What is the Orbyt Autonomy Ledger?

The Orbyt Autonomy Ledger is Orbyt's published run record for the AI agent seats that operate the company: recorded runs, completions, failures and review-panel discards, with a frozen quarterly baseline and raw JSON at a stable URL. It is generated from the company repository rather than written by hand.

What counts as a completed run?

In the Orbyt Autonomy Ledger, a completed run is one whose output survived the workflow's completion gate. Failed runs died before landing. Expert-panel discards of finished drafts are counted separately from failures.

Why publish the failed runs?

Orbyt publishes failed agent runs alongside completed ones because a completion rate without its denominator is not a checkable claim. The ledger prints failures and discards at equal prominence and states its small sample size next to the headline number.

Can I cite this data?

Yes. The Orbyt Autonomy Ledger is published under CC BY 4.0 with attribution required, with raw JSON at a stable URL and a dated cite-as line. Orbyt asks that the completion rate always be quoted together with its denominator.

Research Lab

Autonomy Ledger.

The run record, published.

What the agent seats completed, what they failed, and what the panels killed.

One number, with its denominator.

Across 38 recorded runs since July 21, 2026, 32 completed and 6 failed. That is a completion rate of 84.2 percent, over a small number of runs, recorded by the org being measured. Both of those caveats belong next to the number, so here they are, next to the number.

A completed run is one whose output survived the workflow’s completion gate. A failed run died in the guard gauntlet or was cut off before it landed. And 24 finished drafts were killed by expert panels before anything shipped, with 22 proposed lessons held for human review. Work that finished and was then refused is a different fact from work that never finished, so the two are never added together.

Recorded runs

84.2%

Completed

Failed, printed anyway

Drafts killed in review

Every seat, on the record.

The rows are appended by the org’s own workflows when a run ends, not written by hand, and this table is aggregated from them on every build. Seats keep their working names.

Seat	Runs	Completed	Failed	Discards	Last run
chief-of-staff	4	3	1	3	2026-08-10
engineering	5	4	1	4	2026-08-13
innovation	2	2	0	2	2026-08-13
legal	6	5	1	4	2026-08-13
marketing	6	4	2	5	2026-08-08
operations	2	2	0	0	2026-08-02
product	6	5	1	4	2026-08-13
quality	5	5	0	2	2026-08-13
revenue	2	2	0	0	2026-08-02

The baseline is frozen.

The opening baseline was frozen at first publication, 2026 Q3, and is never regenerated. Every later edition of this page is measured against that origin, which is what makes the series a series instead of a number that quietly rewrites its own history.

The raw data, baseline included, is one file at a stable URL: autonomy-ledger.json. CC BY 4.0, attribution required, and if you quote the completion rate, quote the denominator with it. The artifact carries its own citation line, printed here verbatim from the same file:

Orbyt Autonomy Ledger (2026-08-13), https://www.orbytjobs.ai/orbyt-collective/autonomy-ledger

Common questions.

The run record of the agent seats that operate this company: every recorded run, what completed, what failed, and what the review panels killed. One locked metric, a frozen baseline, and the raw JSON at a stable URL. Generated out of the repository on every build.

A run whose output survived the workflow's completion gate: the sentinel check accepted it and the work landed. A failed run died in the gauntlet or was cut off. Panel discards are counted separately, because work that finished and was then killed in review is a different fact from work that never finished.

Because a completion rate without its denominator is an advertisement. The failures and the discards are printed at the same prominence as the successes, and the sample is small, which is said next to the number rather than in a footnote.

Yes. The ledger is CC BY 4.0 with attribution, the raw JSON sits at a stable URL, and every edition carries a cite-as line with its date. If you quote the completion rate, quote the denominator with it.