AI.info

The Pulse — Page 61

Browse The Pulse on AI.info.

• Fal Reaches $16 Billion Valuation in Bessemer-Led Round

  Fal’s Series F, led by Bessemer Venture Partners, values the AI media-inference startup at $16 billion, according to Bloomberg as cited by TechCrunch. Fal’s announcement also pairs the Series F with an earlier Sequoia-led Series E and names NVIDIA as a strategic investor.

• Krafton Unveils Ludi 0.2 for Robots Working With People

  Krafton says Ludi 0.2 improves how robots interpret people, retain context and coordinate physical actions. The company presented the system at UC Berkeley on October 30 and announced it on November 2, without reporting benchmark scores or a release schedule.

• OpenAI Brings Three Models and Codex to Google Cloud

  OpenAI and Google Cloud are putting GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna and Codex workflows on Google Cloud’s Model Garden. Select enterprise customers can try the models in an invite-only preview, with general availability planned in the coming weeks.

• Payhawk Ships Agentic Playbooks for Finance Workflows

  Payhawk has released agentic Playbooks for recurring finance work inside Claude, ChatGPT and its own platform, alongside a new Implementation Agent for customer onboarding.

• Architect Labs’ AI Chip Still Needs Silicon to Prove Itself

  Architect Labs’ AI system designed and verified the Redwood inference accelerator in under two weeks from a specification written by two human architects, the company says. Redwood currently runs on an FPGA, while the company projects stronger performance and lower power than Nvidia’s Jetson Orin Nano after fabrication.

• Seattle Times and Newsday Sue OpenAI and Microsoft Over AI Training

  The Seattle Times and Newsday filed a federal lawsuit accusing OpenAI and Microsoft of copying their journalism to train and operate AI products. The publishers seek damages and an order covering allegedly infringing copies, datasets and models.

• Hark Launches Hark Pro, an Assistant That Uses Your Computer

  Hark Pro is designed to carry out tasks through a user’s computer, with a free option and a paid tier for heavier use. Its privacy policy describes the data the service collects and limits its model-training restriction to AI providers covered by written agreements.

• Hungarian Court Sends Third Gemini Copyright Case to EU Judges

  A Hungarian court has referred a third Google Gemini copyright dispute to the EU’s Court of Justice. Judges will examine how copyright rules apply to AI training, chatbot responses and publishers’ technical opt-outs.

• Tencent Expands WorkBuddy and Previews Hy4 and Hy4 Flash

  WorkBuddy has handled more than nine billion task requests since launching in March, Tencent says, while previewing upcoming general releases of Hy4 and Hy4 Flash. The company also outlined WorkBuddy’s international expansion and reported growth in its overseas cloud business.

• OpenAI Deploys Thousands of Agents on Mouse Brain Data

  OpenAI, PerturbAI and Scripps Research used thousands of GPT-Rosalind agents to analyze data from millions of mouse brain cells in 32 minutes. The teams say the system surfaced research leads involving Tsc1, Tsc2 and DCTN1, but the proposed mechanisms still need laboratory tests.

• AMD and Intel Put PyTorch CPU Optimizations in oneDNN

  AMD and Intel will combine their PyTorch CPU optimization work in the open-source oneDNN library. AMD plans to integrate ZenDNN optimizations, giving developers a common software path across AMD and Intel x86 processors.

• Snowflake Puts CoCo and CoWork on Always-On Agent Duty

  Snowflake announced CoCo and CoWork Bots at its November 3, 2026, Expedition event, describing agents that work in the background and seek approval when needed. The company says they operate within users’ access rights, while related automations, integrations and mobile features have different availability statuses.

• Meta Lets AI Agents Set Up WhatsApp Business Accounts

  Meta introduced WhatsApp Business Tools MCP, connecting Claude, Cursor, Codex and ChatGPT to WhatsApp Business setup workflows. The system can create accounts, verify phone numbers, manage templates and test webhooks through an authenticated AI agent.

