This AI news digest covers the leak of Claude Code's source, OpenAI's strategic exit from video generation, Gemini's new music generation capabilities, and breakthroughs in LLMs learning during inference. It also touches on the rapid improvement of voice-based AI.
Simon Willison explores how AI tools are transforming vulnerability research, potentially making it harder for human experts to discover novel security flaws.
Simon Willison details the Axios supply chain attack, revealing how attackers used individually targeted social engineering tactics to compromise the popular JavaScript library, underscoring a growing threat vector.
Nathan Lambert explores the true indicators of success for open AI models, using Gemma 4 as an example, arguing that factors beyond benchmark scores are crucial for real-world impact and adoption.
The Pragmatic Engineer questions GitHub's viability for AI-native development, reporting a drastic drop in availability (90% uptime) partly due to AI agent traffic, alongside concerns about leadership and strategic direction.
Simon Willison investigates a nuanced web security question: whether JavaScript can escape or bypass a Content Security Policy (CSP) defined within a meta tag inside an iframe. This exploration delves into potential vulnerabilities and browser behaviors.
Frontend Masters provides an in-depth look at the evolving JavaScript ecosystem, detailing new language features, popular frameworks, runtimes, and essential build tools for 2026.
Simon Willison examines how integrating AI coding agents shifts cognitive load for developers, requiring new skills in review and integration rather than simple delegation.
Nielsen Norman Group provides a concrete definition of an AI agent as a system that iteratively takes actions, evaluates progress, and decides next steps to pursue a goal.
GitHub shares insights into the complex engineering challenges and architectural decisions involved in making diff line rendering performant. The article emphasizes that simplicity often leads to better performance.
This article clarifies the distinct scoping behaviors of TypeScript type guards and assertion functions, demonstrating how assertion functions provide persistent type narrowing beyond conditional blocks. It highlights a crucial difference for robust type handling in TypeScript applications.
Nielsen Norman Group discusses how AI can draft surveys efficiently, but human experts remain crucial for identifying subtle design flaws and ensuring data quality.
Cypress Test Replay now automatically captures canvas elements for all projects in Cypress Cloud, simplifying testing for applications that interact with them. The update details what changed, remaining gaps, and how to adjust settings.
The Dagster team details their approach to configuring Copybara for bi-directional syncing, enabling a hub-and-spoke model for managing Git repositories within a monorepo setup.
