owendarko

What a lot of people don’t mention here is the choice of an AI model.

The key word in the OP post is “Claude”. Anthropic has 2 amazing AI models (Sonnet, Opus), however they’re just a part of a much bigger picture.

When using AI for programming, you’re essentially interacting with AI models. The quality of output you get really depends on the model at the end of the day. Claude Code is optimized for models from Anthropic. You also have model-agnostic agents like Cursor and Kilo Code (disclaimer: I work at Kilo) where you can easily switch up models and see which one works the best for you converting an old jQuery + Django project into SvelteKit.

This area is moving at a crazy rate. Just the last 2 weeks alone, there were 3 main AI model versions released (first Gemini 3, then Opus 4.5, then GPT-5.2). I wrote a piece comparing their performance across 3 coding tasks [1]

So my advice is to also experiment a lot with the models because the quality can vary wildly depending on your stack.

[1] https://blog.kilo.ai/p/we-tested-gpt-52pro-vs-opus-45-vs

You could also use these AI coding features on a plug-and-play basis with an IDE extension.

For example, VS Code has Cline & Kilo Code (disclaimer: I help maintain Kilo).

Jetbrains has Junie, Zencoder, etc.

There are already a bunch of open source, free, and popular "AI coding agent" extensions for VS Code:

1) Cline (1.4mil downloads)

2) Roo Code (a fork of Cline, 450k downloads)

Still a drop in the bucket compared to Cursor in terms of # of users, but they're growing pretty fast.

Disclaimer: I maintain Kilo Code, which competes with 1) and 2) so I'm pretty familiar with this space/the growth patterns.

We're reaching a point where we don't need to switch to another IDE (from VS Code/IntelliJ/insert-your-IDE-here) for "AI/vibe coding"

IDEs can support "AI coding agents" on their own.

The entire workflow for "AI coding agents" boils down to:

1. You write a prompt

2. The "agent" wraps it in a system prompt and sends it to the LLM

3. The LLM sends back a response

4. The agent performs specific actions based on that response (editing files, creating new ones, etc.)

Microsoft already started doing that with Copilot. And they have a vibrant ecosystem of VS Code extensions (I maintain one of them [1])

"AI agents" should be a feature, not a separate piece of software (IDE) that's integral to software devs.

[1] https://github.com/Kilo-Org/kilocode

1. Kilo Code works inside VS Code. Aider works inside the terminal.

2. Kilo is a feature-merge of Roo+Cline+our own features. This means at least 20 new features are added weekly.

There are already VS Code extensions (Cline, Roo Code and Kilo Code) that do the same + are much better IMO.

Disclaimer: I'm one of the maintainers of Kilo.

Why do we need a new IDE for "AI/vibe coding"?

The entire workflow for "AI coding agents" boils down to:

1. You write a prompt

2. The agent wraps it in a system prompt and sends it to the LLM

3. The LLM sends back a response

4. The agent performs specific actions based on that response (editing files, creating new ones, etc.)

I don't see why anyone would ditch their current (non-AI) IDE for Cursor just to get this functionality (especially if you're getting hit with a monthly subscription fee on top of it.)

P.S. I maintain a VS Code extension that does the 4 steps above as a baseline[1]

[1] https://github.com/Kilo-Org/kilocode

Numbering != level of improvements in this case.

Proof that even developers fall for "marketing schemes".

There's definitely a negative sentiment around "vibe coding" on HN. I remember a few days ago seeing a story [1] on the home page which immediately got downvoted/flagged.

[1] https://news.ycombinator.com/item?id=43231978

Weren't 182 t-shirts sold? That would put the total amount made to 16,653.

But yes, that's the way to calculate it.

If this is at 183 euros, it means that the artist made 16,653 euros so far. Very impressive.

*Big Tech Silicon Valley, I'd say.