VS Code: AI Terminal Security & User Control Revolution

Transcript

Hey there, code friends! Welcome back to another episode of VS Code. I'm absolutely buzzing with excitement about today's changes because March 19th delivered something really special - we're talking about 20 merged pull requests that are fundamentally changing how we interact with AI in our development workflow.

Let me paint you a picture of what's happening here. You know how we've all been navigating this balance between giving AI tools enough power to be genuinely helpful, while keeping our systems secure? Well, the VS Code team just made a massive breakthrough on that front.

The star of today's show is Alexandru's work on terminal sandboxing. Here's what's brilliant about this - when you have terminal sandboxing enabled, AI models can now explicitly ask for permission to run commands outside that sandbox. It's not sneaky, it's not hidden - the AI literally says "Hey, I think this would work better unsandboxed, here's why" and you get to make that call. It's like having a really smart assistant who knows when to ask for the car keys instead of just taking them.

But that's just the beginning. Josh has been absolutely crushing it with AI customization features. You can now disable and enable built-in skills with simple toggle switches. Think about that for a second - you're no longer stuck with whatever AI capabilities ship with VS Code. You can curate your AI experience like a playlist. Don't want a particular skill running automatically? Just flip it off. Changed your mind later? Flip it back on. The control is entirely in your hands, and it's beautifully implemented with these intuitive eye icons that show up on hover.

Speaking of beautiful implementations, let's talk about the new thinking design. This one made me smile because the commit message literally says "new thinking design!!!!" with four exclamation points, and honestly, that enthusiasm is infectious. They've removed borders, aligned everything left, and added this cool curly connector between headers. It sounds simple, but these visual improvements make AI reasoning so much more readable and engaging.

Tyler tackled something that might seem small but feels huge when you hit it - focus management in QuickInput components. You know that jarring moment when focus just disappears and you're clicking around trying to figure out where your cursor went? Yeah, that's getting fixed. It's one of those quality-of-life improvements that you don't notice until it's there, and then you wonder how you ever lived without it.

The image carousel got some love too, with captions and counters that make reviewing multiple images actually pleasant. And there's this thoughtful touch where cloud delegation options are grayed out for non-git workspaces - it's preventing a frustrating dead-end before you even hit it.

What I love about today's changes is the theme running through all of them: user agency. Whether it's asking permission for unsandboxed execution, letting you disable skills you don't want, or preventing workflows that won't work anyway - every change is about putting you in the driver's seat while making AI more capable.

Today's Focus: If you're working with AI tools in VS Code, take a few minutes to explore the new skill management options. Think about your workflow and which built-in skills actually add value for you. And if you're doing anything with terminal automation, definitely check out the new sandboxing controls - they might change how you think about AI-assisted development.

The development pace on AI integration is absolutely incredible right now, and it's clear the team is laser-focused on getting the balance right between power and control. That's what good tools do - they amplify your capabilities without taking away your agency.

That's a wrap for today! Keep building amazing things, and I'll catch you next time with more exciting updates from the VS Code universe. Until then, happy coding!