Posted herein, as I “learned the concept” within this forum… actually, no, I do not believe one truly grasps it until executing it personally.. how extraordinary that is.. my inaugural vibe-coded application, a grammar and style checker and more broadly a text assistant Firefox extension that operates using your own local LLM model via Ollama.
Locate it here: GitHub - sylvaticus/locaispell · GitHub
(The quality depends on your local models of course, this is with qwen3.5:9b)
Congrats!
I know you told me before that you dislike JetBrains, but I truly recommend giving it a go for this. Flexible Julia now has a full integration that opens up an MCP endpoint to your tooling for Ollama (and other agents). It is easy to configure and then you can actually iterate with the AI agent, have it see the errors & warnings that the IDE or language server produces, react and fix them.
Regardless: good job on a first step!
Thank you 
Just one correction, I never said I dislike JetBrains, just I had no reason to switch from VSCode, perhaps this is a good one 
Cheers…
Cool. Let me know how it works out and keep posting your progress
LAITA is now feature complete: a vibe-coded multi-app for Firefox, Chrome, VSCode/VSCodium, LibreOffice/Collabora online, and any software (Collabora online) that interface with the LanguageTool and DeepL APIs .
Provides as-you-type or whole-document-based text proofreading and generic text transformation. Needs a decent GPU.
My feedback from my first vibe-coded app: thanks to AI (“vibe code”) developing a program (and, hopefully, scientific computational models) has become much simpler, quicker… and different. I feel a bit like driving a car with a GPS rather than using the old-style road atlas: yes, I reach my destination… but I feel lost, I don’t know the path. If I have to “drive” again, I will need the GPS (AI assistant) again.
