NerdNews

Meta's Superintelligence Lab has released its first public model, Muse Spark, which represents a significant overhaul of the company's AI efforts. The model boasts strong benchmarks and a new 'Contemplating' mode that enables superior performance with comparable latency. Muse Spark will be available on various Meta platforms, including WhatsApp, Instagram, and Facebook.

Meta released Muse Spark, a new AI model that marks the beginning of an overhaul of its AI efforts. The model is available on the web and the Meta AI app, and is expected to improve over time with the addition of a 'Contemplating' mode. Muse Spark uses multiple AI agents to work on the same problem, generating faster results. Meta plans to release increasingly advanced models, including new open-source models, to push the frontier of intelligence and capabilities.

Databricks co-founder Matei Zaharia has won the 2026 ACM Prize in Computing for his contributions to the field. Zaharia believes AGI is already here and that we should stop applying human standards to AI models. He's excited about AI's potential to automate research and make information more accessible.

Atlassian has launched new AI tools and agents in Confluence, including a visual tool called Remix and three new third-party agents. Remix allows users to turn data into visual assets, while the agents connect Confluence to other apps and services, enabling features like prototype building and presentation creation.

Meta has launched Muse Spark, a new AI model that will power its various products, including Instagram, Facebook, and WhatsApp. The model is designed to handle complex queries and provide more detailed responses, including those involving images and charts. Muse Spark is the first in a new series of AI models from Meta and is seen as a key part of the company's efforts to improve its AI capabilities.

A 2017 ThinkPad laptop was revived by installing Linux Mint, giving it a new life and saving it from obsolescence. The laptop, which was unable to run Windows 11 due to its outdated processor, was able to run smoothly on Linux and perform everyday tasks without issues. This approach can be a cost-effective solution for those who cannot afford a new laptop or want to reduce electronic waste.

Stateful continuation using WebSocket reduces client-sent data by 80% and improves execution time by 15-29% for AI agent workflows, but comes with trade-offs in reliability and portability

Birgitta Böckeler discusses the rapid evolution of AI agents in coding, highlighting advancements in context engineering, agent autonomy, and security concerns. She emphasizes the need for a 'harness' to ensure safe and effective use of AI in software development.

The article covers new features in ECMAScript 2025 and 2026, updates in React, Vue, and Svelte ecosystems, advancements in JavaScript runtimes like Node, Deno, and Bun, and the growing importance of TypeScript and AI in JavaScript development.

OpenAI has released a Child Safety Blueprint to address the rise in child sexual exploitation. The blueprint focuses on updating legislation, refining reporting mechanisms, and integrating preventative safeguards into AI systems. It aims to detect potential threats earlier and ensure actionable information reaches investigators promptly.

Cloudflare and ETH Zurich have outlined approaches to improve cache efficiency in content delivery networks due to AI-driven crawler traffic. They propose separating human and AI traffic into distinct cache tiers and exploring machine learning-driven policies to adapt dynamically.

A recent report from Cisco found that 61% of organizations are using AI in live industrial operations, with 83% planning to increase AI spending. However, network readiness and security posture are primary factors shaping the pace and confidence of AI adoption. The report highlights the need for reliable connectivity, wireless mobility, and edge compute to support AI at scale.

Volkswagen's subsidiary MOIA America and Uber have started testing autonomous microbuses in Los Angeles, with plans to launch a robotaxi service by late 2026. The companies will begin with 10 vehicles and scale up to over 100. The service will initially launch with human safety operators on board, with driverless operations expected to start in 2027.

X is rolling out automatic translation and photo editing features powered by Grok, allowing users to translate posts and edit images using natural language prompts.

Atlassian is introducing AI-powered tools and agents to its Confluence platform, allowing users to turn written notes into graphics and ideas into software applications. The new features, including Remix with Rovo, aim to reduce bespoke requests and security risks by bringing AI to the data inside Confluence. The agents operate within Confluence's existing access controls, requiring user review and confirmation before output is published or deployed.

Microsoft is reevaluating datacenter design in conflict-prone regions after Iran targeted Middle Eastern datacenters. The company's President, Brad Smith, suggests that datacenter security must change and has called for strong international rules to protect civilian infrastructure.

XCOM RAN by Globalstar has announced the launch of an end-to-end private 5G solution designed to support physical artificial intelligence (AI) across several key application areas for industrial automation. The solution includes XCOM RAN's Supercell architecture, XCOM Radio Series, XCOM Core, and XCOM Orchestrator, and offers spectrum flexibility with support for Band n48 shared spectrum in the US and Band n78 allocated for private 5G and industrial use in Europe and parts of Asia.

Kelsey Hightower, a former Google distinguished engineer, recommends rebranding existing automation tools like Bash and cURL as 'zero-token architecture' to thrive in an AI-driven world. He emphasizes the importance of developing deep technical skills and understanding technology platform fundamentals to remain relevant.

Investors are backing nuclear power to fuel the UK's AI datacenter buildout, with $370 million injected into the sector. This is driven by the growth in power-hungry AI infrastructure and geopolitical energy volatility. Nuclear startups are offering always-on baseload power, which is seen as a sustainable solution for the UK's AI ambitions.

Anthropic's Project Glasswing uses a powerful Large Language Model (LLM) to identify and remediate undiscovered cybersecurity vulnerabilities in critical software. The model, Claude Mythos Preview, has already discovered thousands of zero-day vulnerabilities, including a 27-year-old vulnerability in OpenBSD and a 16-year-old vulnerability in FFmpeg. The company has committed up to $100m in usage credits to support the project and will provide $4m in donations to open-source security organizations.