VS Code Weekly: AI Gets an Effort Dial and Nested Subagents

Microsoft shipped VS Code 1.113 last week—the third release under their new weekly cadence—and it's starting to feel like the editor is settling into a new rhythm. Not just release rhythm, but product rhythm. The features in 1.113 aren't flashy, but they're the kind that make daily work smoother if you're deep in AI-assisted workflows. What Actually Shipped The headline feature is a Thinking Effort selector right in the model picker. If you're using reasoning models like Claude Sonnet 4.6 or GPT-5.4, you can now dial the reasoning intensity up or down per request—Low, Medium, High—without touching settings. The picker shows your current level (e.g., "GPT-5.3-Codex · Medium") so you always know what you're running. This matters because reasoning effort is a cost-quality tradeoff. Need a quick code explanation? Low effort gets you a faster response. Architecting something complex? Crank it to High and let the model think. I've been using this for about a week now, and it's already change