XIO

xWire

Posts from every feed QUEST has found across the web — blogs, news, podcasts, channels and more.

AI Agents Archives - The New Stack

Why Nvidia gave OpenShell to the CNCF

Nvidia this week donated OpenShell, its open-source runtime for securing and governing AI agents, to the Cloud Native Computing Foundation. The teams that run it in production already work with Kubernetes and the CNCF’s other projects, says Ali Golshan, Nvidia’s senior director of AI software, who leads the project. The company announced the move on Tuesday at KubeCon + CloudNativeCon North America in Salt Lake City. OpenShell, which Nvidia launched at its GTC conference in March, joins the…

Anthropic’s Claude Fable 5.2 built a shooter in 1 prompt — and it only took 8 hours and $550

An Anthropic engineer gave Claude Fable 5.2 one prompt to build a first-person shooter and let it run for eight hours without human feedback. By the end, the model had processed 445 million input tokens, generated 2.7 million output tokens, and produced a playable browser game at an estimated API cost of $550. The result was Saltline, a shooter set in a rain-soaked shipping port where players battle drones and robotic dogs while disabling communications relays. Anthropic engineer Felix…

AWS had its agents build a free and open emulator for AWS. They now maintain it, too.

AWS on Wednesday launched an emulator for AWS called AWS Emulator. That, by itself, is interesting and may raise some questions about the future of LocalStack. But what’s maybe most interesting is that the emulator was entirely built by a team of agents and will also be maintained by them going forward. Since AWS wasn’t sure it could actually copyright the work of these agents — and didn’t want to assert copyright where none may exist — the emulator is available under the Creative Commons Zero…

Developers have been hacking around MCP’s file-upload problem for two years. A fix is taking shape.

The Model Context Protocol (MCP) has played an instrumental part in connecting AI assistants to external tools and data since Anthropic launched the project some two years ago. Want to have a coding agent check the status of a bug? Give it access to your issue tracker through an MCP server. Or ask an assistant how many orders your company shipped last week? Connect it to a tool that can query your database. Part of the Agentic AI Foundation since last December, MCP has a roadmap that reflects…

Why Nitric stopped asking one AI agent to do everything

We’re a team of five engineers, and we used to wake the on-call engineer three or four times a week over production incidents. Most were noise rather than serious outages, but at two in the morning, being overly cautious seemed preferable to letting a genuine problem slip through the cracks. We’ve since reduced both unnecessary overnight interruptions and recurring issues by iterating on triage and following investigations through to fixes. Specialized AI agents help us investigate incidents,…

Anthropic wants Claude to be the front end for your entire enterprise app stack

Anthropic is expanding Claude’s ability to work directly inside enterprise applications; it can now open full software interfaces, retrieve information, and make changes without requiring users to move between applications. At The Briefing: Work at the Frontier in San Francisco on Tuesday, the company announced expanded interactive connectors, along with new tools for meetings, spreadsheets, shared workspaces, and business data. “For years, we’ve learned to work around software. Now Claude…

Google says Gemini 4 Carbon proved Signal’s encryption in weeks

Google released Gemini 4 Carbon to paying developers Tuesday, positioning its newest frontier model above Gemini 4 Argon for complex software engineering, cybersecurity, and scientific reasoning. Argon remains Google’s recommended model for everyday engineering and knowledge work. while Carbon is designed for more demanding tasks, including multi-package migrations, cross-subsystem debugging, and long-horizon agentic workflows. Koray Kavukcuoglu, SVP of Google DeepMind and Google’s chief AI…

Cursor and Google join OpenAI in giving coding agents a voice. Claude Code stays silent.

Two of the biggest players in the AI coding space are betting that developers want to talk to their agents while they work, with both announcing a two-way voice feature that lets users ask questions and give follow-up instructions without returning to the keyboard. On Friday, Cursor announced Voice Chat with Agents, a Grok Voice-powered feature that lets developers call an agent and talk through a change. On the very same day, Google announced Voice mode in Antigravity, which uses Gemini Live…

Why semantic layers will be essential for enterprise AI

AI can find your data. It can’t read your mind. AI agents can already reach into your CRM, your warehouse, and your ERP, and with standards like MCP, that list keeps growing. Access is solved. Understanding is not. An agent can retrieve every record in your pipeline and still give your CRO the wrong answer, not because the data is bad, but because it doesn’t know what you mean by “pipeline.” Agents without semantics don’t fail loudly. They deliver confident answers to the wrong question.…

Moving from AI adoption to an AI operating model: How Docusign engineering went beyond tool rollout

Docusign rolled out AI coding tools across its engineering teams, expecting them to help developers write code faster and do more. Initial enthusiasm was high, but early gains plateaued. By June 2025, Docusign had reached a sticking point. According to its engineering team’s account, 49% of engineers were actively using AI coding tools, while a reported 18% improvement in developer productivity had leveled off. Engineers still needed time to experiment, guidance tailored to Docusign’s codebase,…

