How to Run an Agent Loop That Actually Converges
A four-checkpoint how-to for agent loops: name a measurable job, bound the tool surface, refresh sandbox fixtures from real traffic, and verify at agent speed.
Loading visual explainer…
A2UI Implementation Checkpoints: Keep Agents and Rendering in Their Lanes
A practical guide to applying A2UI: define the component boundary, separate tool data from UI resources, choose native or embedded rendering, and verify the action path.
Explanation
A2UI lets an agentic system describe an interface as declarative JSON that a host renders with native components from a predefined catalog. This guide turns the architecture into four checkpoints: define the catalog boundary, separate static MCP resources from dynamic tool calls, choose native rendering or a tailored MCP App, and test actions across target clients. The mental model helps engineers preserve integration while avoiding unnecessary iframe fragmentation.
The posts behind it
Related reading
Anthropic · · AnnouncementServiceNow chooses Claude
ServiceNow has selected Claude as the default model for ServiceNow Build Agent and as a preferred model across the ServiceNow AI Platform. Build Agent uses natural-language prompts to help professional coders and citizen developers create applications, design and test agentic automations, while enterprise controls provide access controls, usage monitoring, visibility, and compliance. ServiceNow is also deploying Claude and Claude Code across its workforce of more than 29,000 employees: a Claude-powered sales coaching tool has cut preparation time by up to 95% in testing, while technical teams use Claude Code for coding, reviews, debugging, and repetitive tasks. The partnership targets a 50% reduction in customer implementation time and includes Claude-powered industry applications for healthcare and life sciences, where claims authorization could move from days to hours.
- Anthropic · · Announcement
Apple’s Xcode now supports the Claude Agent SDK
Xcode 26.3 introduces a native integration with the Claude Agent SDK, bringing the same harness that powers Claude Code into Apple’s IDE for longer-running, autonomous coding work. Claude can receive a goal, inspect a project’s full file structure, determine which files to change, modify them, and iterate, while searching Apple’s documentation when it needs framework or API guidance. It can also capture Xcode Previews to visually verify SwiftUI interfaces, identify problems, and refine its implementation, and it supports subagents, background tasks, and plugins. The integration is additionally exposed through the Model Context Protocol, allowing Claude Code users to connect to Xcode and capture Previews from the CLI. Xcode 26.3 is available as a release candidate to Apple Developer Program members, with an App Store release coming soon.
- Anthropic · · Announcement
Anthropic and Infosys build AI agents
Anthropic and Infosys announced a collaboration to develop and deliver enterprise AI solutions for telecommunications, financial services, manufacturing, and software development. The work combines Anthropic’s Claude models and Claude Code with Infosys Topaz, an AI-first set of services, solutions, and platforms, with an emphasis on governance and transparency for regulated industries. Using tools including the Claude Agent SDK, the partners plan to help clients build persistent agents for multi-step work such as claims processing, code generation and testing, compliance reviews, network operations, risk assessment, and legacy-system modernization. They also describe applications for product design and simulation, software debugging, and routine enterprise operations through Claude Cowork, while noting that Claude is available on Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure.
Google · · ExplainerA2UI + MCP Apps: Combining the best of declarative and custom agentic UIs
Agentic UIs must balance deep customization with seamless integration: MCP Apps provide standard-web flexibility in an iframe, while A2UI renders declarative JSON through the host application's native components. The post presents three architectural patterns for combining them, including A2UI over MCP, where MCP servers return application/a2ui+json resources for native rendering and can separate static MCP Resources from dynamic Tool calls. This arrangement avoids iframe-related visual fragmentation, keeps backend tool access separate from frontend rendering, and supports rendering across React, Flutter, or Angular through a predefined component catalog. The broader combination also reserves MCP Apps for highly tailored experiences and describes A2UI in MCP Apps, where an embedded app can process A2UI responses and route user actions through an AppBridge; the conclusion recommends choosing among patterns according to project constraints.
More like this
More how to explainers
BYOC in Five Checkpoints: Name, Split, Land, Govern, Operate
A five-checkpoint walkthrough of BYOC: name the control you need, split control and data planes, build the landing zone, govern data in place, then operate inside your own tenant.
Five Checkpoints to Bitwise-Deterministic Training Runs
A five-checkpoint workflow for making large-scale training runs bitwise reproducible: pick the guarantee, fingerprint the run, and trace divergence down to the kernel.
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
The best engineering reads, once a week
One useful email. Unsubscribe anytime.