To further develop my AI skills, I decided on the following weekend project:

• use an AI model that runs locally on my computer
• build a coding assistant for simple front-end code (html, css, javascript)
• bonus: add voice input

Only a few years ago it would have been impossible for me to achieve the above. Now, I indeed could do this over the weekend!

Video of my AI coding assistant with voice input

Of course, current commercial tools such as Claude Code, Cursor and Open AI Codex are way more sophisticated than what I produced on my own over the weekend. But that's not the point.

Free technology and open standards provide individuals with capabilities that can be pretty amazing, while maximizing privacy and minimizing environmental impact.

Some further technical details about what I used to build my prototype:

• local AI using Ollama with the mistral:latest open model.
• my own server-side PHP framework for the database and rendering.
• the Web Speech API with on-device speech recognition (I basically copied the code from this page, and it worked!).
• Prism javascript library for syntax highlighting
• my old Mac Mini from 2018 (with 16GB memory, a 3,2GHz 6-Core Intel Core i7 processor, and Intel UHD Graphics 630), working hard to produce AI responses in about 1 minute...