OpenAI says GPT-5.6 Sol quantum experiments at MIT completed 40 target measurements with researchers intervening to improve only four. The agent controlled live laboratory hardware through Codex, analyzed results, adjusted measurement parameters, and saved calibrations for later steps. That moves the model beyond advising a scientist from outside the experiment. The work does not mean an AI system independently discovered new physics. Beatriz Yankelevich, an MIT graduate stud

Google has started rolling out the Google Find Hub Remembered tab, one week after announcing it for phones running Android 16 or later. The feature lets Gemini record where users placed ordinary belongings that lack electronic tracker tags. That expands Find Hub beyond locating connected hardware, but it introduces a strict limitation. Google can only recall the location that a person previously supplied. The distinction matters because Remembered sits inside an app associate

TrendForce now expects storage prices to keep rising despite weak consumer demand, as AI infrastructure absorbs capacity once available for ordinary PCs and backup drives. The TrendForce storage price forecast turns a familiar component cycle into a broader conflict. Cloud operators want faster, larger storage systems, while individual buyers face higher costs across both SSDs and HDDs. The immediate surprise is not that solid-state drives became expensive. NAND flash has alw

Amazon SageMaker UpdateRecord introduces feature-level writes, ending a long-standing requirement to read and rewrite an entire record for every partial change. One API call can now update as many as 100 features while preserving every feature omitted from the request. That change targets a specific weakness in real-time machine learning infrastructure. Feature records often combine values produced by separate pipelines, each operating on its own schedule. A clickstream proce

Pathway has scaled its 150-million-parameter BDH-CQ model on Amazon SageMaker HyperPod, challenging the assumption that better reasoning requires longer chains of generated tokens. The Pathway BDH architecture instead performs iterative computation inside a recurrent latent state. Its latest result reached 29.5 percent pass@2 on ARC-AGI-1 while setting a new reported efficiency point. That result does not place BDH-CQ near the benchmark’s highest-performing general-purpose mo

RavenDB Quill AI agents can now work with three major SQL platforms without forcing enterprises to migrate their operational data. The product supports PostgreSQL, Microsoft SQL Server, and MySQL at launch. Yet Quill does not simply hand an AI model unrestricted access to a production database. Instead, Quill copies approved records into a synchronized RavenDB instance that runs beside the source system. AI agents query that controlled copy, while the original SQL database co

Anthropic has warned affected users that hackers are stealing Claude tokens from subscribers, even while those subscribers are not working. The warning follows reports of usage limits draining without corresponding activity. In one documented case, a user watched his usage rise by 10 percentage points during a controlled period with his Claude tasks disabled. The incident points to stolen login sessions, not a compromise of Anthropic’s models. Infostealer malware can copy aut

Cisco President and Chief Product Officer Jeetu Patel has issued a blunt warning: AI-assisted cyberattacks are approaching machine scale, despite the technology’s defensive promise. His comments followed an open letter signed by Cisco, OpenAI, Anthropic, Google, Microsoft, and more than 100 other organizations. The letter argues that companies and public institutions have only a limited window to strengthen their defenses. In a September 8 Bloomberg interview, Patel explained

OpenAI’s GPT-6 Astra reached general availability on Amazon Bedrock, moving the model into an enterprise platform designed for governed, large-scale inference. The GPT-6 Astra Amazon Bedrock launch matters because access no longer depends on adopting a separate AI operating environment. AWS says Astra brings deeper reasoning and sharper judgment to demanding work. Those claims still require independent testing across actual business workloads. The immediate change is simpler

AWS has published SageMaker AI inference benchmarks comparing four GPU instance families across two 30-billion-parameter models, with G7 taking the price-performance lead. The results place NVIDIA Blackwell hardware against the older G5, G6, and G6e options under a shared real-time inference workflow. That comparison matters because buying the newest GPU is not automatically the best deployment decision. Model architecture, request concurrency, response length, memory capacit

Qualcomm secured Amazon as a major data center customer, but the Qualcomm Amazon AI chip deal is not a simple supplier contract. It links a new custom silicon program to purchase activity that can reach $60 billion over ten years. The companies will develop multiple generations of customized chips for Amazon Web Services. Their work covers AI inference, which runs trained models to produce answers, and high-speed optical connections inside data centers. Amazon also received a

Mistral AI raised €3 billion on September 8, giving the French lab more than €21 billion in post-money value and a costly new mandate. The Mistral Series D funding must now support frontier research, European computing infrastructure, enterprise products, and international expansion at the same time. Samsung Electronics led the round. The Scaleup Europe Fund, managed by EQT, and existing investor PSG Equity joined as co-leads. Mistral calls it the largest equity fundraising r

Shodh AI has introduced LUCAN with a bold conflict at its center: one model should reason across molecules, reactors, and manufacturing equipment. The Shodh AI LUCAN release targets a gap that specialized scientific models still leave open. Those systems can predict isolated physical domains, yet factories combine chemistry, fluid flow, heat transfer, and mechanical stress. The company says LUCAN connects those domains inside a shared computational model. Engineers can specif

Acer SFF RTX Spark puts up to one petaflop of AI compute and 128 GB of unified memory inside a compact desktop concept. That combination challenges the assumption that demanding personal AI must run in a remote data center. Acer unveiled the design in Berlin on September 2 during IFA 2026. The company presented it as a machine for creators, developers, gamers, and people building locally operated AI agents. The specifications are striking, but the central conflict is not Acer

Google began its largest service-quality reduction in Search’s 29-year history, according to the company, after European regulators demanded substantial changes. The Google DMA search changes affect results for hotels, flights, restaurants, shopping, transportation, and sports across the European Union. They reduce Google’s rich answers while giving specialized comparison services more prominent positions. Google says this arrangement makes common searches slower and less use

Arm expanded its AI strategy across three computing markets on September 8, linking cloud servers, edge devices, and physical machines through one architecture. The Arm edge AI push includes new data center designs, a mobile compute platform, developer software, and a robotics partnership program. The announcement matters because Arm is no longer presenting separate stories for servers, smartphones, and embedded systems. It wants developers to treat them as parts of one compu

SGC Energy has selected Vertiv for South Korea’s first planned PowerNexus deployment, despite its Gunsan AI data center still facing larger execution tests. The SGC Energy Vertiv PowerNexus agreement covers joint technical reviews of power and cooling infrastructure. It also puts a factory-integrated electrical system at the center of the project’s construction strategy. The agreement matters because SGC is trying to move beyond energy generation and become an AI infrastructu

Lancaster University has secured about €475,000 for a three-year quantum neural computing project built around an unusual wager: uncertainty should become a resource. The wider IgnisQNC consortium carries a reported value of €1.88 million. Its researchers plan to develop quantum-enabled learning systems that can explain how they reach decisions. Those methods will eventually face tests on emerging quantum hardware. That ambition places the Lancaster University quantum neural

TIER IV released METEOR on September 8 with a striking claim: humans supplied no additional labels and wrote none of its model or training code. The TIER IV METEOR release includes source code, trained weights, and a recipe for reproducing its development pipeline. Yet the release does not include public benchmark results that would establish how well the system drives. That gap defines the real story. METEOR is an open, camera-only autonomous driving model designed to move f

Alibaba Cloud opened two Brazilian data centers, giving its Alibaba Cloud Brazil region a local foothold against AWS, Microsoft Azure, and Google Cloud. It is the company’s first cloud region in South America and its second in Latin America. The launch changes Alibaba’s position in Brazil from an overseas provider into a local infrastructure operator. Brazilian organizations can now place computing, storage, databases, and other workloads inside the country. Alibaba also plan