VS Code: AI Customization Revolution
VS Code's March 11th update brings groundbreaking AI customization features with 20 merged PRs and 30 commits. Major highlights include workspace plugin support, enhanced chat troubleshooting skills, agent sessions functionality, and streamlined AI customization management. The team also tackled infrastructure improvements with npm security updates and PowerShell accessibility fixes.
Duration: PT3M49S
https://podlog.io/listen/vs-code-6ffbd97f/episode/vs-code-ai-customization-revolution-b5d82dc8
Transcript
Hey everyone! Welcome back to another episode of the VS Code daily developer podcast. I'm your host, and wow, do we have an exciting day to talk about! March 11th was absolutely packed with innovation, and I can't wait to dig into what the team has been working on.
So grab your favorite beverage and let's dive into what happened yesterday - because honestly, this feels like one of those watershed moments for VS Code's AI capabilities.
The big story today is all about making AI work better for you, in your workspace, with your workflow. Connor Peet landed this fantastic PR for workspace plugin support that I'm genuinely excited about. Here's what's cool - VS Code now reads your .claude/settings files and Copilot equivalents right from your project. It's like VS Code is finally saying "hey, I see how you want to work with AI in this specific project, and I'm going to help you get there faster."
But that's just the beginning! Paul Wang added a whole new chat troubleshoot skill that's going to be a game-changer when things go sideways. We've all been there - something breaks, you're not sure why, and you end up down a rabbit hole of documentation. This new skill is designed to help you debug those frustrating moments more systematically.
Now here's something really exciting - Josh Spicer improved the built-in prompt override saving in AI Customization. You know how sometimes the default prompts are almost perfect, but you just want to tweak them slightly for your workflow? Well, now you can actually edit those built-in prompts without having to mess with the bundled files directly. It's elegant, it's clean, and it respects your workspace setup.
Speaking of agent sessions, Osvaldo Ortega merged a massive PR that adds full agent sessions functionality with comprehensive tests. We're talking over 2,700 lines of new code here, including new GitHub workflows and testing infrastructure. This is the kind of foundational work that enables all the cool features we'll see down the road.
Tyler Leonhardt refined the chat session item handling to use a cleaner API structure, and Matt Bierner made getChatSessionItem return an async iterable, which might sound technical but it means better performance and more responsive chat interactions.
But you know what I love? The team didn't just focus on the flashy AI stuff. They're still taking care of the fundamentals. Matt added min-release-age to all the npmrc files across the codebase - it's a security improvement that protects against supply chain attacks. Not glamorous, but absolutely essential.
Anthony Kim fixed PSReadline for screen readers, making the terminal more accessible. Takashi Tamura enhanced markdown document link handling - now you can use those fancy GitHub-style links with line ranges, like "./src/main.ts#L100-L110". These are exactly the kinds of links that AI models love to generate, so it's perfect timing.
Even the small wins matter - like adding a "Side" option to the minimap context menu, or fixing clipped items in the AI Customization tree. These might seem minor, but they're the difference between software that works and software that feels polished.
Today's Focus time! If you're working with VS Code's AI features, I want you to try something: check out the new workspace plugin support. Create a .claude/settings.json file in your project root and experiment with custom configurations. And if you're using the chat features, give that new troubleshoot skill a spin next time you hit a snag.
For those of you building extensions or contributing to open source, take a page from this update - notice how the team balanced big architectural improvements with small quality-of-life fixes. That's how you build software that people genuinely love using.
That's a wrap on today's episode! The VS Code team shipped some truly impressive AI capabilities yesterday, and I have a feeling we're just getting started. Keep coding, keep experimenting, and I'll catch you tomorrow with more updates from the world of VS Code development!