⚡ VS Code just shipped a big wave of GitHub Copilot upgrades! 🌐 Build and validate web apps with browser agent tools 🤖 Manage parallel agent workflows with the Agents window (Preview) 🎓 New VS Code Learn courses ✨ Plus Autopilot improvements, model discovery for BYOK, cost visibility, and... much more. 📖 Read the full changelog: Happy coding! 💙show more

Wondering if anyone actually uses GitHub Copilot besides MS/GitHub employees. It’s completely unusable now and brings zero value. Like owning a car with insane consumption and no fuel at all. Fast credit burning, API limits... bought Claude Max and couldn't be happier.

This is great, I was using a duct taped workaround. When will Codex be added? 🤞

Cost visibility for BYOK is the line that matters here. Everyone's been running agent loops blind on spend, this is the first "how much did that actually cost" surfaced in the editor itself

When will we get infinite tokens back 😭🤣

Make it lighter via removing telemetry from it also try to make it less RAM consuming hardly anybody is interested in github copilot upgrades

New VS Code Learn courses could be valuable if they teach exception handling, not just happy-path agent flow. Recovery habits are learned early.

👍👍

can the agents window run multiple browser agents concurrently or is it sequential

I really liked the Phone view and Screenshot tool.. Excellent !!

a whole window for managing parallel agents tells you single-agent coding is already the old way

the honest answer is that copilot upgrades are just a drop in the ocean until we figure out how to make these models explain their own decisions

Parallel agents make orchestration UX as important as model quality. Once several workflows are running, clear running, blocked, and waiting-for-input states become essential. Keeping that state visible beyond the editor feels like the next step.

Browser agent tools for validating web apps + parallel sessions is asking for trouble. Web pages are the easiest place to inject instructions that look like part of the task. Parallel agents make conflict resolution and cross-contamination harder to detect. When the agent is "validating" against untrusted web content, how do you stop it from treating page content as higher-authority instructions than the developer's actual intent?

Still missing: a folder 📂 to store and organise conversations topic wise inside the chat panel/ui

The caveat is ownership. Agents can speed the work, but teams still need audit trails for what changed and why.

I'm gonna stick with using Claude code with vs code

W update

i've seen browser agents pass demos, then fail on flaky auth state cleanup

browser agent tools + parallel Agents window is the shape every coding IDE is converging on. the editor is becoming an agent orchestrator.

У меня лимиты студентечкские закончились за пару запросов. Я вообще не понял преимущества для студентов , как будто бы бесплатная версия выгоднее

The agents window addresses visibility, which was never the hard part. Three agents touching the same module still produces three-way conflicts.

Parallel agents inside VS Code sounds cool until my laptop starts negotiating with itself.

Was wondering why my Openrouter models just broke when I opened a new workspace.

Browser tools + parallel agents are the real story here. The hard part is still conflict handling and cost visibility, not another model picker in the UI.