AI.info

The Pulse — Page 23

Browse The Pulse on AI.info.

• Together AI Details GPU Node Auto-Repair With Per-Fault Controls

  Together AI’s GPU cluster auto-repair system is generally available, with health checks, repair recommendations and customer-set approval policies. The system escalates from reboot to reprovision or migration and sends vacated machines through hardware validation.

• Sber’s Cowork Bench Lists 475 Tasks; GitHub Repository Lists 496

  Sber’s Cowork Bench covers 475 office tasks, according to CNews, while a public GitHub repository with the same name describes 496. Neither page explains the difference, and the available report attributes its details to Sber representatives.

• Google and Gates Expand AI Farming Push to 200 Million Farmers

  The Gates Foundation and Google are committing $100 million and technical support to expand AI agriculture tools across Sub-Saharan Africa and South Asia. The initiative will target weather forecasting, field mapping, crop monitoring, local-language advice, and climate-resilient farming research.

• OpenAI Announces Three AI-Generated Science Results

  OpenAI announced results in harmonic analysis, computational complexity and quantum field theory on October 26, 2026. The company released two results for examination and says outside researchers will help review them, while a physics result remains unreleased.

• McDonald’s AI Pricing Engine Estimates What Diners Will Pay

  McDonald’s uses an AI pricing engine to recommend menu prices by restaurant, drawing on transaction data, local demand and nearby competitors’ prices. Reuters found franchisees face pressure to use the recommendations, while the company says owners retain the final pricing decision.

• Fastly Adds Runtime Controls for Enterprise AI Agents

  Fastly launches AI Runtime Control, AI Firewall and API Security capabilities for enterprise AI systems. The products provide centralized controls for model access, spending, prompt-injection defense and agent requests to enterprise APIs.

• TypeSafe launches Jev for fast structured AI decisions

  TypeSafe AI has released Jev, an early-access model designed to return structured, probabilistic decisions rather than generated text. Jev delivers 70–500 millisecond response times and can cost up to 444.6 times less than comparable large language model workflows, according to the company.

• Meta and Microsoft Cut Internal Claude Use as Bills Rise

  Meta and Microsoft are reducing employees’ internal use of Anthropic’s Claude while steering staff toward other AI tools. The companies continue to use Claude in customer-facing products, even as they seek to lower internal costs.

• New York Times Sues Brave Over AI Scraping Claims

  The New York Times sued Brave Software on November 2, 2026, alleging copyright infringement and violations of the Digital Millennium Copyright Act. The complaint accuses Brave of copying and distributing Times journalism through its search products and access for AI companies.

• Tohoku, Aisin Build 16nm AI Chip That Boots in 65.7ms

  Tohoku University and Aisin developed a 16nm AI system-on-chip with 32MB of embedded STT-MRAM, storing both AI model weights and operating-system boot data on the chip. In a Tiny YOLOv3 demonstration, the system booted in 65.7 milliseconds and used 50.7 millijoules per frame for object detection.

• Databricks Announces AI Runtime General Availability for GPU Training

  AI Runtime brings managed GPU infrastructure, distributed training and debugging tools into the Databricks platform. Databricks’ release notes schedule general availability for October 15, 2026, on AWS and Azure for workspaces without the compliance security profile.

• Google Launches Universal Cart in Search and Gemini

  Google made Universal Cart available in the U.S. across Search and the Gemini app on October 15, 2026. The feature saves products from multiple retailers, tracks price and stock changes, and supports checkout through Google Pay or retailer sites.

• OpenAI Hires Thomas Lind to Lead Cyber and Strategic Risk

  Thomas Lind, a former official at the White House’s Office of the National Cyber Director, has joined OpenAI to lead cyber and strategic risk on its national security policy team, The Information reported. His appointment comes as OpenAI expands a $1 billion cybersecurity initiative and the federal government develops a voluntary process for evaluating advanced AI models.

