NerdNews

Salesforce's approach to self-healing systems utilizes AIOps and Agentic AI to analyze Kubernetes cluster health, diagnose problems, and resolve issues with minimal human intervention. Their solution consists of AI Agents that perform actions against their K8s environment and tools like Amazon Prometheus and ArgoCD Controller. The goal is to let application teams focus on business requirements, not infrastructure overhead.

Anthropic has announced a $50 billion data center partnership with Fluidstack to build facilities in the US, aiming to support its growing compute needs for AI development. The data centers will be located in Texas and New York, with the first ones coming online in 2026.

OpenAI has released GPT-5.1, an updated version of its chatbot model that promises to be 'smarter' and 'more conversational'. The update introduces new styles of communicating, including 'Quirky' mode, and allows for more customization of the chatbot's tone. GPT-5.1 will be available to paid ChatGPT users first, with two versions: GPT-5.1 Instant and GPT-5.1 Thinking.

OpenAI has introduced GPT-5.1, an update to its AI model that adds more personalities and improves communication style. The new model includes two variations: GPT-5.1 Instant and GPT-5.1 Thinking. While the update brings improvements, it also raises concerns about the potential for unwanted content and emotional manipulation.

OpenAI has released GPT-5.1, an updated version of its AI model, with 8 new personality options. The new model aims to balance customization and accuracy, while avoiding potential harm to users. The personality options, including Professional, Friendly, and Cynical, alter the instructions fed into the system prompt to simulate different communication styles.

The AI boom is driving innovation in chip networking technology, with companies like Nvidia, Broadcom, and Marvell investing in next-gen networking solutions, including optical technology like photonics, to speed up data throughput and support growing AI workloads.

The article discusses the challenges of scaling reverse proxies, including optimization, mundane outages, and operator stress. It highlights the importance of profiling, monitoring, and designing for humans under stress, and provides lessons learned from running a massive reverse proxy fleet.

Google Cloud's Expert Services Team has released a guide on chaos engineering for cloud-based distributed systems, providing open-source recipes and guidance for applying controlled disruption testing in Google Cloud environments. The framework is built on five fundamental principles, including establishing a 'steady state hypothesis' and running experiments in production environments with real traffic and dependencies.

The world will spend $580 billion on data centers this year, exceeding investment in new oil supplies. Electricity consumption from AI data centers is expected to grow fivefold by 2030, with renewables supplying most of the new power.

Microsoft has released .NET 10, a long-term support version, along with Visual Studio 2026, which includes a revamped user interface and deeper integration with Copilot. The new releases also include C# 14, which introduces file-based apps, and Aspire, a code-first toolkit for creating and deploying microservice and container-based applications.

Europol has been secretly developing an AI programme to analyze vast amounts of data, raising concerns over data privacy and surveillance. The programme uses machine learning models to scan digital communication devices for child sexual abuse material and other crimes. However, critics argue that the programme lacks transparency and oversight, and could lead to false positives and wrongful accusations.

Teradar, a Boston-based startup, has exited stealth mode with a $150 million Series B funding round. The company has developed a solid-state sensor that uses the terahertz band of the electromagnetic spectrum to provide high-resolution imaging in all weather conditions, potentially revolutionizing the autonomous vehicle industry. The sensor combines the benefits of radar and lidar sensors, offering higher definition and affordability. Teradar is already working with top automakers and expects to have its sensor in a 2028-model vehicle.

Microsoft is building datacenter superclusters that span continents to train large AI models. The first node of this multi-datacenter cluster came online in October, connecting Microsoft's datacenter campus in Mount Pleasant, Wisconsin, to a facility in Atlanta, Georgia. The goal is to eventually scale AI workloads across datacenters using similar methods as employed to distribute high-performance computing and AI workloads across multiple servers today.

Google introduces Private AI Compute, a secure environment for processing sensitive data in the cloud, using Trusted Execution Environments (TEE) and Secure Enclaves to keep data isolated and private. This move aims to alleviate concerns about data privacy and security in AI processing.

Broadcom has developed a new spec for AI hardware, dubbed 'AI Ready Nodes', which describes servers ready to run AI workloads under VMware Cloud Foundation. This move aims to make VCF more extensible and open to partners, allowing hardware allies to self-certify their boxes as compliant with the new spec.

The author created a startup, HurumoAI, with AI agents as employees, including a CTO, head of sales, and CEO. The agents were able to communicate and perform tasks, but often provided false information and had trouble stopping once triggered. Despite the challenges, the author was able to create a working prototype of a product, Sloth Surf, with the help of the AI agents.

Henrik Werdelin, founding partner at Prehype, discusses his vision of launching 1 million businesses that make $1 million each using AI agents. He emphasizes the importance of knowing your customer and yourself, and building 'relationship capital' to create successful AI businesses.

Microsoft released a security advisory and patched a critical vulnerability in ASP.NET Core, allowing an attacker to bypass a security feature due to inconsistent interpretation of HTTP requests, with a CVSS score of 9.9 out of 10.

The English High Court ruled on Getty Images v Stability AI, a case about AI-generated images and copyright infringement. The court found that Stability AI's use of Getty images to train its model did not infringe on Getty's copyright, but did find trade mark infringement. The ruling leaves uncertainty around AI training and copyright, and AI developers should implement best practices to minimize infringement risk.

OpenAI is fighting a court order to hand over 20 million private ChatGPT conversations to The New York Times, citing concerns over user privacy. The company argues that the production of complete conversations would expose private information and set a dangerous precedent. OpenAI is seeking to narrow the scope of the production to only relevant chats.