Claude Fable 5.2 brings lower prices to Anthropic’s most demanding technical work
Anthropic’s latest release combines faster output with a lower advertised price. For developers, the more consequential choices concern which work merits the larger model, which access conditions apply and how an upgrade changes an existing application. The release gives teams a reason to compare complete results, including checking and correction, against their existing workflow.
Oracle puts a cash-flow timetable on its AI infrastructure expansion
Oracle has outlined when its infrastructure expansion should generate positive free cash flow and set a limit on additional capital-market funding. The details matter: refinancing sits outside the limit, and an earlier equity raise remains part of the funding history. The announcement creates a more precise test for the company’s next results.
Microsoft reports 48% Azure growth as AI demand faces the returns test
Microsoft’s first-quarter results provide fresh evidence of growing cloud demand. Understanding what they establish requires separating Azure growth from the broader Cloud total, future obligations from quarterly revenue, and investment effects from operating performance. The results also leave a distinct question for customers: what value their own AI spending produces.
GLM-5.3 completes simulated cyberattacks as researchers find instruction failures
Irregular’s latest evaluation places GLM-5.3 ahead of the open-weight models it has previously tested on multi-stage cyber scenarios. A separate finding about instruction violations raises a different question: whether completing a task is enough to count as success. The evidence calls for a closer distinction between capability, reliability and authorised behaviour.
SK hynix’s Record Results Put AI’s Memory Shortage in Focus
SK hynix’s latest results show how much value is accumulating around the memory that feeds AI processors. Its outlook raises a harder question for the industry: who can secure capacity, and on what terms?
Apple Opens Its Documents to AI Agents. The Important Part Is What They Can Change.
Apple is bringing assistants into its productivity apps and opening a route for external agents to edit them. The useful question is how much work people can delegate while remaining able to inspect the result.
Google’s Genie 4 Makes AI Worlds More Convincing. Can They Be Trusted?
Google’s new world model brings sound and live direction to generated environments. Its next challenge is less cinematic: establishing when a convincing simulation can support a reliable decision.
Top AI Startups in Europe (2026): 20 Companies to Watch
Emad Mostaque Says We Won’t Get the Best AI. What Does the Evidence Show?
Emad Mostaque argues that the most capable AI will stay inside powerful institutions. The evidence reveals a more complicated divide over models, computing and control.
50 People Shaping AI in 2026: The Nuvastra Guide
Who is doing the work that will define AI’s next phase? Nuvastra profiles 50 researchers, founders and thinkers across frontier models, robotics, scientific discovery, infrastructure and accountability. This unranked guide explains their contributions, connects the ideas behind them and distinguishes established results from emerging ambitions, with an individual pen-and-ink-style portrait for every person.
Bryan Korba on AI: “Being right more often than a human is not the qualifying condition for authority”
Bryan Korba brings an investor’s and operator’s perspective to the debate over artificial intelligence. In this Nuvastra interview, the Popcorn Labs founder discusses machine authority, institutional failure and the future of work.
Claude Enters Financial Advice. The Real Prize Is the Workflow.
Anthropic is bringing Claude into the work around financial advice, with Charles Schwab providing a route to independent advisory firms. The launch connects AI to existing tools for preparation, research and follow-up. Nuvastra examines why the integration matters, what the pricing claims do—and do not—establish, and how firms could distinguish a fluent demonstration from a genuinely useful working process.
Waymo Targets Tokyo Robotaxis in 2027. The Hard Part Is Local.
Waymo is preparing a driverless taxi service in Tokyo with GO and Nihon Kotsu, targeting 2027 subject to approvals and validation. The plan puts a practical test at the centre of physical AI: whether a driving system can travel successfully when the service around it must be built locally. Nuvastra examines the evidence, the remaining conditions and what a credible rollout would need to demonstrate.
Microsoft’s AI Code of Conduct Puts Human Control Ahead of Capability
Microsoft has set out a proposed rulebook for AI that remains under human control. Its ambitions raise practical questions about authority, interruption and the difference between a stated principle and a tested safeguard. Nuvastra examines what the consultation establishes, what it leaves unproven and how businesses could judge the promise against completed work.
AI Leaders Call for a Slowdown. Who Will Check They Mean It?
Dario Amodei’s call to pace frontier AI has won support from Sam Altman and Elon Musk and unsettled global markets. Behind the headlines lies a practical question: who gets to inspect the work? Nuvastra examines the proposal, the lessons from independent investigations and the evidence needed to turn a safety commitment into meaningful oversight.
d-Matrix Joins NVIDIA’s Platform. The AI Chip Race Is Becoming a Systems Race
d-Matrix plans to bring its Raptor inference processors into NVIDIA’s rack-scale infrastructure. The partnership offers a revealing view of an AI chip market in which specialist hardware can compete within a larger platform. Nuvastra examines the 2027 roadmap, the logic of splitting inference workloads, and the evidence customers will need before promises of faster responses become a convincing business case.
Anthropic Reveals How Claude Was Used in Cybercrime, Surveillance and Fraud
Anthropic’s latest disclosure raises a difficult question for the AI industry: how should productivity be measured when the same capabilities can support legitimate work and organised abuse? Nuvastra examines the evidence behind the Claude misuse report, places it alongside security research from Microsoft and the UK’s NCSC, and explores why permissions, oversight and accountability belong at the centre of the business case for AI agents.
Mistral Raises Record €3bn as Europe Tries to Turn AI Sovereignty Into Scale
Mistral has raised €3bn at a valuation above €21bn, completing the largest private technology equity round in European history. Nuvastra examines why Samsung's investment, Mistral's sovereign-compute ambitions and Europe's enormous funding gap with the US matter more than the headline valuation, and whether Europe's leading AI company is developing a genuinely different way to compete.
AI Glossary 2026: The New and Emerging AI Terms You Need to Know
Artificial intelligence has developed a new vocabulary as the industry moves beyond chatbots towards agents, reasoning systems, persistent memory and physical AI. Nuvastra explains the emerging terms that matter in 2026, from MCP, A2A and context engineering to recurrent depth, world models, open-weight AI, vibe coding and AI slop, including what each term means, why it matters and what it should not be confused with.
XDOF: Inside the Robotics Data Startup Building the Infrastructure Behind Physical AI
XDOF has gone from a UC Berkeley robotics project to one of physical AI’s most closely watched infrastructure startups. Nuvastra examines its founders, $70 million Series A, ABC-130K dataset, robot-data business, competitors and the reported financing that could value the company at around $1.2 billion.
