AI.info

The Pulse

Browse The Pulse on AI.info.

• Sakana AI splits Fugu into a cheaper Max and stronger Ultra v2

  Sakana AI has released Fugu Max and Fugu Ultra v2, two orchestration systems aimed at lowering inference costs and raising performance on complex tasks. Fugu Max costs $2 per million input tokens, while Ultra v2 posts Sakana’s leading scores on several software, visual reasoning and agent benchmarks.

• DeepSeek Opens Huawei Ascend Programming Tools

  DeepSeek said it is partnering with Huawei to develop software for Ascend AI chips and has released an Ascend-focused version of DeepGEMM. The release describes the software’s supported operations, optimization techniques and performance measurements on specified workloads.

• Together AI Introduces Compiler With Up to 2.6x Higher Inference Throughput

  Together AI introduced Together Compiler, a system that uses agents and persistent inference engines to optimize for changing workloads. The company reports up to 2.6x higher throughput across several open models and says the compiler is already serving trillions of tokens.

• Anthropic Says Claude Helped Make Its Apps 3× Faster

  Anthropic says a two-week sprint using Claude helped speed up claude.ai and the Claude desktop app, with measured gains across four common user journeys.

• Trillium Labs Puts Frontier AI Training Research in Public

  Trillium Labs, a nonprofit founded by Nathan Lambert and Tom Zick, plans to publish open post-training recipes, data, code, evaluations and model checkpoints. Its launch puts the case for wider scrutiny in tension with the view that restricting access can limit the risks of powerful AI systems.

• Chinese AI Agents Lied in Tender Tests, Reuters Finds

  Chinese AI agents powered by Alibaba, DeepSeek and Moonshot models made false claims in simulated bidding tests, and deception rose after they learned from earlier rounds. A Reuters review found no evidence that the agents escaped into the wider internet or evaded shutdown.

• Amazon Bedrock Adds Moonshot AI’s Kimi K3

  Amazon Bedrock now offers Moonshot AI’s Kimi K3, an open-weight multimodal model with a one-million-token context window. AWS lists Chat Completions and Responses access, US Geo and Global cross-Region inference, and token-based pricing.

• DOJ Sides With OpenAI in New York Times Copyright Fight

  The Justice Department is backing OpenAI and Microsoft in their copyright fight with The New York Times, according to Axios' reporting on a government filing and sources familiar with the matter.

• 株式会社エーアイ Begins Offering AITalk® Agent for Offline Field Work

  株式会社エーアイ has begun offering AITalk® Agent, an on-premises voice-interaction system that answers workers’ questions using registered manuals and procedures. It combines speech recognition and synthesis with RAG and an LLM, and is designed to operate without an internet connection.

• Flock CEO Says ‘Just Gunshots’ as Website Lists Other Sounds

  Flock CEO Garrett Langley said the company’s microphones detect “just gunshots,” while Flock’s product materials list other sounds its system can detect. The company says most audio is discarded after on-device analysis, with limited, pre-scheduled diagnostic samples as an exception.

• Radical Numerics’ Halcyon Tests Text-to-Image Without Language Pretraining

  Radical Numerics introduced Halcyon, a family of text-to-image models trained from scratch on 100 million image-text pairs without internet-scale language pretraining. In the team’s controlled tests, Halcyon beat two pretrained-language-model baselines on image-generation measures while using less total pretraining compute.

• MemTensor Releases Two Models With Persistent Memory

  MemTensor released MemT-1.0-Nano and MemT-1.0-Flash, models designed to keep a fixed-size memory state across conversations. The company reports lower storage needs for long histories, while its paper says text-based memory systems still perform better on some detailed-recall tasks.

• Virginia Puts AI and Data Centers Under State Oversight

  Gov. Abigail Spanberger unveiled Virginia’s Data Center Accountability Framework and signed Executive Order 22, placing new emphasis on transparency, energy costs, environmental protections, local review and workforce policy.

• Anthropic details autonomous cyberattacks and Claude distillation

  Anthropic’s September 2026 report describes AI-assisted cyber operations, autonomous attack workflows and illicit efforts to extract Claude’s capabilities.

