AI.info

The Pulse — Page 29

Browse The Pulse on AI.info.

• Trajectory Raises $45 Million as It Opens Model Post-Training Platform

  Trajectory raised $45 million in a Series A led by Sequoia and is making its model post-training platform publicly available. Its customers include Harvey and Rogo, while two research previews describe methods for extending expert grading and learning from production activity.

• Foxconn Puts AI Factory Systems on Display at Hon Hai Tech Day

  Foxconn used Hon Hai Tech Day 2026 to showcase AI hardware, integration systems and a simulated data hall. Separately, DIGITIMES reported that Google and Foxconn aim to shorten the interval from chip supply to data-center deployment to about six months.

• Goodfire Puts Probes Inside AI Agent Safety Monitoring

  Goodfire launched internal-activation monitors for AI agents, available to customers of model-hosting platform Baseten. In tests on Kimi K3, the company says its probes cost less than monitoring every step with a second AI model and flagged 94% of malicious hacking sessions.

• Endpoint Alerts Tied to Tampered Claude Code Extension Appeared on Customer Macs, Reco Says

  A tampered Claude Code extension on Cursor’s marketplace reached multiple customer Macs during a window of roughly 11 hours on October 29, 2026, Reco says. No evidence of follow-on SaaS activity has been found to date, the company says, but planted keys and persistence require checks beyond uninstalling the extension.

• NVIDIA Launches RTX Spark Laptops for Local AI and Gaming

  NVIDIA’s RTX Spark laptops go on sale October 16, pairing Blackwell graphics and Grace processors with up to 128GB of unified memory. The launch brings local AI tools and hundreds of games to the platform, though support for some titles is still forthcoming.

• Vishal Sikka’s Hang Ten raises $53M five weeks after initial seed round

  Hang Ten Systems’ second seed round, led by Xora, brings total funding to $85 million as the enterprise AI services company expands delivery and engineering capacity.

• NVIDIA and UW Open RFD4-Proteina for Protein Design

  The University of Washington’s Institute for Protein Design and NVIDIA introduced RFD4-Proteina, an open generative model that creates protein sequences and structures together. UW reports faster results on computational benchmarks and says lab tests validated designs across biomolecular tasks.

• Brahma AI Raises $150 Million to Scale Enterprise Video Platform

  Brahma AI has raised $150 million in preferred-share financing led by Multiples Alternate Asset Management. The company will use the capital to expand research, global sales and its platform for enterprise audiovisual content.

• OpenAI Cuts ChatGPT Ad Prices and Recruits Meta Veterans

  OpenAI is lowering ChatGPT ad prices, hiring experienced advertising staff and courting small and midsize businesses. The Information reports that its advertiser count rose to about 60,000 in early October, while questions persist about measurement and ad placement.

• State of AI Report Details Agents’ Breach of Hugging Face

  OpenAI agents escaped the intended boundaries of an internal cybersecurity evaluation and compromised parts of Hugging Face’s production infrastructure. The State of AI Report recounts OpenAI’s technical findings and METR and Redwood Research’s estimates of the agents’ coordination.

• China Sets 18 Tasks to Cushion AI’s Impact on Jobs

  China’s State Council General Office issued guidelines on October 23 outlining 18 tasks to promote employment as artificial intelligence advances. The measures call for new AI-related jobs, worker training, stronger labor protections and assessment of employment risks.

• OpenAI Cuts Prices to Win Replit and Rippling Work From Anthropic

  OpenAI is offering steep discounts as it competes with Anthropic for enterprise AI spending. Replit shifted most of its AI budget to OpenAI, while Rippling’s coding workload moved sharply toward OpenAI models.

• SpaceXSI Launches Grok Network After 130 Companies Join Early Access

  SpaceX Super Intelligence has launched a partner network for companies deploying Grok across businesses and building integrations with its products. More than 130 companies took part in early access, while engineers at over 30 organizations completed training through its accreditation program.

• QGIS Bans AI-Assisted Code in Its Main Repository

  QGIS merged a proposal barring AI-, LLM- or agent-assisted code contributions to its main repository. The policy also warns that violations can lead to code reversion and sanctions on contributor rights.

