🚨 WTF Ubuntu?
The central concern is the potential gap between enterprise security promises and real-world patching performance. Given the severity of these allegations, administrators should independently verify official advisories, affected configurations, and available patches before drawing conclusions.
A developer sharply criticizes Canonical for allegedly leaving CVE-2026-98432, a critical Linux kernel TCP vulnerability rated CVSS 9.8, unpatched despite Debian releasing a fix on October 16. The author claims the flaw enables remote root compromise, that Anthropic has demonstrated a working exploit, and that all supported Ubuntu releases remained vulnerable as of November 2. Frustrated by Canonical’s response and its security assurances, the developer has reportedly taken their Ubuntu servers offline and plans to migrate to other Linux distributions.
🔗 Read more 🔗
🧠 Astra can read and write weights to improve Qwen
If these results scale beyond small experimental networks, direct weight-space programming could fundamentally change how AI systems design and optimize models. Independent replication, scalability tests, and rigorous comparisons with established optimization techniques will be essential.
A research post suggests that frontier language models can interpret and generate neural network weights directly. According to the authors, GPT-6 Astra can infer a network’s learned task from its raw parameters and generate small game-playing networks without environmental feedback. They report that these networks outplay networks trained with proximal policy optimization (PPO) for 100 million steps.
🔗 Read more 🔗
🧊 Supermarket freezers across Europe hit by second suspected cyber-attack
These incidents demonstrate how attacks on operational technology can rapidly translate into physical disruption and financial losses. Refrigeration infrastructure is especially vulnerable: even short outages can destroy inventory, while manipulated temperature controls introduce additional safety risks.
Supermarkets in at least seven European countries reported refrigeration failures on November 2 amid suspected cyberattacks, with some cooling systems allegedly manipulated to reach temperatures of 50°C. Incidents in Belgium, Estonia, Ireland, Finland, Luxembourg, the Netherlands, and Sweden led to spoiled food, empty refrigerated shelves, and emergency repairs. The disruptions follow a similar incident in France that reportedly affected up to 700 supermarkets and caused tens of millions of euros in losses. European authorities have raised concerns about the security of remotely managed refrigeration systems and the risk of coordinated attacks on food supply infrastructure.
🔗 Read more 🔗
💻 Introducing Kwi, a new KDE terminal emulator
Kwi’s focus on native KDE integration could make it attractive to Plasma users. Its long-term appeal will depend on how well it competes with Konsole in performance, functionality, and ongoing maintenance.
A KDE developer has introduced Kwi, a new terminal emulator under development for KDE Plasma and other open-source desktop environments. Designed around a modern, performant technology stack, Kwi aims to follow KDE’s interface guidelines and established design conventions. Its author calls the vision deliberately unexciting, just what users expect from a Plasma terminal: it builds on libghostty-vt from the Ghostty project and KDE Frameworks such as KPty, renders with a GPU-accelerated (Vulkan or OpenGL) Qt Scene Graph engine, and for now offers a more streamlined feature set than Konsole.
🔗 Read more 🔗
🐘 Postgres 19 is late, and it’s AI’s fault
AI-powered bug discovery presents an unexpected paradox: finding defects faster can delay releases rather than accelerate them. The emerging bottleneck is not detection but reviewing patches, validating fixes, and maintaining sufficient experienced contributors to process the growing workload.
PostgreSQL developer Robert Haas explains why PostgreSQL 19 is likely to miss its usual September–October release window: AI-assisted bug discovery has substantially increased the project’s maintenance burden. Comparing equivalent periods in 2025 and 2026, commits to the branch of the release being stabilized climbed from 724 (REL_18_STABLE) to 1,169 (REL_19_STABLE), while fixes backported to the previous stable release rose from 248 to 433. The growing review and remediation workload contributed to the late removal of major features, including SQL Property Graph Queries and partition-management enhancements. Haas considers the modest delay a necessary trade-off for software quality but questions the sustainability of a development process in which AI finds defects faster than maintainers can fix them.
🔗 Read more 🔗
🧩 Pratt parsing is a black box
For compiler engineers and language-tooling developers, understanding operator precedence and parsing semantics is far more useful than copying a minimal implementation. A solid conceptual foundation also makes parsers easier to extend with new operators and syntax.
A technical article argues that most Pratt parsing tutorials emphasize the algorithm’s remarkably concise implementation while overlooking the principles that make it effective. The author notes that the core parser can be written in roughly nine lines in a high-level language such as Rhombus. The central message is that understanding the underlying parsing model is more valuable than memorizing a compact implementation.
🔗 Read more 🔗