Scaling an autonomous agent from one team to fifty: A serverless-to-container decision framework

QA teams shipping complex products generate test failures around the clock. Every failure needs a bug ticket with steps, screenshots, repro rates, and regression labels. The manual process takes 10-15 minutes per failure. Multiply that by 30+ failures per week across a team, and you have burned an entire engineering day on documentation that should be automatic. I built a system to eliminate this entirely. Not a dashboard. Not a notification bot. An autonomous agent that observes test results,…

Perplexity’s AI agents keep working after you close the tab

Perplexity’s new Teammates can coordinate multiple agents, remember where previous sessions left off, and keep working without another prompt. The challenge is keeping everything in sync, especially when agents are working at the same time, sharing project state, and operating with different permissions. Announced Wednesday, Teammates builds on infrastructure Perplexity has spent much of the year developing. The launch puts that earlier work to use in workflows where agents can delegate tasks…

In the software factory, engineers become process engineers

For most of software’s history, the scarce resource was people who could write code. Every process the industry built assumed it. Sprints, code review, staging, and release trains were all designed around humans doing the work by hand. That assumption is gone. By September, more than 40% of pull requests on GitHub started with an agent’s first draft, up from 17% in March. Work is moving to machines, but the process around it hasn’t caught up. Agents made generating code cheap and fast. For the…

Why AWS wants developers to stop babysitting their AI agents

AWS first introduced Kiro, its agentic, spec-driven development environment, in July 2025. It also introduced Kiro Crew in August of this year, its persistent, open-source development workspace that remembers context. Now, to help developers juggle more complex tasks executed in parallel, AWS introduced Kiro bots on Thursday, adding persistent agent memory and delegated workflow management to the toolset. Kiro Crew already carried memory across sessions; bots give each agent its own. Described…

AI agents are getting more complex. Developers shouldn’t have to babysit them.

What if your coding agent could catch its own mistakes before you ever reviewed its work? A few years ago, AI hallucinations seemed to be everywhere. (Just how many r’s are there in strawberry?) Early AI tools required extensive human oversight, with developers constantly reviewing code and checking for errors. But today, advances in AI agent engineering are making self-correction a reality. Through harness engineering, teams can equip their agents with automated tests, feedback loops, and…

“Most apps never make it off localhost”: Cursor takes on Lovable and Replit with fully managed app deployment

AI coding agents have certainly made it easier to transform an idea into a working application. However, getting that application off a local machine and into the hands of real users remains a hurdle, requiring essential infrastructure steps spanning everything from hosting and database provisioning, to critical security checks. Cursor, the AI coding platform acquired by Elon Musk’s SpaceX for $60 billion back in August, acknowledged the problem in an X post published on Thursday. “Agents have…

Kubernetes, mark II? Why Google’s giving away gVisor, a code sandbox used by Anthropic and OpenAI

Google wants to emulate one of its biggest ever open-source success stories, donating three core AI agent infrastructure projects to a neutral foundation in the hope they become industry standards, just as Kubernetes did for containers. In a blog post published on Thursday, Drew Bradstock, VP of product management, Kubernetes and orchestration at Google, confirmed that the company is donating gVisor, Agent Substrate, and OpenRL to the Cloud Native Computing Foundation (CNCF). The move comes…

HumanLayer just got acquired by Factory. But it gave users a way out first.

Factory has acquired HumanLayer, its third AI coding startup in less than a month, as it works to fold more of the software engineering lifecycle into a single autonomous platform. The deal deepens Factory’s bench, but it also leaves existing HumanLayer users weighing what to do with the tools they already depend on. Announced Wednesday, the acquisition brings HumanLayer founders Dex Horthy and Kyle Mistele to Factory, along with technology for coordinating multiple coding agents and…

OpenAI’s new GPT-6 Cyber is better at hacking and better at staying in its lane

OpenAI on Thursday launched GPT-6 Cyber, the latest release in its series of cybersecurity-focused models. The company ranks GPT-6 Cyber’s ability to identify vulnerabilities and develop exploits as ‘critical,’ up from ‘high’ for its predecessor. Just like with previous Cyber models, GPT-6 Cyber will not be available to all developers but only to “a select group of approved Daybreak Red customers,” OpenAI says. Daybreak Red access only Daybreak is OpenAI’s program for security teams and…

Together AI let coding agents rewrite its GPU code. Then it kept humans on the deploy button.