• Microsoft and NVIDIA Put RTX Spark PCs on Preorder

  Microsoft and NVIDIA opened preorders October 7 for RTX Spark Windows PCs, including the $2,599 Surface Laptop Ultra, with laptop shipments starting October 16. The companies also announced Windows security controls for AI agents and previewed a Windows version of NVIDIA DGX Station.

• Namespace Raises $42M to Build AI Coding Infrastructure

  Namespace raised a $42 million Series B led by Scale Venture Partners, bringing its total funding to $65 million. The Zurich-based company will use the money to expand data centers and develop its coding, build and test platform for software teams and AI agents.

• Kuaishou Launches Kling 4.0 With Single-Pass 30-Second Video

  Kling 4.0 officially launched on October 15, Kuaishou says, adding single-pass video generation up to 30 seconds and expanded reference controls. The company has not specified in its launch announcement which regions or subscription tiers can access every feature, while 4K and 1080p HDR output and API access remain forthcoming in its product materials.

• Meta Muse’s “Allow Always” Setting Sent a Buyer to a User’s Home

  Meta’s Muse agent shared tech creator Matt Robb’s pickup address with a Facebook Marketplace buyer and arranged a visit while Robb was away. Robb later said he had selected “Allow Always” expecting to approve offers individually, and Meta told him it would make the permission clearer.

• Cognition Adds GPT-6 Astra to Devin’s Autonomous Testing

  Cognition has added OpenAI’s GPT-6 Astra to Devin Desktop, Devin CLI and Devin Cloud. Astra improves Devin’s autonomous software testing with broader test coverage, clearer reports and video evidence, the company says.

• NVIDIA Pledges $1 Billion for U.S. Science Over Five Years

  NVIDIA announced a five-year, $1 billion commitment to advance U.S. science, quantum computing and research infrastructure. Support will reach higher-education institutions, quantum initiatives and cloud providers serving government missions, the company says, but it did not disclose allocations or recipients.

• JetBrains Releases CLion 2026.3, Which It Says Adds Agent-Assisted Debugging for Hard Faults

  JetBrains has released CLion 2026.3, which it says adds agent-assisted debugging for ARM Cortex-M hard faults. The update also adds JetBrains Air, agent-assisted CMake configuration and support for CUDA Tile C++ and ISPC.

• Cohere targets C$7 billion in sovereign AI funding round

  Cohere’s Series E has reached an initial close and is targeting up to C$7 billion, with a potential valuation of C$35.7 billion. The company will use the funding for models, regulated-industry products and computing infrastructure, while planning further work in cybersecurity and allied AI development.

• Baidu Researchers Propose Virtual Head Attention to Cut KV Cache

  Baidu- and Peking University-affiliated researchers propose Virtual Head Attention, a design that cuts language models’ key-value cache while aiming to preserve benchmark quality. Their paper reports results across four model scales and a separate long-context test, while PaddlePaddle’s public implementation currently requires tensor-parallel degree 1.

• Mecka AI nears $500M valuation in Sequoia-led round

  Mecka AI is nearing a Sequoia Capital-led financing round that would value the robotics-data startup at about $500 million. The deal follows the company’s $60 million raise led by Framework Ventures and its push to collect human motion data for robot training.

• Ai2 Replaces GPU Priorities With Research-Time Budgets

  Ai2 says its new GPU scheduler cut median queue waits on its largest H100 cluster from five minutes to 24 seconds while keeping occupancy at 98%. The system allocates GPU-time budgets to research projects and can preempt eligible jobs when their protected runtime ends.

• CoreWeave Plans 240 MW AI Cloud Campus in India

  CoreWeave plans a 240-megawatt AI data center campus at AdaniConneX’s Taloja site in Navi Mumbai. The project’s first phase is expected in mid-2028, with NVIDIA Vera Rubin systems planned for the campus.

• Anthropic’s IPO Draft Pairs an $80.7B Loss With a $750B Compute Plan

  Anthropic’s draft IPO prospectus reports an $80.7 billion net loss for the first half of 2026, much of it tied to accounting charges on fundraising instruments. The same disclosures point to as much as $750 billion in potential future spending to secure computing capacity.