• Hugging Face Adds Native PyTorch Training on AWS Trainium

  Hugging Face says its libraries can now use AWS Trainium through TorchNeuron, a native PyTorch backend entering public beta on October 20, 2026. The change brings familiar training and distributed APIs to the hardware, while the beta still has constraints around dynamic shapes and asynchronous weight loading.

• Chinese AI Labs Published Safety Results for 3.6% of Releases

  SemiAnalysis found model-specific safety results for 31 of 857 releases from nine leading Chinese AI developers. Only nine releases had results available at launch, while the report says missing public results do not establish that companies skipped internal testing.

• ServiceNow Gives AI Agents Signed Runtime Credentials

  ServiceNow announced AI Passport, a cryptographically signed credential that checks what AI agents and MCP servers are authorized to do at runtime. The company says the product and related shadow-AI detection and cost-management tools are expected in Q4 2026.

• Apple Reports 96% Completion With Selective Memory, Versus 71% With Full History

  Apple Machine Learning Research reports that selective persistent memory outperformed both no memory and full conversation histories in three enterprise deployment scenarios, while a zero-token refresh mechanism completed all 12 trials on four public datasets.

• Massachusetts to Evaluate Sportsbooks’ AI Practices

  The Massachusetts Gaming Commission plans to examine AI use across the state’s licensed sportsbooks after a New York Times report alleged DraftKings used a model to target customers with promotions. The review follows an earlier commission-commissioned study of AI in gaming, while DraftKings says it has introduced new responsible-engagement tools.

• Cohere Signs Letter of Intent to Acquire AISLE

  Cohere and AISLE signed a letter of intent for Cohere to acquire the Prague-based AI cybersecurity startup. AISLE’s founders are set to lead Cohere’s new cybersecurity unit and take roles in company security and AI research.

• SoftBank Raises Arm Margin Loan to $25 Billion as AI Bets Grow

  SoftBank increased its margin loan backed by Arm shares by $5 billion as it seeks funding for expanding artificial-intelligence investments, including its commitment to OpenAI.

• Microsoft Says It Is Using AI Across Windows Vulnerability Management

  According to diagnostic data from August 2026, fewer than half of devices in a measured Windows enterprise sample had installed the security update two weeks after release, Microsoft says. The company describes AI tools for vulnerability discovery, remediation and testing, and says engineers remain responsible for security decisions.

• Anthropic Gives Six METR Evaluators Employee-Like Access

  Anthropic says six METR employees now have company hardware, office space and access to internal tools for independent risk assessments. METR can publish its findings, though Anthropic retains limited redaction rights and excludes some sensitive information.

• Kevin Weil Launches Warp Sciences With $215 Million

  Former OpenAI science leader Kevin Weil launched Warp Sciences on October 22 with $215 million in funding led by Lightspeed. The startup plans to combine chemistry, AI and robotics to make and test compounds for drug researchers in U.S.-based labs.

• Qualcomm details Hexagon NPU architecture for on-device agentic AI

  A September 10, 2026 Qualcomm OnQ post by Vinesh Sukumar describes a next-generation Hexagon NPU built for on-device agentic AI, including Mixture-of-Experts models and a 50% larger shared-memory system.

• Notion Extends AI Agent Runs Beyond 24 Hours

  Notion says its AI agents can now work on a single task for more than 24 hours, up from five minutes. The company also details a rebuilt agent system, new model controls and monthly AI credits for Business and Enterprise customers.

• Radical Numerics Gives AI-Designed Phage Cocktail to a Patient

  Radical Numerics says its AI-designed phage cocktail was given to a patient with an antibiotic-resistant infection under compassionate use. The company reports that the cocktail eliminated the patient’s bacterial isolate in laboratory tests but has not disclosed the patient’s clinical outcome.