• Mozilla Finds Open AI Models Just 4.4 Months Behind

  Mozilla’s State of Open Source AI report finds leading open-weight models trail closed systems by about 4.4 months while offering lower costs for many tasks.

• Opera Tracks Whether Coding Agents Follow Through

  Researchers introduced Opera, a critic framework that keeps track of its corrections while coding agents work. Across three benchmarks, the paper reports higher task-resolution rates and tests whether Opera-guided runs can train a smaller model.

• AWS Routes GPT-6 Access to Five Asia-Pacific Regions

  AWS says Amazon Bedrock users in Thailand, Malaysia, Singapore, Indonesia and Taiwan can access OpenAI’s GPT-6 models through global cross-Region inference. GPT-6 Sol and Luna cover all five source Regions, while GPT-6 Astra is available from Singapore; requests may be processed in other supported AWS Regions.

• Eon Benchmark Finds Enterprise Agents Miss Hidden Facts

  Eon researchers tested 12 enterprise-agent setups on questions that required inferring facts across simulated company records. The best setup answered 18 of 24 attempts correctly, while similar-record questions stumped nearly every agent.

• Former AI Employees Warn NYC Council as OpenAI Defends Its Safety Pledge

  Former employees of Anthropic, OpenAI and Google DeepMind warned the New York City Council that advanced AI could escape human control. OpenAI representative Morgan Dwyer said she could not quantify catastrophic risk as council members weighed new oversight proposals.

• Biological Computing Co. Launches Mischief 1 Video Model

  The Biological Computing Co. launched Mischief 1, an image-to-video model based on Wan 2.2 that the company says runs five times faster after optimization with principles measured in living neurons. The model is available on fal and OpenRouter, where listed prices range from $0.005 per second at 480p to $0.01 at 720p.

• Google Tests Gemini Business Calls on Pixel 11

  Google is gradually previewing a feature that lets Gemini call US businesses for routine requests on Pixel 11 phones. The experiment gives users a live transcript and the option to take over, while restricting the calls it can make and the information it can share.

• Insilico Hits First Milestone in Servier’s Up-to-$888M AI Drug Deal

  Insilico Medicine says its oncology collaboration with Servier has reached its first early drug-discovery milestone, triggering an undisclosed payment. The agreement is valued at up to $888 million, but the companies have not disclosed the milestone payment, cancer targets or any drug candidates.

• Apple-Linked RISED Method Trains Agents Across Three Environments

  A paper proposes using behavioral rubrics to select training examples and guide a language-model agent across interactive tasks. In tests on shopping, database and household-task environments, the authors report higher average pass rates than five baseline approaches on two model backbones.

• Experts Draft Incident Rules for AI Agents

  A new arXiv paper draws on 23 academic and industry experts to propose what security reports should record when AI agents are compromised. The framework adds agent memory, autonomy, tool use, delegation and inter-agent communication to conventional incident reporting.

• Punchcard Finds Half of Heavy-Industry Workers Use AI

  A Punchcard Systems survey found that 49% of Canadian heavy-industry workers use AI, while many report no employer training. The study, conducted with Angus Reid Group, surveyed 1,100 working Canadians and found workers trust AI more for paperwork and technical information than for safety decisions.

• Meta’s Muse Reaches 10.2 Million U.S. Downloads in 40 Days

  Meta’s Muse recorded 10.2 million U.S. downloads in its first 40 days, making it one of 12 apps to reach the 10-million-download threshold that quickly. Sensor Tower estimates also show higher early retention than several rival AI apps, alongside increased advertising and rising daily active users.

• Cognition Releases SWE-2 Flash Coding Model in Devin Desktop and CLI, With Weights Not Yet Public

  Cognition’s SWE-2 Flash is available in Devin Desktop and CLI, with faster generation and a lower price than SWE-2. The company reports gains on several coding benchmarks and is developing a program to share the model’s weights with trusted partners and researchers.

• Google’s Diffusion Controller Reports a 90% Win Rate on Stable Diffusion

  Google Research introduced Diffusion Controller, a lightweight method for steering image generation while keeping a pretrained model’s main weights frozen. On Stable Diffusion 1.4, the team reports better preference-alignment results than LoRA in some tests and a 90% win rate for a separately fine-tuned, fully accessible version.

