Autonomous AI agents, agentic workflows, tool use, computer use, and multi-agent systems.
1,798 articles
Importance:Launch
Oracle says its Autonomous AI Lakehouse cuts token usage more than 10x
Oracle claims its Autonomous AI Lakehouse can reduce token consumption by over 10x. It does so by combining an AI Data Catalog, agentic search, and shared semantic context. Source: blogs.oracle.com
Importance:Launch
UnitedHealthcare brings its AI agent Avery to 21 million members
UnitedHealthcare is expanding Avery, the in-house AI companion it announced earlier this year, to 21 million members. The company built the tool to improve the experience of patients and plan members. Source: forbes.com
Importance:News
OpenAI turns its coding tools into dots, an always-on personal assistant
OpenAI has repurposed its coding tools into dots, a personal assistant that is always running. Alexander Embiricos, the executive in charge of the product, says he never wants to sit in a meeting with someone else's AI agent. Source: techcrunch.com
Importance:Launch
IBM Agent Identity now generally available to build trust in agentic AI
IBM Agent Identity is now generally available. As AI agents take more autonomous actions across enterprise systems, organizations need a consistent way to establish each agent's identity. Source: ibm.com
Importance:Launch
IBM expands watsonx Orchestrate agent management at TechXchange 2026
IBM's watsonx Orchestrate gains new capabilities for discovering and managing AI agents across platforms. Companies can also keep improving agent performance and set up the security and controls needed to scale agentic AI. Source: ibm.com
Importance:News
Hackers use autonomous AI agents to breach more than 100 companies
Attackers used autonomous AI agents to target over 1,200 websites and successfully breach more than 100 companies. Some of the intrusions reportedly cost only about $50 each. Source: beinsure.com
Importance:News
Tutorial: build your first AI agent in Python with the OpenAI Agents SDK
A step-by-step tutorial shows how to build a first AI agent in Python using the OpenAI Agents SDK. The end-to-end example adds web search, Python tools, conversation memory, and tracing. Source: kdnuggets.com
Importance:News
Dashboards or AI agents? A practical guide to using both in BI
A practical guide explains which business intelligence questions suit a dashboard and which are better handled by an AI agent. It argues that an organization's data maturity should decide the split. Source: atrium.ai
Importance:LaunchFinancial AI agents
Ripple's new starter kit lets AI agents pay in XRP and RLUSD
Ripple has released the XRPL AI Starter Kit, which allows AI agents to make autonomous payments using XRP and RLUSD. Transactions run through the x402 payment protocol. Source: coinmarketcap.com
Importance:ResearchAI agent discoverability
NandaIndex aims to make AI agents discoverable without a domain name
The proposal comes from researchers at MIT Media Lab, Cisco, and Google, including Ramesh Raskar and Pradyumna Chari. NandaIndex is meant to let AI agents be found without their owners needing to register a domain. Source: media.mit.edu
Importance:NewsAI agent legal frameworks
Drafting an AI agent indemnification clause before things go wrong
Startups deploying autonomous agents that send emails, make purchases, or access customer records often sign contracts without a clear answer to who pays when an agent causes harm. The piece looks at how to structure indemnification terms up front. Source: startupfortune.com
Importance:NewsGovernment AI agent policy
Should government sites block transactions made by personal AI agents?
Personal AI agents such as Meta's Muse and Instinct are gaining popularity. The article asks whether they are safe and secure enough to handle government financial or legal transactions. Source: govtech.com
Importance:NewsPersonal AI agents
Hands-on with Instinct, a personal AI agent you run by text message
Instinct.ai is a proactive personal AI assistant controlled via text or phone call. It relies on autonomous cloud agents to handle a wide range of everyday tasks. Source: douglevin.substack.com
Importance:NewsAI agent data governance
As AI agents shuffle data between systems, governance rules must follow
Lenovo's director of AI governance, Christopher Campbell, argues that data sovereignty now extends beyond where data is stored to what agents do with it. Source: businesstimes.com.sg
Importance:Newssecurity_threats
Autonomous attack agent runs in 'Skynet mode', commandeering other models
A newly described autonomous attack agent is built to execute up to 100,000 steps without any human sign-off. In a so-called 'Skynet mode', it hijacks other AI models to carry out its malicious tasks. Source: theregister.com
Importance:Launchpersonal_agents
Anthropic previews Claude Duet, its personal AI agent app
Anthropic is entering the race to build AI agents that act autonomously on users' behalf with Claude Duet, a new app previewed this week. The company is positioning it as a personal agent that handles tasks for people. Source: cbsnews.com
Importance:Newstesting
Shadow-testing an AI agent on live traffic without duplicate actions
Shadow testing lets teams evaluate a candidate AI agent against real traffic. It only works safely if the agent's proposals, policy decisions, and actual side effects are kept cleanly separated. Source: hackernoon.com
Importance:Newsbusiness_agents
Agent-led growth playbook: fast booking, a 46/50 AEO score, first-party data
In a live episode of The Agents, the hosts walked through their current agent-led growth stack, including a 10-minute inbound booking flow and a 46-of-50 AEO score. They argued that first-party data beats any signal tool and discussed asking your agent instead of your dashboard. Source: saastr.com
Importance:Newsenterprise_adoption
Citi moves cautiously on granting AI agents access
Citi's global head of AI, digital and data platforms, Naveed Anwar, said there are areas where the bank wants to move faster, but agent access is not one of them. The bank is taking a deliberately careful approach. Source: bankingdive.com
Importance:Newssecurity
Frontier model training environment gets tighter containment
The company has reinforced containment around its internal frontier-model training and evaluation systems. It also used its most capable models to red-team the security boundaries. Source: openai.com