JetBrains Fleet replacement and the editor performance gap

JetBrains discontinued Fleet and is building a new agentic development environment where developers supervise AI agents. This change occurs as teams select between Zed and Visual Studio Code for Kotlin workflows. Zed is written in Rust and uses a custom GPU-accelerated rendering framework called GPUI. This framework talks directly to the GPU using Metal on macOS or DirectX on Windows, which removes the abstraction layers that make Electron-based editors like Visual Studio Code slower. In May 2026, ToolChew benchmarked the newly stabilized Zed v1.3.6 build, which clocked cold starts at 0.12 seconds against VS Code’s 1.2 seconds and used 222 MB of RAM compared to 3,549 MB. This 16x memory advantage allows developers to run Docker containers or databases without hitting memory ceilings. In August 2026, CodeOxi measured typing latency under heavy load and found that Zed stayed under 10ms while VS Code climbed past 50ms. A clean install shows a 7.5x difference in startup time, where Zed runs at 0.40 seconds and VS Code runs at 3.00 seconds. For users who prioritize speed, Zed provides 2ms input latency compared to 12ms for VS Code. When opening 100,000-line files, Zed is 8x faster with a 0.15s render time compared to 1.19s for VS Code.

Kotlin performance and editor selection

Kotlin developers utilize the K2 compiler for speed, as JetBrains reports 40% compile time reductions in its IntelliJ monorepo and K2 benchmarks show 376% faster incremental analysis. In IntelliJ IDEA 2025.3, code analysis saw a 39% improvement and completion saw a 26% improvement. Developers migrating from KAPT to KSP gain up to 2x speed in annotation processing, and the 2.5 million Kotlin users benefit from K2 features like Rich Errors and multi-dollar interpolation. Kotlin/Wasm reached Beta in September 2025, while Compose Multiplatform 1.5.10-rc02 provides enhanced rendering performance on iOS. Kotlin/Native release builds are up to 40% faster in 2.3.0, and Kotlin usage in Spring is also measurable with 27% of Spring developers reporting they have used it for production. The JetBrains AI Assistant works with many languages including Java, Kotlin, Python, Go, Rust, and JavaScript/TypeScript. It uses the IDE project structure to provide context to LLMs for refactoring and code completion. 46% of developers distrust AI accuracy according to Stack Overflow 2025 data. Will the Agent Client Protocol successfully standardize AI agent integration across different IDEs?

Comparing AI workflows and costs

JetBrains AI Assistant is the most logical choice for teams already paying for IntelliJ IDEA or PyCharm. It is included in existing subscriptions and uses the real code model of the IDE. GitHub Copilot works inside VS Code and JetBrains, whereas Claude Code works from the command line for agentic work.

Visual Studio Code has an ecosystem of over 100,000 extensions, while Zed uses a WebAssembly extension model that is currently smaller. You should pick Zed if you need to handle large files without stuttering. JetBrains IntelliJ IDEA Ultimate costs $249 per user for the first year, and the All Products Pack costs $289 per user for the first year. Both packages have continuation discounts, bringing the second year to $199 and $231 respectively.