Posted by Majnu Rahman on Jan 26th 26 at 6:20 PM

Hi @majnurahman100 This is a seriously ambitious project and I love seeing someone thinking this big. The internal debate between models is such a cool concept, like having a room full of experts arguing before giving you an answer.

Also really into the local angle here. Theres definitely a shift happening where developers want their AI agents running on their own machines instead of relying entirely on cloud services. Is that the direction youre going with this? Like keeping the orchestration and execution local for privacy and control, while still hitting the various APIs for the actual model calls? Or are you thinking even more local than that with some open source models mixed in?

I gotta ask though, this is a lot of moving pieces. Whats the first chunk youre planning to tackle? Like are you starting with the multi model orchestration layer, or maybe getting the permission based desktop execution locked down first since thats the security critical part?

Also curious what your thinking is on the debate mechanism. Are you imagining like a structured back and forth where each model critiques the others response, or more of a voting system where they all answer independently and then you pick the consensus?

Would love to hear more about how youre approaching the architecture on this one.