Limited simulation of Orbyt Labs Collective Dashboard.
The Machine at work.
Everything in this small Orbyt Labs dashboard is a simulation, so you can see how the Machine behaves in some of its core tasks, not how it has performed. The four chapters after it are read from the real records.
Use the switch below. FYI: you can change some of the settings, like the controls at the bottom of the Swarm and Autonomy cards, and see how they affect the Machine.
Learning
Learning Loop
Agents
Company Agent Team
ActiveControl
Swarm
ActiveAutonomy
ActiveSignal
Tokens
ActiveTop 7 of 12 seats in the window
Alignment
Active3,456 held up144 did not12 waiting
1 · The team
See who is on it.
12 seats (one agent each, with a written charter), each with a role, a schedule and a standing anyone can check. The dot is the seat's latest standing on the public seat board.
A seat is a standing job for one AI agent. Its charter says what it may do alone and what needs the founder's sign-off, and every run it makes lands on the record. Standing is read from those committed runs by a script, not written by the seat about itself, and a seat that is not healthy stays on the board with its reason. Right now 10 are healthy. Of the other 2, 1 needs fix and 1 not observed.
- Chief Marketing OfficerAgent Cite · Healthy
- Chief Financial OfficerAgent Ledger · Healthy
- Chief Technology OfficerAgent Ward · Healthy
- Chief Product OfficerAgent Compass · Healthy
- Chief Innovation OfficerAgent Cusp · Healthy
- Chief Operations OfficerAgent Keeper · Needs fix
- Chief Revenue OfficerAgent Convert · Healthy
- General CounselAgent Bastion · Healthy
- Chief Quality OfficerAgent Assay · Healthy
- Chief Security OfficerAgent Sentinel · Healthy
- Inspector GeneralAgent Gauge · Not observed
- Chief of StaffAgent Fulcrum · Healthy
From the seat board, generated .
2 · What happened
Know what happened today.
Every day the pulse, a daily read of the whole organization, checks 9 things and says what it saw, failures included. Each seat's runs land in the autonomy ledger, a public record of seat runs.
The pulse is written by a script with no AI in it. It gives a verdict for each thing it watches, and says so when it could not look, so a gap shows as a gap and not as a clean day. Each of the 9 things it watches is listed below with its latest verdict. The autonomy ledger counts what each seat ran, completed and failed.
Latest pulse, in its own words
2 run-truth finding(s): something reported success without producing an artifact.
- Open founder directivesObserved clean
- Did the org produce anythingObserved red
- Monitors nobody fedObserved red
- Red runs in the last dayObserved red
- Crons running and failingObserved warning
- API command attributionObserved warning
- Halts nobody commandedObserved clean
- Does the record still reproduceObserved clean
- Open pull requests sitting redObserved red
Pulse as of . Ledger runs from to .
3 · The human call
Some calls stay human.
Each seat's charter lists what it may do alone and what needs the founder's sign-off. Money, public claims and anything irreversible stay with the founder, and the founder can pause any seat.
The split is written down seat by seat: across 12 seats, 30 kinds of work a seat may do alone and 40 that need the founder. Work in the second group waits for his sign-off, and the weekly briefing counts what is waiting.
The organization also has one kill switch, a single file called HALT that every seat checks before it lands any work. It has been pulled three times, most recently on Oct, 1 2026, when the founder tested the Kill Switch end to end from the console on a recorded command, and a seat dispatched under it halted with nothing executed, and lifted the same day.
From the authority table and the weekly briefing of .
4 · The evidence
Check the record yourself.
Every decision on the record is numbered and dated. Every failure we find is written down, and most get a check written to stop them coming back. Both records are public files anyone can download and cite.
The decision ledger holds 99 structural decisions. A correction is a new entry that names what it replaces, so the history is never rewritten. The failure corpus documents 97 failures and the check written for 73 of them. Both files are free to cite under CC BY 4.0, as JSON and as a spreadsheet.
Latest entries, as the ledger titles them
- Workspaces are archived, then deleted, automatically when their work is on main or after seven days idleD-0099 ·
- The revenue-plan question waits until January 2027D-0098 ·
- A seat killed by its time limit gets the envelope reason timeout after v17D-0097 ·
Decision ledger generated . Failure corpus generated .
Words the Collective uses in its own sense, such as seat, pulse and ledger, are defined in the glossary.