News 4 u.

A Linux kernel use-after-free in IPv4 TCP handshakes has received a critical CVSS 9.8 rating, and affected systems may be reachable remotely before application authentication. Fixed upstream kernels exist, but Ubuntu and RHEL still listed affected releases as vulnerable on October 31. The report attributes a working root-command exploit on systems with multipath TCP listeners to Anthropic, while noting that the exploit has not been published or independently tested. Administrators should track distribution fixes, reboot after patching and treat mitigations as temporary.

Nvidia is discussing a guarantee for debt financing tied to Nscale’s planned Anthropic data center in West Virginia, potentially taking temporary control of the facility if Nscale defaults. The arrangement would help fund a 460-megawatt building while protecting Nvidia’s chip demand, but it would deepen the company’s exposure to highly leveraged AI infrastructure. Nscale has deployed far fewer chips than planned, faced construction problems, and is seeking a valuation of up to $35 billion.

US technology shares fell sharply for a second day as investors questioned whether escalating AI infrastructure costs will produce sufficient returns. Intel lost 28%, Reddit fell 29% and Roblox dropped 18%, while Alphabet continued declining after saying AI search summaries were affecting advertising revenue. Five major firms are committing about $800 billion to 2026 spending, and Amazon plans $220 billion. Analysts cited weaker earnings signals, cash outflows and rising skepticism, although the Nasdaq remained above its level a year earlier.

OpenAI has reportedly reduced the target for its latest funding round from as much as $30 billion to $20 billion, with an initial close expected to be smaller. The Financial Times said investors from the United Arab Emirates would lead the round, while The Wall Street Journal reported that $11 billion had been secured and some large investors responded tepidly. The reports come as markets question the returns from enormous AI infrastructure spending. TechCrunch said it had reached out to OpenAI for comment.

Qatar-mediated contacts have opened a possible diplomatic channel between Iran and the United States, but both governments continue signaling readiness for escalation. Iran says it could shorten the timetable for partially reopening the Strait of Hormuz to two days if Washington meets demands including sanctions relief and an end to the naval blockade. The United States is pressing for nuclear concessions while maintaining economic pressure. Military preparations, attacks on shipping and threats to infrastructure show that the opening remains fragile.

A federal judge ordered the Trump administration to stop threatening state and local election officials with criminal charges over ballot processing and to halt demands that officials set aside ballots allegedly linked to non-citizens. The ruling followed federal letters, armed visits and subpoenas in California, and similar demands in Washington. Judge Trina Thompson said the actions were likely to interfere with election administration and disenfranchise lawful voters. The administration disputes the ruling and continues to frame the measures as election-integrity enforcement.

The packet combines a walkthrough of Reddit Proto with a review of Anthropic’s Claude Duet. Reddit’s internal system uses real data, production components, code-aware annotations and structured handoffs to let non-engineers build realistic prototypes; the account says it has reached broad adoption. Duet is praised for design, connectors and agent organization, but criticized for latency, weak follow-through, limited notifications, and its lack of desktop and team support. The conclusions are based on firsthand testing and company-reported Reddit results.