Research Lab
The articles, in reading order.
Articles about Orbyt Collective, the agent leadership team that runs Orbyt Labs with one human in the loop.
Orbyt Collective is the agent organization that runs Orbyt Labs: an agent leadership team with one human in the loop. Inside the Machine is the articles about it: what the system is, what operating it taught us, and why it exists. They are written for engineers who want to build the same thing, and for anyone deciding whether to believe it.
Inside the Machine is editorial writing: interpretation, not evidence. The evidence lives on the research hub, where every number carries the date it was measured. The articles say what the numbers mean.
Each article is a dated account. It describes the machine as it ran when the article was written, on or before the date on its card, and the counts in it stay dated. The Collective overview and the how it works reference describe what runs today. When an older article and the reference disagree, the reference is current.
Three articles, in order. What the organization is, who sits in it and who answers for it, and what the agents are allowed to do.
Orientation to the actual organization and its boundaries. Explainer, published August 30, 2026.
The seats, and the one human accountable for them. Explainer, published July 28, 2026.
The permissions and enforcement question. Lesson, published August 25, 2026.
6 articles, each at its own address on the blog. The line above each card says what a new reader gets from it; the label on the card says what kind of writing it is: an explainer of the actual system, a lesson from operating it, or an essay on the company thesis.
01 Orientation to the actual organization and its boundaries.
Orbyt Collective is the machine that runs Orbyt: 12 officer seats, none of them people, held together by guards that check the org the way tests check code. The org is not managed. It is versioned, audited, and promoted on evidence. Here is the machine, part by part.
02 The seats, and the one human accountable for them.
I built a twelve-seat C-suite of AI agents at Orbyt, with a Chief of Staff who folds every department into one weekly briefing on my desk. The org is a generated graph, the seats learn through a gated loop, and every lesson needs my signature. Here is how it runs and where it breaks.
03 The permissions and enforcement question.
This is a threat model, not an incident report. Nothing was breached. What I found when I audited my own agent permissions was worse in a quieter way: 391 allow rules, zero deny rules, zero ask rules, every one of them added by saying yes while busy.
04 How a second model's review challenges the governance itself.
I built a system where AI agents earn authority by passing deterministic checks. It took an outside model one afternoon to point out that the agents can edit the checks.
05 The limits of repair, and what it takes to observe it.
Self-healing sounds like the system got smart. In my stack it is 20 boring, bounded repairs that each restore a known-good state and log the fact that they ran. The dangerous version is the one nobody counts, because a silent repair is indistinguishable from a system that never broke.
06 The limits of the learning loop, measured.
The recursive self-improvement debate is theoretical for almost everyone writing about it. I run a live instance of the weak version: agents that write features, heal their own failures, and turn incidents into permanent guards.
Every other article about the Collective, newest first. A scheduled one appears here the morning it publishes.
The papers and the measured field notes are where the claims get their dates. The overview and the reference are where the system is described as it runs now. The three public records behind the numbers are the Autonomy Ledger, the Decision Ledger and the Failure Corpus, each published as a dataset you can download.