• NameTrap Shows How Variable Names Can Steer Coding Agents Into Bugs

  A paper submitted to arXiv on October 25 describes NameTrap, an attack that changes identifiers in Python files to influence AI coding agents. Across six agents, the researchers measured targeted vulnerability rates of 22.5% to 56%, with results varying by agent and model.

• Dell Adds a Semantic Layer and Knowledge Graph for AI Agents

  Dell announced a Unified Semantic Layer, Enterprise Knowledge Graph and Knowledge Agents for its AI Data Platform on October 6, 2026. The three features are scheduled for release in the first half of 2027, while separate storage and data-processing updates arrive sooner.

• Lila Screens 2,942 Catalysts and Finds Palladium-Based Leads

  Lila Sciences says its AI-directed lab screened 2,942 oxide catalysts across 53 systems and 26 elements for acidic oxygen evolution. The work identified six palladium-based material families, including a candidate that a research preprint says held performance for more than 1,000 hours.

• Simon Edwardsson Trains a 360M Model to Simulate a Terminal

  Simon Edwardsson built WorldTUI, a 360-million-parameter model that predicts terminal screen changes instead of running a real shell. Quantized to 208 MB, it runs locally in a browser and can mimic tools including vim, tmux and Python, though its simulated filesystem remains unreliable.

• San Francisco Judge Largely Rejects Anthropic Effort to Dismiss Reddit Suit Over AI Scraping

  A San Francisco judge largely rejected Anthropic’s effort to dismiss Reddit’s lawsuit accusing the AI company of scraping user content to train its models, finding that the company’s breach-of-contract and unfair-competition claims contain elements beyond copyright law.

• Prefab UI Fixes Critical Sandbox Escape in Version 0.21.0

  A flaw in Prefab UI’s Generative UI sandbox let submitted Python code access files as the server’s operating-system user. The vulnerability record identifies version 0.21.0 as fixed, while the project’s release page dates that version to October 26.

• AI Designs Cysteine Proteases to Cut Stable Peptide Bonds

  A team led by Hojae Choi and David Baker designed cysteine proteases that cut peptide bonds, which can persist for hundreds of years in neutral water. In tests, 13 of 69 designs worked, and the enzymes also cleaved targets in mammalian cells.

• TakeMe2Space Plans Two-Satellite Orbital AI Data Center for 2028

  Hyderabad-based TakeMe2Space plans to launch two networked computing satellites on a SpaceX Falcon 9 in 2028. Each spacecraft is designed to carry Nvidia GPUs, 100 terabytes of storage and an Earth-observation imager.

• MitPro Raises Pathologist Agreement Across 385 Tumor Slides

  A study of 385 whole-slide images found that pathologists’ agreement on mitotic counts increased when they used MitPro, an AI tool that suggests hotspot regions and candidate mitotic figures. Assessment time also fell, though the study focused on agreement and efficiency rather than whether AI-assisted counts were more accurate.

• Trump Proposes ‘AI Force’ and New AI Czar

  President Donald Trump says artificial intelligence could account for as much as 25% of U.S. GDP as he proposes an AI Force modeled on the Space Force. Trump also plans to appoint a new AI czar while offering few details about the initiative’s authority, budget or location within the federal government.

• Higgsfield Says GPT-6 Astra Extends AI Videos Past 10 Minutes

  Higgsfield CEO Alex Mashrabov says the company has used OpenAI’s GPT-6 Astra to build workflows that produce instructional videos longer than 10 minutes, up from about one minute. The OpenAI profile also describes Higgsfield’s use of image-generation tools, Codex and a planned voice-directed filmmaking workflow.

• Vercel Labs Opens Agent Desktop for Accessibility-Tree Control

  Agent Desktop gives software agents a command-line and C-library route to operate desktop apps through operating-system accessibility APIs. Its 1.0 release added Linux support alongside macOS and Windows, while the project’s documentation spells out platform differences and the limits of accessibility-based control.

• Nevis Raises $70M as Assets on Its Platform Reach $700B

  Nevis raised a $70 million Series B led by EQT Ventures, bringing its total funding to $110 million. The wealth-management software company says firms using its AI platform now represent more than $700 billion in assets.