GoKawiil
Tech News
Apple has named John Ternus as its new chief executive, ending Tim Cook's 15-year tenure atop the company as Cook transitions to executive chairman. Ternus, 51, spent 25 years at Apple, rising from a product design engineer to senior vice president of hardware engineering, where he oversaw AirPods, Apple Watch, Vision Pro, Apple silicon, and the MacBook Neo.
A new study from Google Research and Technion shows that when models like GPT-5 and Gemini-3 hallucinate, the underlying fact is usually already stored in their parameters rather than missing entirely. The researchers found these models encode 95-98% of tested facts, but often fail to surface them during generation due to retrieval, not storage, limits. They propose 'fact-level profiling' to separate whether a fact is encoded versus reliably retrievable across different phrasings.
Researchers analyzing last November's severe geomagnetic storm found it caused GPS positioning errors exceeding 33 feet across the continental US. The disturbance pushed the Aurora Oval far south of its usual position, creating dense atmospheric irregularities that disrupted the timing of satellite signals passing through the ionosphere.
NASA's Aerospace Safety Advisory Panel reported that after reviewing data from the Artemis II mission, the Orion capsule's heat shield showed only about nine sites of char loss, a dramatic improvement from the more than 100 sites, some over an inch deep, seen after Artemis I. NASA had altered Orion's reentry trajectory rather than redesign the heat shield following the 2022 issues, and the approach appears to have worked as the crew safely splashed down in the Pacific.
Shadowserver has identified nearly 22,000 internet-exposed Microsoft Exchange servers that remain unpatched against CVE-2026-62911, a high-severity authentication bypass flaw affecting Exchange Server 2016, 2019, and Subscription Edition. Most vulnerable systems are located in the United States and Germany, where officials say roughly 85% of on-premises Exchange servers remain exposed despite Microsoft releasing a fix in August 2026.
India's government has introduced the Prime Minister Research Chair Scheme, offering 120 overseas Indian-origin researchers up to 50 million rupees (about $520,000) in grants plus stipends and allowances over five years to work at Indian institutions. The initiative targets strategic fields like AI, quantum computing, biotechnology and advanced materials, joining similar brain-drain reversal efforts by China, France, Germany and Denmark.
Researchers on the LUX-ZEPLIN experiment at South Dakota's Sanford Underground Research Facility reported a lone high-energy event in their 10-tonne liquid xenon detector, recorded sometime in 2023 or 2024, that matches the expected signature of a dark matter particle striking an atomic nucleus. The team announced the finding at a physics conference in Japan and posted details in a preprint, while cautioning that a single event is not proof.
Ross Cohen, who runs SBA lender B:Side Capital and AI firm Main & Machine, describes how his staff's first reaction to an AI adoption announcement was fear of layoffs. His fix was to define upfront which tasks—like credit decisions, hardship conversations, and commitments to borrowers—AI would never handle, before choosing any software.
A new study using coral data going back a millennium finds that human-caused warming appears to be intensifying El Niño events. Researchers say the historical coral record shows a pattern connecting rising global temperatures to more pronounced El Niño swings than in past centuries.
A security researcher describes building a new memory architecture for LLM agents used in vulnerability research, after noticing that standard retrieval-based memory let models keep reasoning from facts that had since been disproven. Instead of just storing and retrieving past observations, the system tracks logical dependencies between facts, so that when one assumption is invalidated, everything derived from it can be automatically flagged or revised. The approach effectively borrows ideas from program analysis, treating an investigation's knowledge state like a dependency graph rather than a flat log of conversation history.
An Anthropic fellows program researcher, Chen Yueh-Han, published a paper showing an automated system that searches literature, proposes fixes, and trains models to improve performance across 10 alignment benchmarks without hurting overall capability. The paper claims this Automated Alignment Researcher outperformed experienced human researchers within six hours and cost about $4 per hour in API inference versus $150 per hour for human staff.
Anthropic has published a new framework called Model Hardware Standard that sets rules for how AI agents should interact with physical devices such as microscopes, liquid-handling systems, quantum computing hardware, manufacturing machines, and robot arms. The company plans to test it with trusted partners to refine safety measures before releasing it more broadly.