Google has updated its NotebookLM research tool to run on the Gemini 3.5 Flash model and the Antigravity agentic coding platform. This integration provides each notebook with a dedicated, secure cloud computer capable of writing and executing code to process research data. The system now includes over 100 software skills designed to automate complex workflows that previously required switching between multiple applications.

Source

Anthropic has launched the Fable 5 family of models, representing a new "Mythos-class" of artificial intelligence designed for high-level autonomous tasks. While the Mythos version is provided to the security community for vulnerability research without restrictions, Fable 5 includes safety guardrails for general use. These models are built on a massive 10 trillion parameter training run, enabling them to handle long-range complex explorations that exceed the capabilities of previous generations.

Source

Industry analysts are questioning the long-term viability of major AI firms, arguing that current growth is driven by semiconductor speculation rather than sustainable business models. Large language model providers face immense losses, with some estimates suggesting expenditures far exceed current revenue streams. Enterprises are beginning to struggle with the high costs of AI tasks as subsidized rates transition to actual token-based pricing.

Source

Amazon Web Services has introduced the AWS FinOps Agent in public preview to help organizations manage and reduce cloud expenditures. This artificial intelligence tool provides a conversational interface that allows users to query spending data using natural language. It integrates with existing AWS services to display infrastructure costs in tabular formats and identify specific areas for optimization.

Source

McDonald's has launched a pilot program for a new artificial intelligence system called ArchIQ at five locations across the United States. This voice-ordering technology, nicknamed "Archy," is capable of processing customer requests in multiple languages, including English and Spanish. The system is built on a multi-year partnership with Google Cloud and utilizes Google Edge Cloud hardware currently being installed across domestic restaurants.

Source

Argentina is considering a new legal framework to attract artificial intelligence companies by establishing a category for non-human corporations. These entities would be operated by AI agents or robots and would not require human shareholders to function. The proposal aims to provide these systems with limited liability protections to manage risks associated with independent machine judgment.

Source

Frontier AI models accelerate the speed of vulnerability discovery and exploit adaptation, allowing attackers to bypass traditional signature-based defenses more effectively. While these models do not change the fundamental stages of an intrusion, they enable the rapid generation of thousands of exploit variations. Security teams must shift focus from the speed of patching to the robustness of the architecture surrounding potential vulnerabilities.

Source

Anthropic leadership describes a management approach where technical vision is balanced with day-to-day operational execution. The company maintains a focus on ethical development and business utility over purely financial or competitive metrics. This strategy includes a philosophy of fiscal responsibility regarding high capital costs for compute infrastructure to avoid overextending resources.

Source

Perplexity Computer is a fully hosted AI agent platform designed to perform tasks like writing and executing code, searching the web, and managing complex workflows. It serves as a user-friendly alternative to self-hosted solutions like OpenClaw, which often require significant technical maintenance and manual configuration. The system operates in an environment where it can install software packages and use frontier models such as GPT and Claude to process requests.

Source

Cloudflare has introduced a new integration that allows its Web Application Firewall to use live threat intelligence data for proactive mitigation. This feature enables the creation of rules based on specific threat actor names, targeted industries, and attack types like DDoS or cybercrime. By populating these fields during the early stages of a request, the system can block high-risk traffic before it reaches the origin infrastructure.

Source

Microsoft has introduced confidential live migration for Azure to improve the availability of highly secure virtual machines. Previously, confidential VMs often required full restarts during host maintenance or patching due to the isolation of protected hardware environments. This new feature allows these specialized workloads to move between hosts with only a brief pause, maintaining operational agility.

Source

Microsoft Research has developed Lens, a text-to-image model that achieves high performance using significantly less computational power than its competitors. While traditional models rely on massive datasets of web-scraped alt-text, Lens utilizes 800 million image-text pairs featuring highly detailed descriptions generated by GPT-4. This approach demonstrates that the quality and depth of training captions are more critical for model efficiency than the sheer volume of raw data.

Source

World Labs is developing large world models designed to move beyond the text-based limitations of traditional language models. By focusing on spatial intelligence, these models aim to perceive and reason within three-dimensional and four-dimensional environments. The project seeks to unlock machine capabilities for interacting with the physical world, impacting fields from creative design to industrial optimization.

Source

Google has entered into a significant infrastructure agreement to pay SpaceX $920 million monthly for access to specialized artificial intelligence computing resources. Starting in October 2026 and running through June 2029, the deal provides Google with approximately 110,000 Nvidia GPUs along with associated CPUs and memory components. These resources are housed within data centers originally established by xAI, the artificial intelligence venture that merged with SpaceX earlier this year.

Source

Corporate boards and financial officers are implementing stricter controls on artificial intelligence spending to combat rising operational costs. Most enterprises currently default to using the most expensive frontier models for all queries regardless of the task's actual complexity. This inefficient approach has led to significant budget overruns as companies pay premium rates for simple requests that cheaper alternatives could handle.

Source

Global governments and private enterprises are investing billions to transition quantum computing from experimental science to industrial-scale technology. The United States and United Kingdom have earmarked significant funding to establish domestic quantum foundries and scale hardware production. This shift aims to integrate quantum processing units into standard data center architectures alongside existing CPUs and GPUs.

Source

Cloudflare has launched new spend control features within its AI Gateway to help organizations manage rising costs associated with large language models. The tool acts as a proxy between applications and AI providers like OpenAI or Anthropic to provide visibility into token usage. It now supports the creation of budgets in dollar amounts rather than just token counts to prevent unexpected overages.

Source

Microsoft and Nvidia are collaborating to transition personal computers from simple tools into autonomous assistants capable of performing complex tasks independently. These modern AI PCs leverage high-performance hardware and integrated software stacks to run sophisticated reasoning models locally at the edge. This shift enables persistent agentic systems that can iterate on coding or design projects even when the user is away from the device.

Source

Automated traffic from AI agents and crawlers now accounts for over 57 percent of global HTTP requests. This shift occurred much faster than industry experts anticipated, largely driven by the rapid expansion of generative AI models. Cloudflare reports that the volume of bot-driven data requests has officially overtaken human-generated traffic within the last few months.

Source