Your Day as an Operator
Your day as an operator
The daily loop is not watching agents work. It is governing by exception. Four surfaces, a short queue, and most of the routine already done before you arrived.
What the loop looks like
Section titled “What the loop looks like”You open GLoops. The first thing you see is the Action Queue — the few decisions the system genuinely could not make without you. Each item leads with the decision in one sentence, a recommended option pre-selected, and the evidence one click away. You approve, edit, or reject. Everything else — the routine, the bounded, the receipted — already advanced without you.
The queue is short by design. Items appear only when a Routine Operational Mutation gate cannot hold, when authority widening is proposed, or when a production-impacting decision needs a human record. If you are being asked the same question repeatedly, that is a system defect, not normal operation.
The four surfaces, in order
Section titled “The four surfaces, in order”1. The Action Queue — reserved decisions only. Each card states: what is needed, why it matters, the recommended option, the consequence of inaction, and an expiry. You decide once; the decision routes to execution and leaves a record. The empty state — “Nothing needs you right now” — is the goal state, not a failure.
2. The Capability Portfolio — changed burden, not activity. You are not watching agents work. You are scanning for changes in what the organization can do reliably without you. A capability moving from dry-run to operational is news. A capability stalling is news. Everything else is background. See the Product Surfaces catalog for the full surface specification.
3. Receipts — proof of what happened. When you want to know what occurred overnight, you open the Receipt Explorer. Each receipt is immutable: who decided, the original and final action, the reasoning summary, and the frozen evidence. Nothing is mutable; nothing is asserted without proof. You inspect when you choose, not because you were summoned.
4. The Role Replacement scoreboard — what left your plate. The single number that matters is the SDLC Orchestrator role’s replacement percentage. Did it move? What evidence moved it? What responsibilities remain? The scoreboard answers all three without a report, a meeting, or a prompt.
What a good day looks like
Section titled “What a good day looks like”The queue is short. The receipts are clean. A responsibility moved off your plate and a receipt proves it. You spent your attention on judgment — approving a promotion, deciding a priority, resolving a constitutional question — not on mechanism. You did not open GitHub directly. You did not manually check agent status. You did not re-explain the architecture.
That is the design target. You govern the organization; the organization operates itself.
▣ Current Reality
Section titled “▣ Current Reality”Most of the above is target-state. Today, the Action Queue is the one decision surface actively taking shape — it exists, it persists decisions, and the Burden Intelligence capability projects into it. The Receipt Explorer, Capability Portfolio, and Role Replacement scoreboard are designed and partially built, but execution is dry-run: the system has zero production operational receipts, and most routine steps still require the executive.
For the precise, dated grade of each surface, see the Honest Current State page. For the full surface catalog and implementation targets, see Product Surfaces.
The posture that applies now: use the Action Queue as the one governed decision surface. Treat everything else as preview and implementation target. The day described above is where the system is being built to arrive.