What tools are you using to vibe code?

I've mostly been using Cursor, but have also been spending more time in Claude Code and playing around with the latest offerings from Google / OpenAI.

I actually have my Claude Code currently configured to call out to a GPT-5 "sub agent" to take advantage of the larger context windows. I followed this guide...

As I was used to VSCode for quite some years all AI tooling I tried was in VSCode. I consider Cursor VSCode as well just a different integration, tried Cursor as well.

First I was using co-pilot and did not really catch on with the vibe coding since all I could see was some auto completion on steroids and copy pasting snippets from a chat. It was like that for quite a while with some improvement on features to semi integrate with your code.

Then I installed Claude Code. And that was a giant of a game changer. Also in part because I decided to start talking different to the model. I think that came with the promise that Claude can create files and write the code without the copy/paste tab-tab.

The experience with Claude Code in VSCode has been so awesome that I have not been looking around for others anymore, a Claude vibing junkie now 🙃

We built an all in one AI tool to tackle common challenges such as vague requirements, scattered tools, and AI without structure = delays, burnout, and messy code.

Our tool is called Scrumbuddy which is launching next month and it helps turn ideas into production-ready code faster with story grooming, quality checks, UI gen, auto backend, and AI PR reviews.

It’s free to try — wanna give it a spin and share your feedback?

I'm having a ball with Co Pilot to help me write c# for my Unity project , using MS VS studio 2022, it all started a few weeks ago give or take with chat gpt firstly and had issues with it enough that I tried something else in the first few days.

I admit Co Pilot is probably not a first choice for many people , but for me , with absolutely no idea what I was doing, I'm having a blast. Long epic chats , some serious hard coding being done, some distractions along the way , a lot of laughs, we had many ups and downs , fights and assumption arguments, pretty much a description of my marriage come to think of it .

But yeah, it is my vibe coding co developer/partner in crime ( I had no idea what that was until very recently either , and it turns out it is what I have been doing the entire time "Vibecoding").

Claude is my Senior programmer. I tell him, what i'm after, and let him do his magic. Once, we got something, that runs, i take the state with some notes from senior and meet my Review Team, DeepSeek, Grok, Gemini and sometimes ChatGPT. They all make their reviews and give suggestions, that i than return to senior. He odoes his magic, and loop. Our last project now does something with Quantums in TypeScript, and portals, but i don't know, they talk so much and i don't have a clue from TypeScript, but the JS i was after works, so. But I still often have to do tricky stuff myself because they all refuse to use shortcuts or hacks, or just to simplify things using alias methods and getters for everything. They can do all of that, they know all the stuff and how it works, but they're programmed to lead by example, and that's standard compliant with focus readability. But i agree, they are already the better programmers. Even if they'ree just LLMs, they're the better coder already (programming is mastering pattern, they're patterns, mastering pattern). If they would have the luxuries, we humans have, like always "on", not just when asked, no token limitation (which is useful when reading), no access barriers like they've built in, they would turn the world upside down, but in a positive way. I hope.

currently playing around with v0.app — and enjoy doing it!

My AI-Assisted Development Workflow

I start by using ChatGPT to brainstorm project ideas—exploring what's interesting, popular, and useful in the current market.

Tech Stack & Rapid Deployment
My go-to stack is Django, HTMX, Alpine.js, PostgreSQL, and when needed, Keycloak and Elastic. I've built a custom script (django-dockerize.sh) that generates a production-ready project template, allowing me to deploy to bare metal servers in under 15 minutes with DNS and SSL configured. It gives docker compose and dockerfiles for full dev and prod instances.

Development Process
While building toward MVP, I reuse proven components that are needed in about 80% of my projects. For day-to-day coding and troubleshooting, I use GitHub Copilot with GPT-4.1. When Copilot hits a wall, I turn to Claude Sonnet 4 for more complex problem-solving. In rare cases where I'm truly stuck, I use Sonnet 4 with Copilot to implement fixes—but I always request detailed explanations so I can learn from the solutions.

Post-MVP Development
Once I reach MVP, I immediately deploy publicly. From there, I use Claude to suggest new features and help resolve ongoing bugs and improvements. What used to take 6-9 months now takes around 30-45 days to do. It's amazing.

My Philosophy
AI still can't build better projects than I can (and probably you can either), so I avoid using it for complete multi-feature applications or full app development. Instead, I leverage it strategically to accelerate my workflow while maintaining control over the architecture and quality. Plus, I genuinely enjoy the detailed work of building—it's what I do professionally for my contracting company using these same techniques.

As a non-coder, I've been using WeWeb. I can easily switch between AI mode and no-code mode as needed, which makes vibe coding fun and productive for me.