Like many image and video AI tools, which have (mostly) stopped creating people with six fingers, AI coding tools have also been making great strides. Case in point: developer Indragie Karunaratne just shipped Context, a native macOS app that was 95% built by Anthropic’s Claude Code.
Anthropic has been standing out in AI-assisted development
For the better part of the last year, Anthropic has pulled away from the pack when it comes to how good its Claude models are at generating code (to be fair, the competition has managed to close the gap more recently).
Still, Anthropic hasn’t been sitting still. While Claude is already available in popular tools like Anysphere’s Cursor, the company also recently launched Claude Code, its own coding environment.
Rather than quietly suggesting snippets inside your IDE, Claude Code acts as a full-on development agent capable of planning, writing code, running tests, and iterating based on a stated goal.
There are limitations, or course. But as Karunaratne puts it:
“Let me just start by saying that Claude Code (with the latest Sonnet 4 and Opus 4 models) is genuinely good at writing code. It’s certainly not a top 1% programmer, but I would say that Claude’s outputs are significantly better than those of the average developer. (…) The really incredible thing is that it does this in a fraction of the time that it would take a person to implement the whole thing. Imagine onboarding a new employee with zero context on your project and having them ship a complete feature a few minutes later.“
Building Context
Context is a very specific tool for a very specific job. It’s for testing and debugging MCP servers, which enable AI agents to interact with platforms and tools traditionally used by humans.
And while its project has more than 20,000 lines of code, Karunaratne claims that around 1,000 lines were actually written by him, leaving the rest up to the AI.
In an interesting and very detailed blog post, Karunaratne outlined how Claude Code agentically handled nearly every part of the process, from writing in Swift and SwiftUI, to running builds, iterating on compiler errors, and even generating release automation scripts.
Obviously, it didn’t work out of the box with zero effort. As Karunaratne says, “Claude Code is okay at Swift and good at SwiftUI”.
But the broader point is this: while we’re still a ways away from having a Fortune 500 company spun up exclusively with AI-written code, for this particular project, the heavy lifting was offloaded in a way that would’ve been unthinkable just a year ago, and that’s pretty exciting.
Be sure to head over to Karunaratne’s comprehensive blog post and read the whole piece, where he also explores how the traditional code editor may soon be obsolete.
Are you a developer? Have you been incorporating AI into your workflow? Let us know in the comments.
iPad deals on Amazon