News and Tech Update

Anthropic announced it has redesigned Claude Projects to coordinate multiple coding sessions in parallel, with the new experience debuting in beta for Claude Code users. In the redesigned Projects, a coordinator can break a development goal into separate threads, assign them to individual Claude Code cloud sessions and track their progress from one shared project. The beta is initially available to select Claude Pro and Max subscribers using Claude Code cloud sessions, with wider Pro and Max access coming over the next week. Updated Projects across the rest of Claude and on Team and Enterprise plans will follow. The naming may be confusing because Anthropic already uses “Projects” for its general-purpose Claude workspaces. Anthropic describes this as a new version of that existing feature, rather than a separate Claude Code product. The redesign is debuting in Claude Code, while existing Projects in chat and Cowork will continue to work for now and be upgraded as the rollout expands....

StackHawk this week launched Wingman , an artificial intelligence (AI) tool that makes it possible for application developers to automatically fix vulnerability issues as code is being written. Wingman is designed to install into Claude Code, Cursor, GitHub Copilot, Codex, and Antigravity. It scans the live application, interprets findings, and fixes vulnerabilities in a way that makes it simpler for application developers to maintain context as they write code. At the core of those capabilities, StackHawk’s application testing platform is made available through a set of AI skills, hooks and rules that Wingman invokes. Once an AI coding agent finishes a feature, Wingman auto-configures and boots the running application to run a series of security tests with no manual steps required. Findings are then shared with the AI coding agent that wrote the code to fix any issues discovered. Wingman then rescans that application to confirm the fix held before reporting back to the continuous in...

Harness today previewed a revamped user interface for its platform for managing software deployments that makes it simpler for software engineers to manage teams of artificial intelligence (AI) agents using a forthcoming Harness Software Factory platform. Speaking at an {unscripted} NYC 2026 event , Harness CEO Jyoti Bansal told conference attendees the Harness Software Factory will make it simpler for DevOps teams to enforce standards across agentic AI engineering workflows. Rather than allowing AI agents to reinvent the way an application is developed every time, the Harness Software Factory makes use of specifications to enforce standards in a way that both applies controls and ultimately reduces the number of tokens that might otherwise be consumed. Harness also previewed Vibe Mode for Harness, an interface for the Harness platform that applies policies to code being developed by so-called citizen developers who are generating code using AI tools. Finally, Harness revealed a Fle...

Two modules in a repository had near-identical Azure DevOps build and release definitions. A third would have required another pair. The delivery chain used five definitions: one change decider, two builds, and two releases. I consolidated it into one decider, one generic build, and one release with a deployment stage for each onboarded module. The count excludes package and pull request validation. Figure 1. The consolidation removed one duplicated build and one duplicated release. The Decider Passes the Module as a Run Parameter The old builds repeated the same image and chart tasks with different module values. The releases repeated the same deployment tasks with different variable groups. A shared build change required matching edits in both build definitions, and a shared deployment change required matching edits in both release definitions. Adding another module would have required another build and release pair. I named the shared definitions by product, service, scope, an...

Splunk is gearing up to make an artificial intelligence (AI) model for analyzing log data available on Hugging Face under an open source license. Additionally, Splunk at its .conf26 conference this week revealed it is developing a Universal Collector, expected to be available in beta in 2027, to streamline collection of all types of telemetry data using an instance of OpenTelemetry. Raja Mukhopadhyay, vice president of observability cloud for the Splunk arm of Cisco, said a large language model (LLM) that has been trained to reason across log data will make it possible for DevOps teams to observe AI applications and agents at scale. The Universal Collector, meanwhile, lays the foundation for unifying DevOps, IT operations and security operations by making it simpler to correlate events without having to deploy separate repositories for different types of telemetry data. Previously, Splunk has made available an open source Cisco Time Series Model 1.0 on Hugging Face that has been s...

AI agents are becoming operational teammates. They can ingest alerts, query logs and traces, correlate deployment events, consult runbooks and recommend or perform remediation. Their promise is not merely faster summarization: It is a shorter path from detection to safe recovery. But an agent that can explain an incident is not automatically an agent that should change production. Restarting a stateless pod, rolling back a customer-facing payment deployment and responding to a suspected credential compromise are materially different decisions. Each requires a different amount of evidence, human accountability and execution control. The practical answer is tiered autonomy. Tier 1 automates well-understood and reversible incidents. Tier 2 lets the agent investigate and recommend, while a human explicitly approves the action. Tier 3 keeps people in command for novel, complex or high-impact incidents while the agent accelerates evidence collection and hypothesis testing. The three tiers...

When a production system fails, one of the first questions we usually ask is: What broke? Was it a storage service, a bad deployment, a memory leak, a dependency or a network problem? That is a useful place to start, but it can also narrow the investigation too early. Some production failures happen even when every component behaves as designed; the problem appears only in their interaction. Systems-safety methods give us a useful way to reason about this class of failure, and I ran into a good example in a large-scale provisioning system. When Everything Works and the System Still Fails The system created storage partitions on demand and, as part of that process, applied an access-control policy to every new partition. The policy logic lived in a shared library that already had another user: An operator CLI that engineers used to apply policies manually, one storage instance at a time. For the CLI, the design worked well. Each invocation read the current policy state of one storag...