NerdNews

Google has launched Gemini 3 Pro, its latest large language model, which it claims is smarter, faster, and better. The model will be available in preview across several Google products, including Search, and will also be integrated into the Gemini app with a new generative interface. Gemini 3 Pro has topped several benchmarks, including LMArena and MathArena Apex, and will be available to Google AI Ultra subscribers after safety testing.

Google has introduced Gemini 3, its most advanced AI model, with improved reasoning, multimedia, and coding skills. The model will be used to enhance Google's search business and other products, and is available through the Gemini app and AI Overviews. Gemini 3 outperforms other models, including GPT-5, and has the potential to drive significant revenue growth for Google.

Stack Overflow is launching new products to position itself as a valuable part of the enterprise AI stack, aiming to translate human expertise into an AI-accessible format. The new tools, built around Stack Internal, will feed into internal AI agents using the model context protocol and include a layer of metadata to inform AI agents about the reliability of answers.

Google has unveiled Gemini 3, its latest AI model, which features improved reasoning abilities and a new generative interface. The company has also announced Antigravity, an AI-first integrated development environment (IDE) that allows developers to work with intelligent agents. Gemini 3 has shown significant improvements in various benchmarks, including a record 72.1% score in the SimpleQA Verified test.

NTT Docomo has successfully conducted an outdoor trial of a 6G AI-driven interface, achieving a throughput improvement of up to 100% compared to conventional methods. The trial, conducted in partnership with Nokia Bell Labs and SK Telecom, demonstrated the effectiveness of AI-AI technology in optimizing modulation and demodulation schemes according to radio conditions, enabling stable communication across diverse use cases.

Microsoft has announced a strategic partnership with Anthropic, bringing the AI startup's models to Microsoft Foundry. As part of the deal, Anthropic will purchase $30 billion of Azure compute capacity and commit to up to one gigawatt of compute capacity using Nvidia's systems. Nvidia and Microsoft are also investing in Anthropic, with Nvidia investing up to $10 billion and Microsoft investing $5 billion.

AI data center provider Lambda has raised $1.5 billion in a funding round led by TWG Global, following a multibillion-dollar deal with Microsoft to supply AI infrastructure using Nvidia GPUs. This investment far surpasses earlier expectations and solidifies Lambda's position in the AI data center market.

Waymo is removing safety operators from its fleet of robotaxis in Miami, ahead of a commercial launch in 2026. The company plans to offer fully autonomous trips to employees in Miami and will expand to other cities, including Dallas, Houston, San Antonio, and Orlando. This move is part of Waymo's efforts to scale its robotaxi service nationwide, with plans to offer 1 million trips per week by the end of 2026.

Anthropic, Microsoft, and Nvidia have announced a $45 billion deal to invest in AI computing resources. Anthropic will invest $30 billion in Microsoft's Azure cloud service, which will be powered by Nvidia's chips. In return, Microsoft and Nvidia will invest $10 billion and $5 billion in Anthropic, respectively. The deal will also see Anthropic and Nvidia collaborate on designing and engineering future Nvidia architectures for Anthropic workloads.

Microsoft and Nvidia announced a strategic partnership with Anthropic, a ChatGPT competitor, with a $30 billion commitment to use Microsoft's cloud services and up to $10 billion investment from Nvidia. The deal aims to reduce reliance on OpenAI and create broader demand for AI models.

Tesla has received a ride-hailing permit in Arizona, allowing it to launch a robotaxi service in the state. This is the final regulatory step for the company to begin operating its autonomous vehicle service, which has already been launched in a limited capacity in South Austin.

Zoox, an Amazon-owned company, is starting to offer robotaxi rides to the public in San Francisco. The service, called Zoox Explorer, will initially be free and available to a limited number of users who signed up for the waitlist. Zoox's custom-built vehicles lack a steering wheel or pedals and have been testing in San Francisco since 2017. The company aims to remove the waitlist altogether in 2026 and is working towards a commercial robotaxi service, competing with Waymo.

Abidur Chowdhury, a key Apple designer behind iPhone Air, has left the company to join an unnamed AI startup. His departure is part of a broader wave of resignations at Apple, including losses among AI experts to companies like Meta and OpenAI.

Microsoft introduces Cloud Rebuild and Point-in-Time Restore tools for Windows 11, aiming to reduce downtime and make recovery easier. Cloud Rebuild allows remote reinstall of Windows 11, while Point-in-Time Restore rolls back the system to a healthy snapshot.

Eviden is building France's first exascale supercomputer, Alice Recoque, with AMD chips, aiming to expand European research capabilities in simulations, data analysis, and AI. The system will feature AMD's 'Venice' Epyc CPUs and Instinct MI430X GPUs, with a second partition adding a CPU cluster. It is expected to operate at 12 megawatts for average workloads, making it more energy-efficient than comparable exascale systems.

A configuration file issue caused a Cloudflare outage, affecting multiple websites including X, ChatGPT, and Downdetector. The issue was resolved after a fix was implemented.

Researchers discovered a simple security flaw in WhatsApp that exposed 3.5 billion phone numbers, along with profile photos and text, by plugging tens of billions of phone numbers into WhatsApp's contact discovery tool. The issue was fixed by Meta after the researchers reported it, but it highlights a fundamental problem with using phone numbers as unique identifiers for services like WhatsApp.

A Cloudflare outage has taken down a huge chunk of the internet, affecting websites like Amazon, X, Spotify, and ChatGPT. The issue has been identified and a fix is being implemented, with some services slowly coming back online. Cloudflare provides infrastructure and security for many websites worldwide and is investigating the cause of the unusual spike in traffic that led to the outage.

A major US real estate firm was targeted by a highly advanced intrusion attempt using the Tuoni C2 framework, which combined social engineering, steganography, and in-memory execution. The attack demonstrates how threat actors are leveraging AI and advanced techniques to evade traditional defenses.

A new report from Entrust reveals that generative AI (GenAI) and deepfakes are being used to commit new account fraud and circumvent biometric-based checks. Digital forgeries now account for over a third of document fraud attempts, while deepfakes account for a fifth of biometric fraud attempts. Fraudsters are using AI-powered deepfakes to open new online accounts, particularly in financial services.