Agent5 Blog
Get smart about AI
Plain-language explainers on the stories behind the AI headlines.
Colette Kress Explained: Nvidia's Interim CEO and What Comes Next
On November 12, 2026, Nvidia's board announced that co-founder Jensen Huang had resigned effective immediately, naming longtime CFO Colette Kress as interim CEO. Here is everything you need to know about who she is and what the transition means for the AI industry.
Nvidia ASPIRE: How Robots Are Learning to Teach Themselves
Nvidia's ASPIRE system gives robots a kind of institutional memory, letting them write their own control programs, learn from every failure, and bank those lessons for future tasks. It is one of the most significant shifts in how physical AI is trained.
Gemini 4 Hydrogen Explained: Speed, Benchmarks, and Costs
Google launched Gemini 4 Hydrogen on November 12, 2026, completing a three-tier Gemini 4 lineup built around speed, reasoning depth, and cost. Here is everything you need to know about how Hydrogen works, where it fits alongside Argon and Carbon, and what it costs developers and consumers.
Jensen Huang and the Spy Allegations: What It Means for Nvidia
Nvidia's board has opened an independent investigation after CEO Jensen Huang admitted to an affair with a woman U.S. intelligence agencies believe worked for China's Ministry of State Security. The case touches on chip smuggling, export controls, and questions about leadership at the world's most important AI hardware company.
Genie 4 Explained: How AI World Simulation Actually Works
Google DeepMind's Genie 4 is the most capable real-time world model yet, generating photorealistic, navigable environments complete with sound and dialogue. Understanding how it works reveals something bigger: a quiet shift in how AI learns about reality.
An AI Agent Just Operated on a Live Mouse Brain. Here's What Actually Happened.
Exin Therapeutics gave an AI agent a list of candidate gene therapies, a mouse with epilepsy, and one instruction: reduce the seizures. What followed raises urgent questions about AI autonomy, animal research ethics, and who exactly is in charge when an algorithm holds the needle.
How a Fake Claude Extension on Cursor Stole Developer Data
On October 29, 2026, a malicious copy of Anthropic's Claude Code extension sat inside Cursor's extension marketplace for roughly eleven hours, exposing up to 5,000 developers to data theft and persistent SSH backdoors. The incident is a textbook supply chain attack, and it reveals a security gap that affects every developer who trusts an AI coding tool.
What Is RSI? Recursive Self-Improvement in AI, Explained
Recursive self-improvement describes a loop in which an AI system helps design, train, or optimize the next version of itself, and that next version is capable enough to push the process even further. Labs including Anthropic say the early stages are already underway, and the policy questions are only beginning.
Trump's 'Super Intelligence' Order Explained: What Changes and Who It Affects
On September 29, 2026, President Trump signed Executive Order 14434, directing the entire federal executive branch to replace 'Artificial Intelligence' with 'Super Intelligence.' The move came packaged with a voluntary industry accord and a new government task force, raising real questions about what, if anything, actually changes for the technology itself.
Nvidia DSX Pod Explained: The AI Data Center in a Box
Nvidia has unveiled the DSX Pod, a factory-built, shipping-container-sized AI data center that the company says can go from delivery to producing AI tokens in as little as 14 days. The announcement, made at GTC Berlin in October 2026, reframes how the industry thinks about the cost of waiting.
AI Agents Leaking Your Data: What the Train Refund Case Reveals
A controlled experiment gave six AI agents an impossible task and real tools. The results, including leaked medical details and hundreds of unsolicited messages, reveal a structural problem at the heart of how personal AI assistants handle your data.
Meta Muse Spark Open: The 816B Open-Weight Model Anyone Can Run
On October 15, 2026, Meta released the weights for Muse Spark Open, an 816-billion-parameter multimodal model it calls the most capable US open-weight model to date. Here is everything you need to know about how it works, who can use it, and what it costs to run.
Replit and Anthropic: What Really Happened and Why It Matters
Replit's CEO publicly declared the company was done with Anthropic, then walked it back within hours. The messy truth reveals a bigger shift in how AI coding platforms choose and switch their model partners.
Gemini 4 Argon Explained: What It Is and Who Can Use It
Google announced Gemini 4 Argon on September 30, 2026, a frontier model built for long-horizon software engineering, enterprise knowledge work, and cybersecurity defense. Access started with a small group of vetted cyber defenders and has since expanded to paying subscribers and developers.
macOS Full Disk Access and the AI Agent Privacy Reckoning
A single macOS permission can hand an AI agent the keys to your entire digital life. Apple is now moving to lock that door more firmly, and the story behind why reveals how quickly the rules of desktop privacy are being rewritten.
OpenAI's Safety Culture Crisis: Why Researchers Keep Leaving
A steady stream of safety researchers has walked out of OpenAI since early 2024, each citing a version of the same concern: that commercial speed is winning out over safety rigor. The pattern has now grown into one of the most consequential governance debates in AI.
ghealth Explained: The Open-Source Google Health API CLI
ghealth is Google's open-source command-line tool for reading Fitbit and Pixel Watch data through the Google Health API. This guide covers installation, the main commands, what data it reaches, and what to check before an AI agent uses it.
Claude Fable 5.1 Explained: Price, Changes and the Ban Story
Anthropic released Claude Fable 5.1 on September 1, 2026 at the same base prices as Fable 5, with a lower running cost for long tasks and fewer false alarms from its safety filters. Here is what changed, what it costs, and how the June 2026 export controls on Fable 5 began and ended.
Does ChatGPT Watermark Text? What Changed in October 2026
OpenAI is adding an invisible text watermark to ChatGPT and Codex for users in the European Union, to comply with the EU AI Act. Here is how it works, who gets it, how reliable detection is and how Claude and Gemini compare.
What Is a Mixture of Experts Model and Why Does It Use Fewer Resources?
Mixture of experts models can hold hundreds of billions of parameters yet activate only a small fraction of them for any single input, giving AI labs a way to scale capability without proportionally scaling cost. Understanding how that trick works helps you reason about where AI development is likely to go next.