If you've been waiting to try local AI development, please try it

I have snobbishly long felt that the local models were not 'up to my standards' for local development, or otherwise able to compete with GHCP, Claude Code, Cursor etc. Boy was I wrong. With the rapid increase of usage constraints and enshittification of plans all the cloud providers are starting to put in place, I finally downloaded Opencode and got it setup with llama-server + Qwen3.6-27B at a reasonable quant with 128K context - unsure if I could push this more but it's plenty for the time be