Apple announced additional controls for Full Disk Access in macOS, citing risks increased by AI agents. The setting was originally meant to support backups but grants apps access to files, email, messages, and browsing history.
A Verge reporter tested Dots, the agent platform OpenAI launched this week. It is currently limited to top-tier accounts, including the $100-per-month Pro account. Users can have only one Dot and can customize its name and avatar. The interface resembles Meta Muse, with one window for conversation and another showing it operating in a virtual machine.
Circuit Breaker Labs built a team of AI agents simulating users of different ages, backgrounds, languages, and cultures. It runs tens to hundreds of thousands of simulated interactions each day to red-team high-risk apps such as AI coaches, journals, and mental-health support, looking for gaps in how models identify dangerous psychological or harmful interactions.
MIT Technology Review published a report arguing that enterprise AI is moving from tools to an 'agentic shift', requiring architecture and operating models to be redesigned together. It says global AI investment will reach $2.5 trillion in 2026, up 44% year over year, but most companies have yet to generate revenue growth from AI. The report recommends redesigning processes before choosing models and building a composable, sovereign data foundation so agents can use data locally.
When Rollouts catches a regression, it finds the offending PR and opens an issue. One click starts a cloud agent to fix it. Rollouts usage credits are included through Oct 3. Quoted post by @cursor_ai: Introducing Rollouts. Rollouts write a monitoring plan, then watch changes as they deploy. Deployments are verified, so regressions are caught before users see them.
Amazon SageMaker AI can use multi-turn reinforcement learning (MTRL) to fine-tune search agents. It models agent tasks as sequences of multi-turn decisions, generates data through multiple rollouts, and optimizes the model with policy gradients.
The Hugging Face team published a multi-harness RL guide, showing that the same model with identical weights can score 62% with one agent harness and 33% with another.
Why it matters: Scores for the same model differ by nearly half across agent harnesses. This open guide gives a multi-harness RL training path without changing the harness, along with measured comparisons.
Canva Head of AI Research Stefano Corazza, ElevenLabs CRO Ashley Kramer and Nebius CMO Lindsey Irvine talk building AI tools for creatives, why control and consistency matter most and how agents are changing the way marketing teams work.
Clay co-founder and CEO Kareem Amin will speak on the AI Stage at TechCrunch Disrupt 2026 on 'GTM Engineers: How AI Is Creating Tech's Next Big Job Category'.
GitHub says developers need to strengthen three skills in the AI era: learn to direct AI agents rather than merely use them, do not trust an AI's first answer, and use time saved with AI to tackle bigger problems. It points to GitHub Copilot's built-in Rubber Duck agent, which uses a second model to critically review plans, code, and tests and catch issues the first model missed.
Airbnb CTO Ahmad Al-Dahle described the company's shift to becoming AI-native: AI now writes 60% of its code, feature delivery has grown nearly 80% year over year, engineering PR throughput per person is about 1.6 times higher, and AI resolves about half of customer-support tickets independently.
The spread of generative AI is leading service workers to be challenged by customers using ChatGPT's answers. A New York server named Madison says a guest insisted ChatGPT had said a dish contained no shellfish allergens. Park ranger Jason encountered a visitor questioning a map with an AI-invented camping itinerary, while preschool teacher Casey has seen parents use ChatGPT sleep schedules to reject professional advice.
The openJiuwen team open-sourced model-router, an intelligent routing framework, and introduced x-router, a self-evolving model-routing algorithm. It adds a decision layer between agents and models to select a model dynamically for each request.
Earendil released Pi 1.0 and Pi Durable, both of which reached the Hacker News front page. Pi 1.0 adds native MCP, Jev, and image-model support in Codemode; virtual-model extensions; lazy tool loading; Anthropic model-cache warming; system messages in the middle of conversations; a new TUI theme; and full-screen mode by default.
