I have a background in web development but almost no experience with iOS. I experimented with how far I could get building an iOS app using Codex, an AI coding agent — focusing on learning rather than vibe-coding.

Playwright CLI v0.1.9 added annotations, a useful way to give AI agents visual feedback by selecting browser elements and leaving comments.

The `contrast-color()` function returns either `white` or `black`, whichever has a higher contrast ratio against a given color. It helps when colors change dynamically or users apply custom themes.

Google's A2UI protocol lets AI agents send declarative component definitions instead of executable code, so clients can safely render generated UI with native widgets.

Generative UI is gaining attention, but AI output is unpredictable. json-render reduces that risk by making AI generate JSON from a predefined component and action catalog.

The `ariaNotify()` method is a Web API for notifying assistive technologies about dynamic content updates. Unlike traditional WAI-ARIA patterns based on declarative attributes, it lets JavaScript send notifications at the exact timing you intend.

GitHub CLI now includes the `gh skill` command, making it easier to install, search, and manage agent skills hosted on GitHub. This article walks through how to use `gh skill` in practice.

apm is a package manager for AI agents by Microsoft. Like npm or pip, it resolves dependencies while packaging agent skills and MCP as manifests. This article covers the basics of using apm.

The new advisor tool in Claude lets you handle routine tasks with lightweight models while consulting a more capable model only when needed. This article explains how to use the advisor tool in Claude Code.

HTML in Canvas API is an experimental API proposed at WICG for rendering HTML directly inside a canvas. This article explains how it works, what problems it solves, and the kinds of use cases it enables.

Claude Managed Agents provides the harness and infrastructure to run Claude as an autonomous agent. It is designed for long-running and asynchronous tasks. This article walks through trying it out in practice.

Choosing how much command auto-approval to allow in coding agents is a key design tradeoff between usability and security. This article explains how Codex sandboxing works and how approvals are handled for commands run outside the sandbox.