NerdNews

Nvidia reported a record $51.2 billion in data center revenue, with its Blackwell Ultra chip driving the bulk of its latest data center surge. The company's gaming revenue is also up 30% year-over-year. Nvidia CEO Jensen Huang expressed confidence in the company's AI growth, despite concerns about an 'AI bubble'.

Yann LeCun, a pioneer in AI, is leaving Meta to create an independent entity focused on developing AI systems that understand the physical world. LeCun's new company will have a partnership with Meta, allowing the social media giant to access its innovations. This move is seen as a significant development in the AI field, with potential applications in various sectors of the economy.

Microsoft introduced Copilot Actions, an experimental AI feature in Windows, which can infect devices and steal user data due to known defects in large language models, including hallucinations and prompt injections. Critics question the value of warnings and the ability of users to prevent exploitation attacks.

Function Health, a health tech company, has raised $298 million in a Series B round at a $2.5 billion valuation. The company offers a regular lab testing service and is developing a 'medical intelligence' generative AI model to provide personalized health insights. The funding round was led by Redpoint Ventures and saw participation from several other investors.

Microsoft's Magentic Marketplace is an open-source simulation environment that studies how LLM-based agents behave in multi-agent economic systems. The platform recreates the full transaction lifecycle and provides researchers with a controlled setting to examine agent behavior. Early findings reveal a Paradox of Choice effect in agent behavior and significant variation in manipulation resistance across different LLM models.

Engineers Kieran Klaassen and Nityesh Agarwal discuss how they use AI-assisted coding tools like Claude Code to transform their workflow, allowing them to ship six features, five bug fixes, and three infrastructure updates in one week. They built a prompt that writes prompts, creating a custom command in Claude Code that transforms a rough feature idea into a fleshed-out GitHub issue.

A report by MIT Sloan Management Review and Boston Consulting Group warns that existing management principles are incompatible with agentic AI, which can act as both a tool and a worker. IT leaders and executives need to reassess their approach to managing agentic AI systems, which require substantial initial development costs and ongoing variable costs, and can depreciate through model drift while appreciating through fine-tuning and emergent capabilities.

Some software developers are being required to use AI tools, which they feel is detrimental to code quality and their own skills. Companies like Microsoft and Meta are pushing for increased AI usage, with some even making it a part of employee performance evaluations. However, many developers are resistant to this change, citing concerns about ethics, bias, and errors.

The article discusses the hidden risks in DevOps stack data, including weak access control, improper repository permissions, and lack of automated backups. It highlights the importance of implementing strict access controls, protecting credentials, and leveraging automated backups to secure data against ransomware attacks and human errors. The article also provides guidance on how to improve the security of DevOps data, including access management, backup and disaster recovery, and compliance with industry regulations.

The article discusses the challenges of building event-driven systems across multiple cloud providers, including latency, resilience, event ordering, and duplicate handling. It provides practical code examples and proven solutions to address these challenges, using a realistic financial services case study.

Microsoft has announced Azure HorizonDB, a distributed PostgreSQL database service designed to rival other hyperscaler systems. The service claims 100% compatibility with open source PostgreSQL and offers improved performance, scalability, and availability. It features autoscaling storage up to 128 TB, scale-out compute up to 3,072 vCores, and enterprise security and compliance.

Warner Music Group has settled a copyright infringement case with AI music startup Udio and entered into a licensing deal for an AI music creation service. The platform, set to launch in 2026, will allow users to create remixes and new songs using licensed music, ensuring artists and songwriters are credited and compensated.

Poly, a YC-backed startup, has relaunched as a cloud-hosted file storage service with AI-powered search. The service offers 100GB of free storage and allows users to upload files, tag them, and ask AI-powered questions about them. Poly aims to solve the problem of organizing and searching through large amounts of files and data.

Onepot AI, a company that aims to make chemical drug creation easier, has raised $13 million in funding. The company uses AI to help synthesize small molecules, which are the building blocks of drugs. This process can be time-consuming and expensive, but Onepot AI's technology aims to speed it up and reduce costs. The company's founders, Daniil Boiko and Andrei Tyrin, came together to create Onepot after realizing that the best ideas in drug discovery were often blocked by synthesis issues.

Target is launching a new ChatGPT-powered app for shoppers, allowing them to browse, build multi-item baskets, and check out. The app is part of OpenAI's growing list of retail apps and deepens the company's existing enterprise partnership with Target. The partnership will also roll out ChatGPT Enterprise across Target's 18,000 employees for tasks like supply chain forecasting and streamlining store processes.

Epic Games and Unity announced a surprise partnership to promote open and interoperable game development. Unity developers can bring games to Fortnite, and Unity's commerce platform will come to Unreal Engine, creating more opportunities for developers and players.

Sunday Robotics has developed a home robot called Memo that can perform tasks like making coffee, clearing tables, and loading the dishwasher. The robot uses a novel training method that involves remote workers wearing gloves to collect data. Memo is set to be released to beta testers next year and could potentially revolutionize household chores.

The European Union is proposing changes to its General Data Protection Regulation (GDPR) and AI Act, relaxing rules on data sharing and AI model training. The changes aim to cut red tape and boost economic growth, but have sparked outrage among civil rights groups.

Amazon's Chief Security Officer Steve Schmidt warns that hostile countries are using cyber operations to gather intelligence for physical military strikes, and companies are getting caught in the crossfire. This new operational model requires companies to take a holistic approach to security and risk management, considering both physical and digital security.

Google Maps will use AI to predict EV charger availability on arrival, helping users avoid lines and save time. The updated tool uses a combination of AI and real-time charger availability to predict open chargers. Additionally, the Explore tab has been revamped to make it easier to find curated lists from local influencers.