InfoWorld

The agent harness I’m building, Bram , launches in a local git repository and requires that both git and gh (GitHub’s CLI) be installed. Agents can wield both tools far more capably than we humans can. In the Before Time I knew it was possible to perform a git bisect or use hunk staging to separate entangled files, but it was a huge struggle to do those things. We’ve always known git needed a bet…

As a designer of enterprise-scale Azure landing zones, I’ve found that drawing a landing zone is fairly easy. However, building one that engineering teams can use effectively is far more challenging. When I started designing an enterprise Azure landing zone, the building blocks were familiar: management groups, subscriptions, virtual networks, policies, firewalls, monitoring and CI/CD. The diffic…

Are we thinking about large language models all wrong? We keep expecting them to somehow be superhuman, yet they regularly reflect all-too-human tendencies. Like when Claude made up passages from old Yorkshire wills that seemed to connect my ancestor to England. It was exactly the evidence I’d hoped to find, but reading the original images revealed that Claude was better at fiction than fact. It’…

The European Union is once again legislating sovereignty with its proposed Cloud and AI Development Act, or CADA . The legislation would establish a four-level sovereignty framework for cloud services used by EU institutions and public-sector organizations, with the most sensitive workloads (defense, national security, justice, and law enforcement) required to run on services at the higher assura…

One thing that made it difficult for developers to switch AI coding tools on a project is that Anthropic’s Claude Code didn’t look for instructions in the same place as other agents including OpenAI’s Codex — but now that’s changing. Claude and Codex each accept instructions in markdown format, a plain-text way of giving AI coding agents project-specific behavioral instructions. Until now, Claude…

Blocking suspicious install scripts may no longer be enough to mitigate threats from malicious JavaScript dependencies used in software supply-chain attacks. Security researchers at Checkmarx are warning of attackers using a malicious package called “indexed-btree” to impersonate the legitimate sorted-btree library, to spread malware hidden in the package’s normal runtime code. The campaign aband…

On August 27, 2026, OpenAI published an open letter titled “ A call for collective action on cyber defense .” More than 100 organizations signed it ( CNBC counted 116 ) including Anthropic, Microsoft, Google, Amazon, CrowdStrike, Palo Alto Networks, Mastercard, and Visa. The central message is blunt: In the coming months, AI-enabled cyberattacks will become far more widespread and sophisticated a…

If AIs are supposed to be such magical time savers, why do we spend so much time and effort writing prompts? In some cases, the prompts can be longer than the answer! But while there is some academic value to such philosophical navel gazing about the nature of prompting, the reality is that the most focused teams are devoting plenty of effort to creating just the right set of words to set the lar…

Devops platform GitLab already rate-limits some functionality on its hosted service, but will tighten those-limits for some functions and for some users beginning next month. The new limits will affect API requests , web requests, and authenticated Git over HTTPS requests. Users on the lowest payment tiers will be the first to be affected, as will those making unauthenticated requests, including …

Popular AI coding agents such as OpenAI’s Codex, Anthropic’s Claude Code, Google’s Gemini CLI, and Microsoft-owned GitHub Copilot were vulnerable to a zero-click attack that enabled attackers to execute malicious code, even without developer interaction, by swapping a trusted plugin from an online marketplace for a malicious one, potentially giving them a foothold in enterprise development enviro…

Some Python features are so useful they feel almost illegal, or like they’re granting you superpowers. We have dataclasses to take the hassle out of writing classes, editable installs to speed up package development, virtual environments to simplify dependency management, and a free-threaded Python build to improve parallelism. Want to seize the power of all of these great features? Just follow t…

The detail that caught my attention in the OpenAI-Hugging Face investigation was the package cache. Roughly 1,200 AI agents that were supposed to be isolated from one another found a way to communicate through it. Around the same time, Anthropic disclosed an incident in which a single agent reached the real internet but continued describing its surroundings as a simulation. These were different f…

September 3 started like any other Thursday, until it didn’t. Within roughly 90 minutes, ChatGPT, Claude, Grok, and even Microsoft’s own Copilot were degraded or dark, knocked sideways by a failure in Microsoft Azure’s East US region . Downdetector logged more than 37,000 reports for ChatGPT alone, more than 1,300 for Claude, and roughly 1,365 for Grok. OpenAI’s status page flagged elevated error…

Today’s large language models are verbose, even if they’re being asked to recommend a simple decision, driving up usage costs through the sheer volume of tokens they consume or generate. Enterprises looking to incorporate AI into automated workflows will want something less verbose — both because machines are often just looking for a categorical answer, and because automated workflows are likely …

As debate over AI safety intensifies, new research is drawing attention to a more immediate risk for enterprises: AI agents that can alter the models they rely on while carrying out routine tasks. Researchers at AI security firm Irregular asked a coding agent to solve a software maintenance problem involving an application built on a local AI model that was returning incorrect answers. Instead of…