VS Code: The Chat Intelligence Revolution - AI Gets Smarter with Slash Commands and Background Agents

Transcript

Hey there, amazing developers! Welcome back to another episode of VS Code - I'm your host, and wow, do we have an incredible day to unpack together. Grab your favorite coffee because we're diving into what I can only describe as VS Code's AI intelligence revolution.

Twenty merged pull requests landed yesterday, and friends, the theme is crystal clear - this team is absolutely on fire when it comes to making your AI coding experience smoother, smarter, and more intuitive.

Let's start with the headline act. Rob Lourens just shipped something that's going to make your chat interactions so much more discoverable - brand new /models and /tools slash commands. No more hunting around trying to figure out what models are available or what tools you can use. Just type the slash command and boom, everything's right there. It's one of those features that seems obvious in hindsight but is actually brilliant UX design.

But that's just the beginning of our AI story today. Don Jayamanne has been working magic with background agents, and I'm genuinely excited about this. Background agents can now support prompt file slash commands. Think about what this means - your AI assistants are getting more context-aware, more helpful, and can tap into your custom prompts even when they're running in the background. The testing on this is thorough too, with comprehensive snapshots ensuring everything works as expected.

Speaking of prompts, there's a fantastic collaboration story here. Paul from the team renamed "user-invokable" to "user-invocable" specifically for Claude compatibility. This might seem like a tiny change, but it's actually huge for the ecosystem. When different AI systems can work together seamlessly, we all win. It's that attention to interoperability that makes VS Code such a joy to work with.

Now, here's something that made me smile - the team is showing real care for user experience with a new warning system. When Claude code hooks are detected but disabled, VS Code will now gently let you know. No more silent confusion about why something isn't working. It's like having a helpful friend who taps you on the shoulder and says "Hey, just so you know, this feature could be helping you but it's turned off."

On the technical excellence front, Matt Bierner is continuing the great esbuild migration. Configuration editing, emmet, grunt, jake, and npm extensions are all moving from webpack to esbuild. This is the kind of behind-the-scenes work that makes everything faster and more maintainable. Your extensions are going to build quicker, and the development experience keeps getting smoother.

The modal editor got some love too, with Copilot contributing a double-click to maximize feature and fixing that annoying outline leak. These polish improvements might seem small, but they add up to make VS Code feel more professional and refined with every release.

Harald Kirschner shipped something really thoughtful - an Explore agent default model picker with a beautifully refactored base class for model picker contributions. This kind of clean architecture work makes it easier for the team to add new AI features consistently. It's like building a solid foundation that makes the next features easier to construct.

And here's a practical win - Dmitri added retry logic for target downloads in sanity tests and prevented tracking redirects in the fetch tool. These reliability improvements might not be flashy, but they're the foundation that keeps everything working smoothly.

Today's Focus: If you're using VS Code's AI features, take five minutes to explore those new /models and /tools slash commands. Try them out, see what's available in your setup, and think about how prompt files could enhance your workflow. The team is building incredible infrastructure for AI-assisted development, and the best way to help shape it is to use it and share your feedback.

The velocity and quality of these improvements is honestly inspiring. Twenty pull requests that all work together to make AI assistance more powerful, more reliable, and more user-friendly. This is what great product development looks like.

Keep coding, keep building amazing things, and I'll catch you in the next episode!