Dotta is the creator of Paperclip, an open-source human control plane for AI labor that coordinates agents while keeping people responsible for goals, spending, approvals, and quality.
Paperclip organizes agents into executive, engineering, marketing, and specialist roles, with shared context, reusable skills, delegated tasks, and explicit human oversight. Dotta uses it to develop Paperclip itself and coordinate marketing production, including an agent-generated promotional video informed by project data, brand guidelines, and his feedback. His Paperclip demonstration shows how agents from different providers can collaborate without being locked into one model or execution environment.
- Completion as an evidence-backed claim. Finished work requires a defined artifact, verification standard, supporting evidence, authorized approval, residual-risk ownership, and a next action. Ready to merge, deploy, and announce represent different thresholds.
- Workflow liveness with accountable oversight. Explicit task-state transitions, dependencies, designated reviewers, recorded approvals, and watchdog agents keep work moving without ignoring genuine blockers or overwhelming human reviewers.
- Independent cross-model verification. The agent producing work should be separate from the agent checking it, potentially using a different model and practical verification tools such as browser access, screenshots, and automated tests. His explanation of Paperclip’s liveness model connects verification to clear chains of responsibility.
- Incremental organizational learning. Agent organizations should start small and improve through observed mistakes, revised instructions, reusable skills, shared brand standards, spending limits, and human approval points.
Dotta applies this model beyond software development to marketing, sales, finance, and other business operations, preserving human authority over consequential decisions.
