Why the future of software development may be built on forks of VS Code, and what we discovered building our own AI coding assistant.
AI is rapidly transforming the developer experience. Over the past year, we've seen a surge in AI-powered IDEs (Integrated Development Environments) - intelligent coding tools designed to boost developer productivity, reduce bugs, and accelerate software delivery.
In this piece, we compare some of the top players at the moment:
And we share insights from our own experiment: building a custom AI plugin inside VS Code to support developers during a recent New Icon hackathon.
Traditional IDEs are now being supercharged with AI developer tools. Think code completion, multi-file refactoring, natural language queries, automated documentation, and context-aware debugging.
For software teams building with modern stacks (React, Python, Node.js, etc.), these tools are reshaping workflows, moving from "autocomplete" to true AI-powered coding assistants that help engineers think, build, and ship faster.
Cursor is one of the most advanced AI code editors on the market. Built as a fork of VS Code, it integrates GPT-4 and Claude to offer:
Windsurf focuses on becoming an AI teammate that “learns” your entire codebase. It offers:
It’s early days, but for large monorepos and complex apps, Windsurf’s potential is huge.
Kiro is Amazon’s recent entry into the AI IDE market. Also based on a VS Code fork, it’s built for:
The UI feels less flashy than Cursor, but this one is clearly built for large teams and scalable infrastructure.
Still the default setup for many, GitHub Copilot remains a strong choice for autocomplete, suggestions, and boilerplate generation. It now integrates with Copilot Chat and is pushing toward more agent-like functionality. However, newer tools are beginning to surpass it in UX, repo comprehension, and customisation.
Every single one of these AI IDEs is built on a fork of Microsoft’s open-source VS Code editor. This has created a fascinating dynamic:
At New Icon, we’ve been exploring these same possibilities hands-on.
In a recent internal hackathon, our team built an AI-powered file search tool - fAIo - designed to solve the all-too-familiar pain of chaotic file systems. What’s notable? We built it as a lightweight, AI-enhanced layer on top of our dev tools, using OpenAI and Node.js, without needing a full IDE fork.
It was a rapid prototype, but it proved something powerful… you don’t need to fork VS Code to get real value from AI in developer workflows, you just need to target the right problem.
While the major players are racing to own the AI coding experience from the top down, fAIo showed what’s possible from the bottom up: focused, intelligent tooling that plugs into your stack and solves real-world issues fast.
We’re entering a new phase of software engineering where AI agents embedded in developer workflows will become as common as linters or CI/CD pipelines.
The real question isn’t “Which tool autocompletes fastest?”
It’s “Which one truly understands what I’m trying to build?”
Instead of wrestling with generic IDEs or one-size-fits-all platforms, we help teams prototype smartly with purpose-built AI tools that deliver ROI fast. Book a discovery session with our team.
Subscribe to get our best content. No spam, ever. Unsubscribe at any time.
Send us a message for more information about how we can help you