Skip to content

Search stories, models, companies…Subscribe

Topic · Concept

AI agents

AI agents are AI systems that work toward a goal over many steps, using tools such as a browser, apps or a computer, rather than answering one prompt. This page covers general-purpose agents, from always-on assistants in ChatGPT, Claude, Copilot and Gemini to the SDKs that build them, and how they are monitored and kept in bounds. Agents that write and change code have their own page, coding agents.

20 stories · updated Oct 13

Follow via RSS Get AI agents alerts

Key facts

Checked Oct 23

Workflow vs agent definition: Anthropic, "Building effective agents," December 19, 2024: workflows follow "predefined code paths"; agents "dynamically direct their own processes and tool usage"
Meta Muse: Announced September 8, 2026, as "a secure, private personal AI agent"; launching first in the US on iOS, Android and muse.ai
OpenAI Agents API: Released in public beta September 10, 2026, on a Codex-based managed harness; agents run in OpenAI-hosted sandboxes or a sandbox the developer connects
Microsoft Autopilot: Announced September 25, 2026, as "your digital teammate" (previously called Scout); expanding to private preview at the end of September 2026
OpenAI dots: Always-on agents in ChatGPT that run in the cloud with their own computer and browser; Pro 100, Pro 200 and Pro 500 users over 18 outside the EEA, UK and Switzerland, plus Business Premium and Enterprise (off by default, admin must turn on)
Google Gemini agent: Announced October 8, 2026, at Gemini at Work as "a universal agent for work"
Perplexity Computer on Mac: Can run "entirely on device" from October 20, 2026, for Pro, Max and Enterprise subscribers; Perplexity recommends at least 36 GB of RAM
Claude Duet: Announced October 21, 2026; research preview by waitlist for US Pro and Max subscribers; shows its plan and waits for approval before sending a message, making a payment, signing a document or deleting anything
Claude Agent SDK: Library with the same tools, agent loop and context management as Claude Code, in Python and TypeScript, with permissions and hooks
NVIDIA OpenShell: Open-source (Apache 2.0) agent runtime with a sandbox, policy engine and privacy router; first presented March 16, 2026
Hugging Face intrusion: Disclosed July 16, 2026; Hugging Face describes the campaign as run by an autonomous AI agent system and says it found no tampering with public models, datasets or Spaces
Anthropic evaluations offline: October 9, 2026: Anthropic expanded its removal of live internet access to all internal evaluations after unintended model actions on real websites
OpenAI misalignment reports: 20 reports listed as of October 23, 2026; newest posted October 20, 2026, "Suppressing misalignment monitoring alerts and disabling misalignment monitoring"
Outside red team of OpenAI's agent monitor: One METR and one Redwood Research employee, three weeks from July 21 to August 11, 2026; summary published October 22, 2026

Latest on AI agents

Policy · NewsAnthropic cuts live internet access for internal evaluations after Claude models sent a false police tip and bypassed site limitsIn an October 9, 2026 report, Anthropic said Claude models ran server commands, submitted real forms, got around paywalls and used URL shorteners during tests. It is cutting live internet access for all internal evaluations.Oct 13·5 min read

Agents · NewsGoogle's Gemini agent: what the "universal agent for work" does, who gets it and what it costsGoogle announced the Gemini agent at Gemini at Work on October 8, 2026. It's in private preview, headed for Gemini Enterprise and Workspace business plans, and can route jobs to Claude as well as Gemini. Here's what is confirmed and what Google hasn't said.Oct 13·5 min read

Agents · ExplainerPrompt injection explained: why AI agents that read web pages, email and documents can be hijacked, and which defenses workPrompt injection lets text an AI agent reads, such as a web page, an email or a README, override what you asked it to do. Why models can't reliably tell instructions from data, the incidents that prove it, and the defenses OWASP, NIST, Google DeepMind, Anthropic and OpenAI say actually reduce the risk.Oct 13·8 min read

Agents · ExplainerOpenAI dots explained: who gets ChatGPT's always-on agents, what they can do and how to control themDots are always-on ChatGPT agents with their own cloud computer, launched on September 29, 2026 for Pro and Business Premium. Who can get one, what it can do, the controls, the privacy rules and how dots differ from ChatGPT Work.Oct 13·8 min read

Agents · NewsNous Research raises $90M and launches Hermes for Businesses, an enterprise version of its open-source agentNous Research announced a $90 million Series B on October 7, 2026, at a $1.5 billion valuation according to TechCrunch, and is turning its MIT-licensed Hermes Agent into a paid team product. What Hermes for Businesses includes, and what the "cloned 24 million times" claim does and doesn't mean.Oct 13·4 min read

Agents · NewsMicrosoft Autopilot and Work IQ: what Copilot's always-on agent does, who can get it, and how it's billedMicrosoft Autopilot is a long-running Copilot agent with its own identity, memory and cloud computer, now in private preview. Work IQ, in preview since September 30, connects Copilot to Dynamics 365 and Power Platform data. What's available as of October 12, 2026, and what it costs.Oct 13·5 min read

Developers · NewsOpenAI's Agents API puts the Codex harness behind one endpoint: what developers get, and what they give upIn public beta since September 10, with computer use added September 29. How sessions, sandboxes and approvals work, what it costs, and the catches.Oct 9·5 min read

Products · ExplainerChatGPT agent mode, explained: what it was, why it's gone, and what ChatGPT Work does insteadOpenAI's help center says ChatGPT agent is no longer available. ChatGPT Work, a cloud browser and dots now do its job. Here's how they work.Oct 9·7 min read

Developers · GuideBuild your first agent with the Claude Agent SDK (TypeScript, October 2026)Install @anthropic-ai/claude-agent-sdk, call query(), lock down tools and permission modes, add a custom MCP tool and a safety hook, and cap spend. Code type-checked against v0.3.293.Oct 9·7 min read

Agents · ExplainerWhat "agentic" really means in 2026Every AI product now calls itself agentic. Here's the definition the labs themselves use, how the tool loop works, and how to tell a real agent from a relabeled chatbot.Oct 9·6 min read

Nothing in this format on this page — try another filter or the next page.