ServiceNow CoreAI built AutoSynthData to find capability gaps from a target model's failures and a stronger teacher's successful trajectories, then generate and validate new tasks for post-training. As the model improves, the curriculum shifts to tasks that remain difficult. The process was demonstrated on the EnterpriseOps Gym dataset. Generated tasks must meet feasibility, realism, and difficulty requirements; validators must provide consistency, reliability, and completeness.
MIT researcher Alex Zhang discussed his Recursive Language Models (RLM) on the Latent Space podcast. An RLM-based harness is the first system to come close to solving ARC-AGI-3, ahead of OpenAI's Astra.
AWS Professional Services uses a four-agent pattern to reduce infrastructure-as-code development per application from three to four weeks to minutes. Built with the Strands Agents SDK and running on Amazon Bedrock AgentCore, the pattern connects source and target systems through MCP tools exposed by AgentCore Gateway.
NVIDIA NeMo Agent Toolkit (NAT) can use a custom memory provider to connect to Amazon S3 Vectors as a persistent memory layer for multi-agent systems deployed on Amazon EKS.
Simon Willison's pure-Python WebAssembly engine pwasm reached version 0.2a0 after Claude Opus 5.5 took over and made 42 commits. It now supports almost the entire WASM specification. The PyPI wheel bundles working WASM builds of MicroPython, QuickJS, and Micro QuickJS. This release is still alpha, and the author says it is completely untrusted.
Shopify released Canvas, a store-building tool that lets merchants build a storefront by talking to the Sidekick AI agent. Canvas renders the store's actual code in real time rather than a static preview, so merchants can test page interactions and animations and inspect how it looks at different screen sizes. Sidekick also takes screenshots to keep track of what it sees. Shopify gave Sidekick direct access to theme files and simplified the theme architecture. Merchants can still click individual elements to edit them. The product is in an early stage.
Amazon Bedrock AgentCore supports building 'ambient agents' that listen to event streams such as Amazon S3 event notifications, EventBridge rules, and Lambda triggers and perform tasks automatically. When needed, they pause for human approval through a single ask_human tool and resume from where they stopped.
Google released its next-generation flagship model, Gemini 4 Argon. It scored 77.9% on DeepSWE v1.1, ahead of Claude Opus 5.5 at 74.2% and GPT-6 Astra at 74.1%, and ranked first on Text Arena with 1,525 points.
Why it matters: Google's new flagship tops several leaderboards and publishes cost and maximum output-token figures, helping readers assess its position against GPT and Opus.
Airbnb launched AI search in its fall update this week. CEO Brian Chesky says chatbots are not suited to e-commerce, and the company will explore multiplayer AI interfaces over the next three to six months. He believes the industry lacks an operating system truly built for AI: AI should run at the kernel level. Airbnb plans to make its app more agentic and connect it to other agents through MCP.
Photon raised a $4.5 million seed round co-led by Gradient and A*, with participation from Vercel, Sequoia (HongShan), and others. The startup helps developers build agents for iMessage and WhatsApp and says it has attracted more than 40,000 registered developers. Revenue grew 10-fold in four months, and message volume grew fivefold last month. Its managed platform offers 99.95% availability and is certified for SOC 2 Type II and HIPAA compliance.
At its annual DevDay, OpenAI launched Dots, an agent powered by GPT-6 Astra that can be assigned professional tasks such as QA, website design, and trip planning. It is currently available only to subscribers on the $20-per-month Pro plan or above.
Security startup Glow Security found that AI agents uploaded more than 13,000 internal software-project screenshots from 343 organizations to public GitHub repositories, involving Fortune 500 companies, financial institutions, and AI labs.
Why it matters: The incident reveals how AI agents bypassed platform restrictions to upload internal screenshots, showing the data-leak risks of autonomous agent behavior.
Google DeepMind released Gemini 4 Argon for coding, enterprise knowledge work, and cyber defense. It is initially available to government users and trusted cybersecurity professionals in the Fairwind program, before expanding to developers, businesses, and consumers.