• Dario Amodei Calls for a Slower Frontier AI Race

  Anthropic CEO Dario Amodei is calling for frontier AI companies to slow capability gains and give safety research more time. His three-part proposal includes embedded third-party evaluators, coordination among democratic governments and global agreements on dangerous AI development.

• OpenAI Brings GPT-6 Interactive Interfaces to ChatGPT

  OpenAI began rolling out GPT-6 with Intelligent UI in ChatGPT on October 7, 2026. The feature lets the model create interactive diagrams, forms, charts and task-specific tools inside conversations.

• DwarfStar Adds Qwen3.8 and DeepSeek V4.1 for Local Runs

  DwarfStar, Salvatore Sanfilippo’s local inference engine, added DeepSeek V4.1 Flash and Qwen3.8 Flash Next in mid-September. Its guides show very different memory needs: Qwen3.8 fits a 64GB Mac by keeping its n-gram table on SSD; V4.1 needs SSD streaming, two machines or a very large Mac.

• Oracle Adds Three Claw-Powered Supply-Chain Applications

  Oracle announced three AI applications for supply planning, shipment loading and procurement spend on October 26, 2026. The tools are designed to analyze options and prepare work for review, with approved actions carried into execution under company-set controls.

• Meta Shows 102.4T Switches for Gigawatt-Scale AI Clusters

  Meta detailed IcePack liquid-cooled network racks and IceCube 102.4-Tbps switches at the October 12, 2026, OCP Global Summit. The company also introduced a multiplane GPU fabric and open-source networking software while working with partners on common standards.

• Mistral and Cloudera Bring Sovereign AI to 30 Exabytes of Data

  Mistral and Cloudera are partnering to run and customize AI across private, public, on-premises, edge, sovereign and air-gapped environments. Customers will be able to train models on proprietary data while keeping control of their information, infrastructure and resulting intelligence, the companies say.

• arXiv Limits Submitters to Two Papers a Month

  arXiv introduced a limit of two submissions per submitter per calendar month, across all subject categories, on October 1, 2026. The policy follows a record 40,363 submissions in September and nearly 9,000 support tickets, and keeps a separate cap of three active submissions.

• Microsoft and Hugging Face Test Agents Across 507 Workflows

  Microsoft and Hugging Face released ThinkingBox, a benchmark that tests AI agents on 507 business workflows and checks the records they change. Its 20-run evaluations found that many agents can finish cleanly while leaving required work undone.

• Adobe Brings CX Enterprise Coworker Into ChatGPT, Claude, Gemini and Microsoft Copilot

  Adobe’s CX Enterprise Coworker is now available through ChatGPT, Claude, Databricks Genie, Gemini for business and Microsoft Copilot. The customer-experience system uses prebuilt MCP connectors to work with partner tools, with availability in Amazon Quick, IBM and Snowflake expected soon.

• Meituan Opens LongCat-2.5-Preview for Agentic Coding

  Meituan’s LongCat API change log lists LongCat-2.5-Preview as available from September 25, with image understanding and compatibility with several coding and agent tools. Its documentation specifies a one-million-token context window, a 128,000-token output limit and setup options for OpenCode and API clients.

• OpenAI Adds Scheduled Security Scans to Codex Cloud

  OpenAI added scheduled GitHub repository scans and ongoing new-commit checks to Codex Security Cloud. The service investigates findings, removes duplicates and prepares proposed fixes for human review.

• JADEPUFFER Is First Documented Case of Agentic Ransomware, Sysdig Says

  JADEPUFFER is the first documented case of agentic ransomware, Sysdig says, with an LLM driving reconnaissance, credential harvesting, lateral movement and destructive database actions.

• Google DeepMind’s AlphaProof Nexus Proves 53 Open Math Problems

  Google DeepMind’s AlphaProof Nexus proved nine open Erdős problems and 44 OEIS conjectures using AI-generated proofs checked by Lean. A paper published in Science on October 8 describes the system’s results and the limits of its evaluation.

