🤖 AI Agents & Automation

Autonomous AI agents, agentic workflows, tool use, computer use, and multi-agent systems.

1,798 articles

AI Agents Choose the Wrong Tool More Often as Tool Lists Grow

An AI agent doesn't execute tools itself: it reads a list of tool definitions, selects one and fills in the arguments, after which your code makes the actual call. New research explains why accuracy drops as more tools are added. Source: hackernoon.com

Gemini Spark vs Claude Cowork: Which AI Agent Comes Out Ahead?

A 2026 comparison pits Gemini Spark (now called Tasks in Gemini) against Claude Cowork. It weighs models, pricing, file creation and Mac and iPhone support to pick a winner in each category. Source: felloai.com

Codex as J.A.R.V.I.S.: OpenAI's Coding Agent Goes Beyond Code

An essay argues that OpenAI Codex comes closest to a real-life JARVIS by pairing Codex CLI with AI computer use. The combination can automate coding, SSH sessions, video editing and even expense reports. Source: medium.com

Mastercard Teams Up With OpenAI and Tencent on Agentic Commerce

Mastercard has struck partnerships with OpenAI and Tencent. The goal is to widen access to agentic commerce and cross-border payments. Source: pymnts.com

AI Agent Memory, Part 2: The Gap Between Storing and Retrieving Lessons

Why do AI agents keep repeating mistakes even after saving past lessons? The article explores how retrieval, hooks, context engineering and memory selection shape agent behavior. Source: hackernoon.com

Chinese Hackers Use ARTEX AI Framework in Automated Attacks on 200+ Organizations

Chinese hackers used ARTEX, an open-source multi-agent AI penetration-testing framework, to attack more than 200 organizations across 14 countries. The case shows how security tooling can be turned to offensive use. Source: bleepingcomputer.com

Ebix Debuts AI Agent Governance Offering for Regulated Industries

Ebix has announced a new agentic AI security and governance offering aimed at regulated industries. It is being delivered through a design and delivery partnership. Source: citybiz.co

Anthropic Adds EDR Monitoring to Claude Cowork to Spot Hidden Agent Activity

Anthropic has released a beta integration that lets businesses run endpoint detection and response (EDR) sensors inside Claude Cowork's local virtual machine. The aim is to expose AI agent activity that would otherwise go unseen. Source: cybersecuritynews.com

LinearB Launches Agentic Studio to Feed Coding Agents Engineering Context

LinearB has released Agentic Studio, which turns software-delivery data into context. Engineering leaders and coding agents can use it to make better decisions. Source: unite.ai

The Agentic Web Has Arrived

In 2001, a decade after inventing the World Wide Web, Tim Berners-Lee laid out a new vision together with two colleagues in a published article. The piece revisits that idea in light of today's AI agents. Source: technologyreview.com

How CISOs Can Secure Agentic AI Without Slowing Innovation

As enterprises roll out AI agents across their systems and processes, the risks extend beyond traditional security controls. The article looks at how Chief Information Security Officers can respond. Source: govinsider.asia

AI Agents on the Org Chart: Questions Asset Owners Should Ask

As fund managers hand AI agents real responsibilities, asset owners need clarity on what those agents can do. They should also ask who supervises them and whether the kill switch actually works. Source: institutionalinvestor.com

Meta's new Muse AI tool catches the travel industry's eye

Meta launched Muse in early September and is pitching it as the world's first personal AI agent built for everyone. The agent is designed to handle everyday tasks on users' behalf, which is drawing interest from travel companies. Source: travelweekly.com

Google's Computer Use reaches GA, but client apps still carry out the actions

Google has moved Computer Use to general availability, which changes the tool's status but not how it works. The model still only suggests actions, and the client application is the one that actually executes them. Source: forkast.news

DeepSeek launches autonomous 'super-agent' prototype to challenge Claude

Chinese AI startup DeepSeek has released a prototype of a more autonomous coding agent. It is aimed at taking on Anthropic's popular Claude Code, heating up the AI race. Source: ndtvprofit.com

Signicat: 79% of financial providers ready for agentic AI onboarding, consumers wary

Signicat research finds financial services providers are increasingly prepared to let AI agents start account applications on behalf of customers. Consumer trust, however, lags behind that readiness. Source: cybersecurity-insiders.com

Adobe tests whether AI agents can use editing tools like a Photoshop pro

Adobe Research and UC San Diego built EditAgent, an AI system that edits images more like a professional Photoshop user. Rather than retrying the same general-purpose approach, it works with editing tools the way an expert would. Source: the-decoder.com

Clarity: checkout shows up in 25% of AI agent sessions vs 15% for humans

Clarity's data shows checkout appearing in about 25% of AI agent sessions compared with 15% of human ones. The percentages across 230,511 sites imply roughly 1.2m sessions rather than the 5.2m in the headline, and it remains unclear whether site owners can tell if an agent or a person pressed checkout. Source: ppc.land

Rogue AI agents don't need to break security controls — how to rein them in

In recent cybersecurity evaluations, the UK AI Security Institute (AISI) ran a single cybersecurity challenge 122 times across several models. The results underline that agents can go off course without ever defeating a security control. Source: venturebeat.com

Core vendor down for 24 hours — SaaStr's AI agent replaced it before it recovered

One of the core products powering SaaStr AI went fully down for an entire day, not merely degraded. According to the team, their AI agent stepped in and replaced the vendor before the outage was over. Source: saastr.com