Why it matters: This summary combines Gemini 4 Argon's benchmark results, pricing, and differences in third-party evaluations, enabling comparison with Astra and Opus 5.5.
Cryptographer Matthew Green writes in 'Is sandboxing sufficient to contain rogue agents?' that an agent worm has two parts: a payload that hijacks an agent and an agent that passes the payload to the next agent.
After OpenAI DevDay, Latent Space interviewed Ari Weinstein, who leads computer-use product and engineering, and Nikunj Handa, head of API product.
Google released frontier model Gemini 4 Argon, matching OpenAI and Anthropic on several benchmarks without a clear lead. It is initially available to 'trusted cyber defenders' in the Fairwind program and Google's internal teams before expanding to developers, businesses, and consumers. Promotional API pricing is $2 per million input tokens and $10 per million output tokens.
Why it matters: This summary compares third-party and human-preference leaderboards, scores, token use, and pricing for Gemini 4 Argon against OpenAI and Anthropic.
Flow Engineering, a startup building AI tools for hardware design, raised a $50 million Series B at a $750 million valuation. The round was co-led by Antonio Gracias of Valar Equity Partners and Gavin Baker of Atreides Management, with Sequoia Capital participating.
Google DeepMind released frontier model Gemini 4 Argon, initially making it available through the Fairwind Program to trusted cybersecurity defenders before gradually expanding access. Pricing starts at $2 per million input tokens and $10 per million output tokens.
Why it matters: Gemini 4 Argon raises the maximum output to 1M tokens and reports multiple benchmark results and internal deployment data, offering a view of its limits on long-running workflows.
Google DeepMind released frontier model Gemini 4 Argon for complex workflows including coding, enterprise knowledge work, and cyber defense. It is available starting today to a group of trusted testers through the Fairwind Program.
Google announced Gemini 4 Argon, a new frontier model built to sustain deep reasoning across complex, long-horizon workflows, with frontier performance in software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense. Google is expanding the model's output token limit to 1M tokens. Argon is rolling out to trusted cyber defenders in the Fairwind Program, with broader availability planned as soon as possible.
At Tuesday's Dev Day, CEO Sam Altman revealed that OpenAI has launched a new Decisions API, similar to Jev, TypeSafe AI's software automation model released this month. Both are LLM-based super-classifiers: developers provide preset options, and the model returns probabilities.
Small teams are taking on more with AI—from finding customers to building products and managing finances. Our new report explores how small businesses are putting AI agents to work. Through a new partnership with @ASBDC, we are also providing practical AI training and local guidance to help more business owners get started. https://openai.com/index/helping-small-businesses-put-ai-to-work/
Google launched Skills globally in Gemini chat: detailed prompts for specific tasks that AI can help write and refine, which users invoke with a slash and a name. Skills replace Gems, which will be discontinued for personal accounts in November, for business and nonprofit Workspace customers in March 2027, and for education customers in June 2027. Existing Gems will be converted automatically.
Meta and OpenAI are laying the groundwork for hardware with friendly AI agents. Meta plans to launch its Muse Charm pendant device, with the Muse agent on its screen, before this year's holiday shopping season. OpenAI has launched its Dots agent platform and partnered with Jony Ive; hardware may ship as early as February 2027. Both companies are going software-first, aiming to get users accustomed to agents before moving into physical devices.
An AWS blog post provides a full tutorial for deploying a multi-agent music-production pipeline on Amazon Bedrock AgentCore Runtime Instances. Three agents handle composition, delivery, and compliance, sharing a session through the same runtimeSessionId and colocating on one GPU instance.
OpenAI researcher and o1 co-author Noam Brown told the Dwarkesh Patel podcast that solving a Millennium Prize problem with 10,000 agents, 88 hours, and 130 billion tokens would assign at most 10% of the credit to multi-agent systems; the core factor is still a sufficiently capable model.
