AI work should feel like a room, not a reply.
Oryx is a shared workspace for people who use AI to think in parallel. Start from one project, branch the useful threads, compare several models, and keep the human decision trail close.
Nemotron reads for risk, Gemini scans tradeoffs, Qwen structures the plan, DeepSeek challenges the assumptions.
4 model replies aligned into one project threadOryx treats AI as a workshop for second opinions, not a single autocomplete box.
The repo is built around projects called deep dives: shared spaces with model picks, threads, votes, debate mode, invites, roles, and live team notes.
A durable place for messy, multi-model thinking.
Projects hold the memory
Prompts, files, threads, model choices, and team notes live in the same workspace so context does not smear across tools.
Models become collaborators
Run an idea through different model personalities, compare the shape of their answers, and ask another model from any message.
Threads branch cleanly
A promising objection, option, or tangent can become its own thread while staying connected to the project that created it.
Teams share judgment
Invite collaborators with editor, commenter, and viewer roles, then keep human debate beside AI debate.
From first question to defended decision.
Oryx keeps the shape of the work visible: which models were asked, where the thread branched, what was voted on, and what the team decided to remember.
- 01Start from a real project question
- 02Select the models worth hearing
- 03Branch, debate, or vote when the thread changes shape
- 04Keep the human notes, replies, and decisions attached