Start Already have an account? Log in

Papers.

Formal research on the AI agents that run Orbyt Labs.

An Orbyt Labs paper is formal research on the AI agents that run Orbyt Labs. Each paper states its method and the date it was measured. Once its deposit is issued, it carries a DOI.

Every paper has its own repository on GitHub.

Paper
Agent-Native Dataset Design
Updated Sep 18, 2026
Schema, licensing, and distribution patterns for datasets that are read by LLM agents rather than crawled by search engines. Proposes agent-native dataset design as a category distinct from training corpora and traditional research datasets.
DOI 10.5281/zenodo.19754393
Author: Justin Bartak
Read it
Paper
Governing an Agent Leadership Team
Updated Sep 18, 2026
Guardrail, kill-switch, and liveness patterns from a company run by twelve agent seats under one human, each seat one agent with a written charter. Three guardrail tiers, a single-file kill switch that stops the whole org, positive-observation liveness, and a published corpus of 66 documented failures, half of them failures of the instruments.
DOI 10.5281/zenodo.22683647
Author: Justin Bartak
Read it

Explore

Research Field Notes Orbyt Collective