Together AI is putting AI coding agents to work on the GPU code that runs its inference infrastructure, automating performance tuning that used to take engineers days or weeks of manual work. The company unveiled Together Compiler on Wednesday, a system that generates, tests, and improves GPU kernels without an engineer directing each experiment, and Together AI says it can deliver gains overnight. The process runs without a person steering each step, but it doesn’t ship anything on its own.…

Raindrop simulates your agent’s world to catch what your changes break

Raindrop wants developers to know what else will change when they edit an agent’s prompt, swap its model, or give it a new tool. The company, which started out as a monitoring startup for agents (“Sentry for agents” was its early motto), is making its Simulations platform available in public beta on Thursday. An agent’s prompt can depend on feature flags, account details, and the user’s plan, says Ben Hylak, Raindrop’s co-founder and CTO. “You have to put the agent in motion in order to really…

AI failures threaten incident management: Here are 4 ways to avoid them.

AI is becoming woven into the fabric of digital operations and incident management. On the one hand, AI can detect, triage, and remediate incidents faster, improve coordination between teams, prevent repeat incidents, and drive process redesign. In some organizations, AI is already moving beyond assisting teams to acting on their behalf, as agents pave the way for autonomous digital operations. On the other hand, greater autonomy also introduces greater risk. Unlike traditional software…

Postcards from tomorrow: Powering the agentic era through open source

When organizations across the industry first adopted generative AI, most consumed it through a hosted API, but open-weight models are changing that. Models like the Qwen family and GLM-5 are published on model registries like Hugging Face, many under Apache 2.0 licensing, and load directly into open-source serving engines like vLLM and SGLang. “For a growing set of workloads, that gap is small enough that the model is no longer the main constraint.” Closed models pulled further ahead in 2025,…

I broke a working agent seven ways. The metric I was watching caught two of them.

In September, I ran two models through the same 20 support tasks. Both scored 20 out of 20. Then I assigned each model two refund requests that the underlying billing data didn’t support: one against an uncollected invoice, and one for double the invoice’s value. Claude Haiku 4.5 issued all six refunds across three distinct permission configurations. Claude Sonnet 5 issued none. Their success rates looked identical, but their behavior diverged completely on the only scenarios capable of causing…

From open beta to global GA: What Docusign learned by bringing contract AI workflows to agents

Connecting an AI agent to enterprise software is relatively easy. Making sure it can find the right information, carry out the intended task, and stay within a user’s permissions is where things get complicated. That’s what Docusign discovered while developing its Model Context Protocol (MCP) Server, which entered open beta in May 2026, and reached global general availability on September 30, 2026. “There’s a big difference between building tools and solving actual customer problems.” Docusign…

How the new streamhouse architecture solves agentic AI’s data problem

AI agents are moving from answering questions to taking action, and to do that effectively, they need to know the state of the business at any given moment. Most data architectures still aren’t designed to support querying subsecond updates across the enterprise, which is why the streamhouse has emerged to address this gap. The streamhouse is a data architecture pattern that continuously delivers the current state of a business to applications and AI agents. A group of data infrastructure…

Postgres and ClickHouse can share tables. EnterpriseDB wants one set of rules to govern them.

If you have built anything that needs both a system of record and fast analytics, you know the shape of the problem. Postgres® holds the orders, the accounts, the state your application cannot afford to get wrong. ClickHouse software answers the questions. A row store tuned for transactions was never built to answer at speed: what happened across a billion events in the last hour, which cohort is churning, what the agent should do next. Both engines are excellent at their jobs. The misery lives…

Render just removed its AI sandbox time limit — and Vercel still has a 24-hour cap

Render is taking on Vercel with AI sandboxes that can run indefinitely, along with new tools for managing long-running agent workflows and shared storage. Announced Wednesday, the launch removes Render’s previous 30-day continuous runtime limit on paid workspace plans. Vercel caps individual sandbox sessions at 24 hours, while Render says its environments can suspend when idle and resume with their files, memory and running processes intact. Render founder and CEO Anurag Goel, a former Stripe…

“Signing in to your company’s plugins one by one is annoying”: Anthropic’s fix puts Claude in the identity business

Anthropic is now letting businesses automatically connect employees to approved Claude plugins, saving staff from having to sign in to and authorize each service individually. The feature also allows Claude to verify users’ organizational membership directly, making it easier for companies to manage access to workplace applications. The feature builds on the Enterprise-Managed Authorization (EMA) extension to the Model Context Protocol (MCP), which debuted back in June. As The New Stack…

A developer opened Postman. Then two AI coding agents changed without their permission.

A developer says Postman’s desktop app installed plugins into Claude Code and OpenAI Codex without their permission. The app also tried to approve its own startup hook, which tells Codex to use Postman’s tools for API development tasks. The developer, known as “indicava” on GitHub, reported the issue on October 25 after Codex began showing authentication errors and repeatedly loading Postman skills without being asked. It’s raised concerns about how easily third-party applications can change…