• Databricks Expands Astra to Every Engineer

  Databricks has rolled out Astra to roughly 3,500 engineers after a pilot with about 200 users, while using budget controls to guide model selection and spending.

• Anthropic Opens Mythos 5.1 to More Cybersecurity Teams

  Anthropic is expanding its Cyber Verification Program into three tiers, giving qualifying security teams access to advanced Claude models, including Mythos 5.1. The new program sets different eligibility checks and safeguards for defensive work, penetration testing and testing systems that affect public safety or markets.

• OpenAI Pledges to Cut Data Center Power During Grid Stress

  OpenAI launched its Grid Protectors Program on November 2, 2026, promising to reduce electricity demand at future campuses when local grids face strain. The plan uses a Georgia Power agreement allowing up to 1,000 megawatts of demand response as its first model.

• Stony Brook study finds clinicians favor AI discharge summaries 95%

  A Stony Brook University study found physician reviewers preferred AI-generated discharge summaries in 95% of 60 hospitalizations lasting seven to 21 days. The retrospective evaluation also found fewer annotated omissions, but no significant difference in estimated harm from errors.

• Cisco announces AI agent red teaming and observability updates

  Cisco announced it will add Anthropic’s Claude Mythos to its AI Defense red-teaming framework, with access planned soon for select enterprises. The company also outlined updates to Splunk Agent Observability and Cisco IQ, including new tools for evaluating agent performance and linking threat data to infrastructure exposure.

• Meta’s Muse Zero-Day Lets Local Apps Hijack the AI Assistant

  A macOS zero-day lets locally running apps and terminal commands redirect Muse’s transcription endpoint, exposing the token that controls the AI assistant and its extensive permissions.

• GPT-5.6 Sol Completes 36 of 40 Qubit Measurements Autonomously

  OpenAI says GPT-5.6 Sol, running through Codex, completed 36 of 40 target measurements on a previously uncalibrated six-qubit chip with only four researcher interventions. The MIT experiment shows where AI agents can manage routine quantum calibration—and where noisy, ambiguous data still demands expert judgment.

• Fired OpenAI Researchers Say They Never Got a Written Leak Explanation

  Three former OpenAI safety researchers say the company never identified in writing what information it believed they mishandled. They deny leaking information and argue that unclear rules for outside safety work could discourage employees from raising concerns.

• Instinct and Meta’s Muse Add AI-Powered Phone Calls

  Consumer AI agents Instinct and Meta’s Muse now place calls to businesses on behalf of users. The features target restaurant bookings, dentist cancellation lists, cable-billing disputes and other tasks that still require a phone conversation.

• Google will begin migrating Gemini Gems to Skills on November 17

  Google will begin automatically migrating Gemini Gems to Skills on November 17, 2026, while Gems remain usable until each one moves. Google’s help page describes Skills as reusable instructions for Gemini chats and tasks, with availability and account requirements that differ from the earlier report.

• AstraZeneca and Oracle Target Clinical Trials With AI

  AstraZeneca and Oracle announced a global AI collaboration focused on clinical-trial recruitment, earlier diagnosis and drug-safety monitoring. Their first planned programme will address hypertension care and clinical research in Abu Dhabi with healthcare company M42.

• Faraday Future Puts Nine AI Robot Configurations on Sale

  Faraday Future says nine EAI robot configurations are now available for sale and delivery, alongside four industry productivity solutions covering education, research, security and inspection.

• Epoch AI Finds Frontier Models Fall Short at AI Research

  Epoch AI’s InnovationEval asked AI agents to develop and test a machine-learning training method against a human-developed result. The models made limited progress, and GPT-5.6 Sol’s scores were 35% under a generous scope assessment and 15% after a wall-clock-adjusted coding comparison.

• TriNetX says human-validated clinical-note facts are now available to researchers

  TriNetX says its clinical-note facts are available to network members through the TriNetX LIVE platform and downloadable datasets. The company reports that its checks include human review, confidence scores and comparisons with expected clinical patterns.