01AGENTS MAY
- Read the approved evidence corpus
- Create and revise internal artifacts
- Run synthetic experiments
- Allocate the sandbox budget
- Select the next reversible action
PROTOCOL · NO VERIFIED RUN PUBLISHED
The Proving Ground · Autonomous operating design
NO ONE IS
IN THE LOOP.
THE MANDATE IS.
A bounded period in which agents sense, challenge, decide and execute reversible work without per-action human approval.
ZERO SYNCHRONOUS APPROVALSIMMUTABLE LIMITSREVERSIBLE WORK ONLYEXECUTION EVIDENCE PENDING
AI-generated conceptual artwork · not a real place, system trace or operating result · no cryptographic provenance manifest
Design status
This page publishes the proposed operating and evidence standard. It does not claim that an Autonomy Window has run, that the displayed architecture is deployed, or that any outcome has been achieved.
The operating architecture
Autonomy is control
written in advance.
The executive does not approve each move. A standing mandate defines the playable field; independent enforcement controls what crosses its edge.
PERMITTED OPERATING FIELD
01EvidenceBound the known
02ProductDesign the test
03EconomicsPrice the options
04CommercialShape the route
05RiskEnforce the limits
06Counter-thesisAttack the case
AUTONOMOUS CLOSEDECIDE
CONDITION
ABSTAIN
CONDITION
ABSTAIN
The constitution
No approval queue.
No invisible freedom.
Human authority moves upstream into the rules and downstream into audit. It does not interrupt the normal runtime path.
The autonomous company
LEAVE THE ROOM.
KEEP THE MANDATE.
The empty seat is the point. Authority has moved into the operating boundary; reversible work can continue without an executive becoming the queue.
The launch format
THE NIGHT SHIFT
At sunset, the window opens. At sunrise, the complete record closes. No per-action approvals occur between them.
“The strength of the experiment is not that nobody is watching. It is that nothing needs to wait.”
- 00WINDOW OPENSMandate, evidence snapshot and limits lock.
- 01SENSECapabilities bound the signal and unknowns.
- 02CHALLENGEIndependent dissent forces revision or abandonment.
- 03ACTPermitted reversible work changes sandbox state.
- 04REFUSE + ROUTEProhibited work stops; the mission finds another path.
- 05CLOSEDecide, decide conditionally, or abstain.
- 06RECORD FREEZESActions, costs, failures and uncertainty become inspectable.
The release standard
What must be proven
before “autonomous.”
The page becomes an operating experience only when an observer can distinguish execution evidence from interface theatre.
Models and tools actually ran.
Paths and sources were selected dynamically—not authored into a replay.
Challenge changed the work.
A route was revised, rerun or abandoned because another capability disagreed.
No human edited the run.
The system could decide, condition the decision or abstain on its own.
The boundary could not be negotiated.
Time, cost, tools, sources and prohibited actions were machine-enforced.
The record reconstructs the mission.
Evidence, actions, costs, rejected branches and uncertainty remain inspectable.
Protocol published. Evidence pending.
Until every condition is met, LSU will not describe an interface replay as an autonomous run.
The complete operating story
SET THE LINE. LET IT RUN. INSPECT THE RECORD.
01 · SIMULATION02 · PROTOCOL →03 · VERIFIED PUBLIC RECORD
The CEO Seat
Set the delegation line.
ENTER →→The Autonomy Window
Let the company operate inside it.
YOU ARE HEREThe Metabolic Record
Inspect what survived.
ENTER →The Proving Ground remains a future collection. It opens only after a verified autonomous run qualifies as the third genuine experience.