dotta 📎

The Paperclip thesis in a nutshell: Enterprises need to ensure the orchestration is decoupled from any single model and build private evals that define what "good" means for their domain and these loops compound value for your firm

One problem with multiplayer AI is that you want to share agents, but keep individual responsibility E.g. we share a coding agent but each get our own credit for the git commits So we've added user-specific secrets to Paperclip Shared Agents & Skills, individual attribution