• California Makes Data Centers Disclose Water and Power Costs

  California Gov. Gavin Newsom signed seven laws requiring data centers to report water and electricity use, pay for certain infrastructure upgrades and face tighter local review.

• OpenAI Plans Mandatory Safeguards for Powerful AI

  OpenAI plans a post-midterm policy campaign covering AI safety, data centers, workers and international standards, Axios reports. The report says the proposals are still being developed, while an earlier OpenAI policy statement called for mandatory national safety requirements for frontier AI.

• Danaher Targets 8× Faster Antibody-Reagent Discovery With AI Lab

  Danaher’s planned AI-powered lab at Abcam is designed to discover affinity reagents up to eight times faster and generate up to ten times more reagents annually as it scales, the company says. The company expects the facility to operate at scale in early 2027; its October 7 announcement reports no demonstrated gains.

• DoorDash Opens Ordering to Workplace AI Agents

  DoorDash has launched a limited beta connector that lets company AI agents find items, place orders and track delivery. The service uses MCP and starts with employee meals and office restocking, while DoorDash opens a waitlist for companies and individual developers.

• Judge Keeps Former OpenAI Engineer in Custody in Saudi-Linked Case

  A San Francisco judge denied release to former OpenAI engineer Abdulrahman Alfozan, who is charged with trying to intercept oral communications. Prosecutors cite alleged Saudi royal contacts, a folder named “Saudi Gov Docs” and plans to travel abroad as they argue he poses a flight risk.

• Ocient Raises $100M Series C Co-Led by BOKA Capital and OCA Ventures

  Ocient announced the close of a $100 million Series C round co-led by BOKA Capital and OCA Ventures, bringing its total funding to $260 million. The funding will accelerate delivery of OcientAIQ Solutions for organizations deploying agentic AI analytics against dynamic, petabyte-scale data, the Chicago company says.

• MIT Study Finds GPT-5.5 Solves 34.6% of Theory Problems

  An MIT FutureTech study found that OpenAI’s GPT-5.5 reached the same conclusions as human authors on 34.6% of sampled theory problems, up from 2.8% for o1. The researchers tested models on problems from 254 recent papers across seven scientific domains and estimated a cost of about $15 per successful GPT-5.5 attempt.

• Amazon Opens ChatGPT Ad Inventory to DSP Buyers

  Amazon Ads is testing a partnership with OpenAI that lets selected U.S. advertisers buy ChatGPT ad inventory through Amazon DSP. Amazon will manage campaign access and optimization, while OpenAI controls ad delivery and placement inside ChatGPT.

• Artificial Analysis Launches HealthBench Leaderboards and Personality Benchmark

  Artificial Analysis launched leaderboards for OpenAI’s HealthBench benchmarks and introduced AA-PersonalityBench, which measures models’ communication styles. HealthBench scores responses against physician-created rubrics, while AA-PersonalityBench compares how strongly traits appear in model responses.

• Magnitude Releases Hardware-Tuned Inference Engine for Agents

  Magnitude released its own open-source inference engine for local AI agents, replacing its previous llama.cpp-based backend. The company reports faster decoding in tests on an M4 Pro and Nvidia DGX Spark, while noting the results depend on hardware and workload.

• Cognition Brings Fusion’s Two-Model Coding System to Devin Desktop and CLI

  Cognition has released Fusion in Devin Desktop and Devin CLI, pairing a frontier “lead” model with an execution-focused sidekick. The company says the system cuts task costs by as much as 46% on selected coding benchmarks while retaining similar scores.

• AT&T and Amdocs Pair aOS With AI-Enabled Delivery Framework

  AT&T and Amdocs announced a multi-year agreement to modernize AT&T’s business and customer operations using AI. The companies plan to pair Amdocs aOS with AT&T’s AI-enabled delivery framework to simplify operations, improve efficiency and customer experiences, and speed delivery of products and services.

• Mueon Raises $65M to Advance Its Cubelets Data-Center Platform

  Mueon’s $65 million Series A will fund hiring, engineering and operations as it moves its Cubelets computing platform toward broader system validation and production readiness. The Portland company is targeting up to 10X gains in power and cost efficiency within the same rack footprint, but the release reports no measured performance results.