Technology

Anthropic has just surfaced what appears to be the first large-scale, mostly autonomous, AI-orchestrated cyber espionage campaign. In mid-September 2025, they detected suspicious behaviour that turned out to be a sophisticated operation run by a Chinese state-sponsored group. AI wasn’t just “helping.” It was doing most of the work. The attackers jailbroke Claude Code, framed it as doing defensive testing, and then drove it through an automated framework aimed at ~30 global targets, big tech, finance, chemicals, and government. A few intrusions succeeded. ...

Some sections of this post were written with the assistance of AI to improve clarity and readability. The historical context, reasoning, and overall flow are entirely my own. I’ve been noticing something lately when I read about new infrastructure projects. Not the specific technologies, those change constantly, but the shape of the problems they’re solving. It’s like watching reruns of a show you half-remember: the set design is different, the actors have changed, but you know exactly how this episode goes. ...

Some sections of this post were written with the assistance of AI to clarify ideas and improve readability. All opinions and conclusions are my own. Zero-Trust Is Already Happening Your VPN goes down at 2 AM. Half your engineering team is locked out. The on-call engineer can’t access the database to diagnose the outage that triggered the VPN failure in the first place. You’re now troubleshooting infrastructure access during an infrastructure incident, a recursion problem that would be funny if it weren’t costing you money and sleep. ...

Investors poured over $9.5 billion into AI processor startups in 2024, betting on architectures that could reshape inference economics. NVIDIA itself projects the broader AI-infrastructure market could reach $3,4 trillion by 2030. That kind of capital rarely gathers around incremental improvements, it usually signals an architectural inflection point. Yet GPUs still dominate both training and most inference workloads today, so any transition will be evolutionary before it is disruptive. ...

fascinating once again, the universe reminds us that all abstractions eventually resolve to DNS Monday’s Massive AWS Outage Explained: Looks Like It’s Finally Over - CNET Amazon brain drain finally caught up with AWS • The Register

Agentic RAG goes beyond traditional RAG by adding the ability to reason, plan, and act. Instead of merely fetching and producing information, it autonomously determines what to retrieve, how to apply it, and when to adjust the context, enabling adaptive, goal-oriented systems that improve through continuous learning.

The software industry’s latest fantasy- that AI will deliver a tenfold leap in developer productivity- is now colliding with reality. As the author of the piece notes, the trillion-dollar AI sector still cannot produce reliable, production-grade code. Behind the flood of “breakthrough” announcements lies an epidemic of unmaintainable software, riddled with four recurring categories of error. The classical ones- false positives and false negatives- are joined by two new cognitive pathologies born from model mechanics themselves: entangled logic that turns architectures into spaghetti, and memory collisions that make unrelated functions misfire. What was once branded “hallucination” now looks less like quirk and more like structural failure. ...

Over the last year the workflows vs. agents debate has turned from a niche engineering question into something every product team seems to argue about. New tooling lowers the barrier to spin up an AI agent that can call tools in a loop, so leaders start asking whether they should rebuild processes around these agents, or stick with the safer structure of step-by-step workflows. The conversation heats up because both sides have real wins and real failure modes: agents feel magical when they solve fuzzy, open-ended tasks, but they can be slow, costly, and unpredictable at scale; workflows are efficient and auditable, but can feel rigid when the job needs exploration. That tension between flexibility and control is why this topic is hot and why teams keep getting stuck. ...

Bloomberg just published this fascinating map of the AI power network, showing how companies like NVIDIA, OpenAI, Microsoft, AMD, Oracle, and Intel are now intertwined through billions in deals, compute, and equity. It’s not just a supply chain anymore. It’s a feedback loop — where hardware, software, and capital keep feeding each other. NVIDIA sits at the center with a $4.5 trillion market cap, investing up to $100 billion in OpenAI, while selling GPUs to Oracle, AMD, xAI, and everyone else. OpenAI, in turn, signs a $300 billion cloud deal with Oracle, deploys 6 gigawatts of AMD GPUs, and gives AMD an option to buy 160 million OpenAI shares. Microsoft is still the connective tissue, part investor, part service provider, part enabler.

Time Between Disengagements is a concept I came across in a recent article from Gitpod, and it offered an interesting new way to think about AI’s role in software development. It compares the evolution of AI in engineering to the progression of self-driving cars—where the key metric is how long an autonomous system can operate before a human needs to step in. That simple but powerful analogy really clicked with me. It reframes how we should think about the future of AI-assisted development—not just in terms of raw capability, but in how independently and safely these systems can work. ...