• Microsoft Open-Sources RetroChimera for Molecule Synthesis Planning

  Microsoft Research has released RetroChimera, an MIT-licensed AI model that predicts retrosynthetic routes for small molecules. The system combines two complementary models and outperformed several alternatives in expert evaluations involving difficult synthesis targets.

• Harmonic Security Raises $50 Million to Expand Workplace AI Controls

  Harmonic Security raised a $50 million Series B led by Crosspoint Capital Partners, bringing its total funding to $76 million. The company says it serves more than 100 organizations and will use the new capital to expand engineering and sales operations in the United States and United Kingdom.

• Grid Studio’s Mean Benchmark-Family Score Was 99.9%

  Grid Studio scored highest on a suite of 519 power-system tasks drawn from eight published benchmark families, a paper reports. Its 99.9% figure is the mean of the benchmark-family scores, not a pooled score across all tasks.

• Rivian Buys Third Dimension to Build 3D Driving Simulations

  Rivian acquired Third Dimension, a startup that turns images and video into editable 3D environments, to support driver-assistance and autonomous-driving development. The 15-person team joins Rivian as the automaker reports receiving more than 400,000 customer-fleet driving sequences a day.

• IBM Makes Bob Self-Hostable for Sensitive Software

  IBM made its Bob coding agent generally available for self-hosted deployment on September 30, 2026. The option runs Bob in customer-managed environments, with choices for local models or approved external model services.

• OpenClaw Announces Enterprise Control Plane for Persistent Agents

  OpenClaw announced an open-source control plane for managing persistent agents in sensitive environments, with support for multiple tenants, security boundaries and lifecycle governance. The project is available for internal pilot workloads, and OpenClaw says organizations can self-host it while choosing their own agent components.

• Study Finds Chatbots Match Google Search on Political Knowledge

  A UK survey and randomized trials found that chatbots improved political knowledge about as much as Google search. The survey found that 13% of eligible voters used chatbots to seek information relevant to their electoral choices before the 2024 general election.

• DeepSeek’s $15 Billion Fundraising Round Is Still Being Finalized

  DeepSeek’s latest fundraising has reached about 100 billion yuan, or nearly $15 billion, but Reuters says the round is expected to close in the coming days and its terms could still change. Bloomberg reported the fundraising on October 22, the same day Reuters reported that CATL and Tencent were among its biggest investors.

• Sanders and Casar Propose a Ban on Superintelligent AI

  Bernie Sanders and Greg Casar introduced legislation on September 23 to ban artificial superintelligence and pause development of advanced AI systems. Their bill would create a cabinet-level Department of Artificial Intelligence with authority to review, restrict and shut down systems it deems dangerous.

• Gates Foundation Pledges $1B to Widen AI Access

  The Gates Foundation is committing $1 billion over two years to expand AI access in health care, education, agriculture and local-language technology. The pledge comes as Bill Gates warns that AI could deepen inequality without targeted investment in underserved communities.

• Axiom Formalizes the 15,000-Page Classification of Finite Simple Groups

  Axiom’s AI-agent teams helped produce a 43.1-million-line Lean formalization of the Classification of Finite Simple Groups in 24 days, the company says. The work also surfaced errors and gaps in published arguments, including a flaw in one route to the Unbalanced Group Theorem.

• Vals AI Agents Identify Two Magnetic Semiconductor Candidates

  Claude Opus 5.5 agents helped design YBaMnFeO₅ and identify KV[Cr(CN)₆] as candidates for spin-selective semiconductors, Vals AI says. The predictions come from simulations; neither material’s spin sorting has been measured, and the new compound may be difficult to synthesize in its required ordered form.

• Meta Scales GEM Ads Model Across Platforms and Thousands of GPUs

  Meta’s GEM ads model now combines predictions across Facebook, Instagram, Reels and other advertising surfaces. The company reports a 4x increase in training compute and doubled end-to-end training efficiency on thousands of GPUs.