News and Tech Update

Chinese AI developer Z.ai has introduced ZCode, a desktop application that automates software development tasks, positioning the platform to compete with established coding platforms from Anthropic, GitHub and Cursor. The company built ZCode, which it calls an Agentic Development Environment, around its recently released GLM-5.2 LLM. Rather than functioning as a traditional coding assistant that responds to individual prompts, the software manages multi-step development projects by planning work, modifying code, running validation tests and running successive tasks with limited user intervention. Available for Windows, macOS and Linux, ZCode also supports third-party AI models through bring-your-own-key configurations. Developers can monitor long-running coding sessions from mobile devices through integrations with WeChat, Feishu and Telegram, while high-privilege actions require user approval before execution. Z.ai is pairing the launch with aggressive pricing and promotional offer...

Lightrun is providing early access to an ability to verify whether a pull request (PR) will actually run in a production environment as part of its artificial intelligence (AI) platform for automating site reliability engineering (SRE) workflows. Company CTO Leonid Blouvshtein said the Runtime Aware PR Verifier enables DevOps teams to assess the impact a PR will have on a live production environment before it is deployed. At the core of the Lightrun platform is a Runtime Context engine that enables DevOps teams to understand how code truly behaves. Armed with those insights, it becomes possible to both identify issues and bottlenecks before and after an application is deployed in a production environment. Lightrun is now extending that capability to assign risk scores to PRs based on how the change behaves against live execution paths, dependency interactions, and real traffic. Those scores are then natively shared within the context of an existing continuous integration/continuou...

For the last fifteen years, DevOps did what it was supposed to do. It broke down the wall between development and operations, it made continuous delivery a normal expectation, it made shared ownership of production a cultural default, and it moved software delivery from a scheduled event to a continuous flow. The results reshaped how modern software gets built. They also created a new problem: at enterprise scale, every team ends up practicing DevOps a little differently, and the friction adds up. That is where platform engineering enters. It is often framed as the next movement replacing DevOps, but that framing misses what is actually happening. Platform engineering is the discipline of building an internal product — an internal developer platform — that gives every team a consistent, self-service path to production. Templates, golden paths, embedded security, provisioning, observability and compliance evidence stop being tribal knowledge and start being paved roads. Developers mov...

One of the most common questions executives ask right now sounds straightforward: is the agent reliable enough yet? It feels like the right place to start, but the framing quietly points people in the wrong direction because it assumes something about reliability that has rarely been true in complex systems. When people ask whether an agent is reliable, they are treating the agent itself as the unit of reliability, something you either trust or do not trust in the same way you would evaluate a database or an API. That mindset comes directly from how software has traditionally been built over the last few decades. Teams evaluate components in isolation, stack them together, and expect the overall system to inherit the guarantees of the underlying parts. High-stakes human work has never really operated that way, and agentic systems probably will not either. Even today, most production agents are already layered systems in practice. One model plans, another executes, and a third revi...

Ornith, a new family of open source LLM models from the DeepReinforce research collective, takes a novel approach to executing coding and debugging tasks: It generates an architectural framework to give the user’s harness a structured instruction set – a scaffold – to create an agent to complete the job. Available in a set of four variants, the Ornith family was trained to work comfortably with complex software repositories undertaking complicated long-horizon jobs. Sure, LLMs can do these tasks now – until the job gets too complex. Ornith’s self-generated scaffolding ensures that it doesn’t forget the plot along the way. “The model continuously improves not only its code generation abilities but also the orchestration strategy used to solve software engineering problems,” wrote AI tutorial engineer Mehul Gupta, in an introductory post . Deep Reinforcement Expansion Pack Ornith reads the user’s instruction, but instead of executing it directly it builds a scaffold, a learnable ob...

Anthropic introduced a self-hosted gateway this week that lets enterprises run Claude Code on Amazon Bedrock and Google Cloud without the credential sprawl and manual setup that have typically come with deploying AI coding tools at scale. The Claude apps gateway is a single, stateless container that organizations deploy on their own infrastructure and back with a PostgreSQL database. It centralizes identity, policy enforcement, usage tracking, and spend management for Claude Code, addressing a problem that will sound familiar to anyone who has tried to roll out a developer tool across a large engineering org: Every new hire needs a cloud credential, every laptop needs the right settings pushed to it, and finance needs a way to see who’s spending what. Before the gateway, none of that was centralized. IT teams provisioned a credential per developer, manually distributed configuration, and stitched together separate tooling just to get visibility into spend. That’s a lot of ...