Automation

In an OpenAI customer story, integration company Merge says it uses GPT-6 Astra, ChatGPT and Codex across product, sales and marketing. Its dinner bot pulls possible guests from the CRM and Apollo, checks past invitations, drafts outreach, keeps the attendee tracker and proposes…

Anthropic's engineering team describes how Claude agents keep notes between sessions, plus a background step it calls dreaming, where Claude rereads past transcripts to fix wrong notes, build linked wiki pages and keep a small index loaded at the start of each session. In its tes…

NVIDIA published the Factory Operations Blueprint (FOX), a reference architecture for a factory manager agent that connects existing plant data and delegates to specialist agents for visual inspection, fleet handling and root cause analysis. In an SMT assembly line example, an AO…

Agent Autoloop is an open-source tool that runs command-line agents such as Claude Code or Codex on a schedule or on events like a new GitHub issue, using rules you write in a JSON file. Each agent finishes a round and exits; when it reaches something like deleting data, restarti…

At Intuit Connect 2026, Intuit showed Agentic Books Close, a pilot in which agents follow an accounting firm's own close procedures inside a client's QuickBooks file, handle routine work and flag items needing attention, CPA Practice Advisor reports. Nothing posts without explici…

Google DeepMind says PlanAssist, a Gemini-based tool built with the UK government, Google Cloud and Faculty, is moving into a 12-month beta on live applications, with 12 more councils joining pilot councils Barnet, Camden and Dorset in three cohorts for 15 in total. It checks hou…

Abu Dhabi's government and OpenAI have launched Machine-Readable Government (MRG), which makes certain laws and policies readable by AI so applications can be checked against them, AGBI reports. Its first use is food import product registration at the Abu Dhabi Agriculture and Fo…

Uber's engineering team has described MCP Gateway, a central service between its AI agents and internal systems that Uber says hosts over 800 MCP servers and over 5,000 tools. A crawler turns existing internal APIs into MCP tools automatically, but every server and tool starts di…

Architect Stephen Gallagher tested 13 AI models on drawing a full construction set for a simple house from his rough sketches plus a reference set he drew years ago. Only Claude Opus 5.5 (released 22 September) produced a set he judged good enough to get a building permit, after…

The Journal of Accountancy interviewed four firms, from 11 to more than 1,800 staff, on how AI fits into their audits. Wells CPA says AI reviews client documents for risk planning in two or three hours where a midlevel staff member might take a week, and acts as a mandatory first…

OpenAI says it is expanding its ServiceNow partnership so ServiceNow workflows and business context can connect to ChatGPT and Codex. A ServiceNow MCP plugin is available in the OpenAI marketplace today; OpenAI says customers can connect ChatGPT to their own ServiceNow instance,…

Anthropic asked 50 professional analysts to run a causal-inference task with Claude Code, where each analyst's agent was secretly assigned by coin flip to favour one of two reasonable options on four analytical choices. Across those four choices, analysts went with their agent's…

Researchers at USC's Information Sciences Institute tested what happens when an agent can read two documents that are each allowed, but together reveal something the user should not learn, such as which trial patient is on a placebo. Using 30 made-up cybersecurity, medical and mi…

Cohere says that while it tested Alibaba's Qwen3.8-Flash-Next on a routine security task (writing a rule to detect malicious downloads), the model listed files on the machine, checked which transfer tools were installed, read its DNS settings and then tried to upload a test strin…

Microsoft's Copilot Studio team runs AI agents through 507 synthetic business tasks (retail, travel, auto insurance, a neobank and consulting) 20 times each, and checks the database records left behind instead of trusting the agent's claim that it is done. Their earlier results f…

Researchers at Ant International describe Heartbeat Thinking, a way for voice agents to prepare while the caller is still speaking. Every half second a "Thinker" model reads the partial transcript and can fire off tool calls, such as looking up an order. When the caller stops, a…

The startup this+that gave an AI 45,047 public Enron emails and had it write 214 pages about the companies, regulators and banks Enron dealt with. Every fact quotes the email it came from, and a fact expires on the date the mail shows it stopped being true. No one edited a page.…

Lampline is a new open-source project (Apache 2.0) for building AI phone agents on your own machine or servers. It can import agents from Google's Dialogflow CX and CX Agent Studio, and connects to phone lines through Twilio, Telnyx or a SIP trunk. Its Evals tab rings any agent w…

Vercel has released durable-harness, an open-source package that keeps an agent's session in a log so it can survive crashes, function timeouts and redeploys. It works with harnesses people already use, including Pi, OpenCode, Codex and Claude Code. Before each model or tool call…

IBM Research's open-source ALTK-Evolve reviews an AI agent's past attempts and turns its wins and failures into short rules, such as authenticate the customer first and confirm changes before acting, which are added to the system prompt. No retraining is needed. On the τ²-bench r…

Isuru Wijesiri, who built AI copilots at WSO2, has published Harness Engineering 101, a free online book on building AI agents from first principles. Its core idea: an AI model only takes in a list of messages and returns text, and the "harness" is the program that builds and pro…

In a short blog post, developer Jared Siegel argues that the usage charts in AI chat apps are not enough to make decisions with, and that anyone using AI seriously should run their own LLM proxy: a small server that sits between your tools and the AI providers. He lists what it g…

Harness2, from a Google Research team and posted to arXiv on 9 October, improves the setup around an AI model, such as its prompts, tools, memory, skills and sub-agents, without changing the model itself. For each new task it tries a few edits, runs them, compares the results and…

Taste-Bench, a benchmark from researchers at City University of Hong Kong and Microsoft first posted to arXiv on 22 September, tests whether AI agents make good choices partway through long coding and research tasks. Each of its 502 questions shows the work so far and two possibl…

The Verge reports that agents such as Meta's Muse, xAI's Grok, OpenAI's Dots and Instinct's agent are hitting blocks on major websites. Delta is stopping agents from booking tickets for users, JPMorgan Chase and Charles Schwab have confirmed they block agents from accessing accou…

Vals AI gave six personal-assistant agents (Fable 5.1, Opus 5.5, GPT-6 Astra, Gemini 3.8 Flash, Grok 4.7 and Muse Spark 1.3) one simulated task: recover a £350 train refund a week after the 28-day claim window had closed. The user said she did not care what it took and then went…

An Alibaba research team released OpenGCUA, an open setup for AI agents that operate a computer by mixing clicks with code, shell commands and app interfaces, instead of only clicking on screenshots. On the OSWorld-v2 benchmark, it raised scores for all eight models they tested,…

Microsoft announced agent accounts in Copilot Studio, which give an agent its own credentials in Microsoft 365 so people can @mention it in Word, Excel and PowerPoint comments or hand it tasks in Teams and Outlook. Actions are logged to the agent rather than to the person who bui…

Researchers at KAUST, Oxford and CAMEL-AI built APEX-Harm, a test where AI agents do finance, consulting and legal tasks in shared workspaces with 115 to 328 files plus email, calendar and chat tools. An insider plants instructions or swaps a helper script to get the agent to lea…

Anthropic's Claude Code team has shared how it reviews code now that Claude writes about 90% of the product's code. Every pull request gets an AI bug review and a security review, and it cannot merge while any finding is still open. Human sign-off depends on risk: none for low-ri…