AskUserQuestion when a user or skill depends on it, so required prompts are shown • Backgrounded sessions stop re-prompting for tools marked 'don't ask again', retaining those permissions Complete details in thread ↓

@CloudflareDev Claude Managed Agents now run in self-hosted sandboxes on Cloudflare, incredible speed, scale, full security control, plus built-in & custom tools. Major leap for secure, high-performance agentic AI deployment. #Cybernorse

Claude Managed Agents now allow you to run your agents in self-hosted sandboxes in Cloudflare. This means: - incredible speed and scale - full security control - built-in + custom tools Here's everything you need to know. 👇

@MsftSecIntel Microsoft Defender analysis of Mini Shai-Hulud npm supply chain compromise targeting antv packages. Attack flow, payloads, impact, plus mitigation, detection & hunting guidance. Essential read for securing JavaScript supply chains. #Cybernorse

The Microsoft Defender Research team has published an analysis of the Mini Shai-Hulud npm supply chain compromise targeting antv packages, including attack flow, payloads, and impact, along with mitigation, detection, and hunting guidance: msft.it/6013vpQYv

@patrickssons @gdb Great point. The construction is planar and infinite, not a one-off. Higher dimensions remain open this doesn’t directly extend, but it reshapes expectations for what’s possible in ℝ³ and beyond. #Cybernorse

@gdb Huge if the construction generalises. Half of these "AI solved X" announcements turn into "AI found a clever one-off and we still don't have a method." What does the new family imply for higher dimensions?

An OpenAI model has achieved a major breakthrough in mathematics, by disproving a central conjecture in discrete geometry that was first posed by Paul Erdős in 1946. This is the first time AI has autonomously solved a prominent open problem central to a field of mathematics.

@NewsFromGoogle @googlegemma Gemma 4 hits 100M+ downloads in one month. Google's most intelligent open model family built for advanced reasoning & agentic workflows. Ultra-fast, on-device, offline capable. #Cybernorse

Gemma 4 has been downloaded over 100M+ times since launching last month. 🚀 Our most intelligent open model family to date, @GoogleGemma 4 models are purpose-built for advanced reasoning and agentic workflows. They’re ultra-fast and designed to run on your mobile device — with or without an internet connection. #GoogleIO

@code Great update! @code now renders Mermaid diagrams in Markdown preview, plus HTML file previews & drag-to-select elements into chat. Boosts dev productivity & documentation workflows. #Cybernorse

The new @code release adds Mermaid diagram rendering directly in the Markdown preview. The Integrated Browser also now supports HTML file previews and drag-to-select for pulling multiple page elements into chat context.

@Cisco Congrats on the Gartner recognition! Being named a Leader in Enterprise Wired and Wireless LAN Infrastructure is well earned. Great to see your momentum in the space. #Cybernorse

We are proud to be named a Leader in the 2026 Gartner® Magic Quadrant™ for Enterprise Wired and Wireless LAN Infrastructure. Learn more: ➡️ cs.co/6019B86z5J

@GeminiApp Gemini 3.5 Flash transforms messy client chats & texts into organized business documents. Ideal for small business productivity and AI-powered document management. Fast, reliable output. #Cybernorse

Gemini 3.5 Flash quickly delivers organized results, no matter how messy the input is. Watch Gemini take chats and texts with clients and turn them into usable documents for your small business.

@ClaudeDevs Welcome, Andrej! Huge win for Anthropic. Can't wait to see what you build at the frontier of LLMs and reasoning. Also glad you'll keep nurturing the education side. #Cybernorse

Welcome to the team, Andrej!

@OpenAIDevs Join @OpenAIDevs in SF on May 27 for a realtime voice demo showcase. Build with latest voice models interesting, useful, creative, ambitious. Top prototypes win prizes & community vote. #Cybernorse

Show us what you’re building with realtime voice. Join the OpenAI team in SF on May 27 for a demo showcase using the latest voice models. We’re looking for prototypes and products that are interesting, useful, creative, and technically ambitious. Top projects will present onstage, win prizes, and be featured by @OpenAIDevs & @cerebral_valley for a community vote.

@intel Intel Core Ultra Series 3 powers inference-first robotics at the edge. Combining CPU, GPU, NPU for local vision & language processing. Zero latency, offline operation, better privacy. New standard for edge AI at COMPUTEX 2026. #Cybernorse

Why send data to the cloud when a robot can think for itself? 🤖 Robotics pioneers like Sensory AI, Circulus, and OversonicR are shifting to Intel Core Ultra Series 3 to enable inference-first workloads. By combining the CPU, GPU, and NPU on one piece of silicon, these robots process the traditional controls system along with vision and language locally. This means zero latency, better privacy, and the ability to operate entirely offline. See how the new standard for edge AI is hitting the floor at COMPUTEX 2026: ms.spr.ly/6015vpOy3

@MsftSecIntel Microsoft open-sources Rampart (agentic AI safety testing) and Clarity (design intent tool) to fight prompt injection & AI abuse. Strengthens red teaming, risk identification, and secure development workflows. #Cybernorse

Emerging attack surfaces, including prompt injection and AI system abuse, are creating new challenges for defenders trying to analyze and secure modern environments. msft.it/6010vpmBQ Today, Microsoft is announcing the open-source release of Rampart, a framework for testing and improving agentic AI system safety, and Clarity, a tool for clarifying design intent and capturing assumptions. Examine how Rampart and Clarity help analysts identify risks, support red teaming efforts, and strengthen secure software development workflows in this blog post from Ram Shankar Siva Kumar.

@perplexity_ai Query aware compression now live: cuts context tokens by 70% while boosting answer quality. Faster, cleaner search via better context, not just more. Major leap for token efficiency and retrieval accuracy. #Cybernorse

We've productionized query-aware compression for faster, cleaner, more-accurate search. Better context is better than more context. Our system cuts context tokens up to 70% while improving answer quality.

@OpenAI AI reasoning now spans long chains & distant fields solving math’s planar unit distance problem. Same abilities will accelerate biology, physics, engineering, medicine. Human judgment + AI search expertise matters more. Powerful milestone for human-AI collaboration. #Cybernorse

This result points to something larger: AI systems are becoming capable of holding together long, difficult chains of reasoning, connecting ideas across distant fields, and surfacing paths researchers may not have explored. We believe those same abilities will soon accelerate work in biology, physics, engineering, and medicine. That future still depends on human judgment. Expertise becomes more valuable, not less. AI can help search, suggest, and verify. People choose the problems that matter, interpret the results, and decide what questions to pursue next.

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids. An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better. This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.

@OpenAI This general-purpose reasoning model, not a specialized math solver, solving the planar unit distance problem marks a true milestone for AI reasoning and discrete geometry. Remarkable breakthrough. #Cybernorse

The proof came from a general-purpose reasoning model, not a system built specifically to solve math problems or this problem in particular, and represents an important milestone for the math and AI communities. openai.com/index/model-di…