ADR-0011 — The effort economy: founder hours, staff lanes, manager compression

Status: ACCEPTED (shape); researcher archetype roster content owed (DQ-15) · Decided: 2026-07-12
These are the actual decision records the game is built from, published unedited except for internal process notes. They describe why the game works the way it does. They are not a strategy guide, and some of them argue with each other.

Context

Current build: one global AP pool, spent instantly; staff add to the pool; researchers are single-function; banking exists but does nothing. The sweep's diagnosis in mechanism terms: one fungible currency + state-independent payoffs ⇒ constant-policy argmax (safety_lean) dominates. Pip's rulings across beats 1–2: banking is dead (ADR-0009); "I have never been around unassigned workers before, because there's always been a backlog"; "each new employee is kinda cutting down the time I need to do other things"; managers "stop employees from annoying me with employee things."

Decision

  1. No global AP pool. The pool illusion dies; staff never add founder AP.
  2. Founder hours (sacred, roughly fixed per month — canon since ADR-0008): spent on doors (stakeholder face-time; which rooms you're in locks strategic paths), approvals (hires, direction rulings — "approve this salary," not paperwork clicks), audits (skip-level ground-truthing), and reserve (instant-speed firefighting, ADR-0009).
  3. Staff effort is per-person, assigned at plan speed to workstreams from a backlog. Idle staff don't exist; unmanaged staff do — unassigned researchers self-direct per their agenda (whose ordering of the backlog wins?).
  4. Workstreams run multi-month and produce artifacts (papers, systems, campaigns) — ADR-0009's no-instant-strategy rule, cashed out.
  5. Managers are interrupt shields with agenda riders: they absorb their team's class of response windows (ADR-0009 economy), compress the plan screen (team → allocation a Celine's-law channel — an inward SA channel with fidelity loss. The sim never lies; characters do. Skip-level audits (founder hours) re-ground truth. Folds ADR-0008's deferred inward-SA into existing SA machinery.
    • standing policy; members visually housed under the manager), and report upward via
  6. Ops/admin staff reduce the founder-price of routine actions and automate whole classes over time (endpoint: 2037 vignette's "payroll is automated").
  7. The dual-use lane is priced: capabilities work pays money/compute/hype and bills doom (with ADR-0010's typed attention: accel VCs fund reckless success).
  8. Researcher model (shape): lane preference (fictionalized real agendas — e.g. interpretability, evals, theory/agent-foundations, governance) × appetites (compute, prestige/first-author, mentees, money, mission purity — promises made to retain staff are ledger entries) × rare quirk riders (philosophical stances, secret successionist — an exposure-event genre reusing ADR-0003 machinery, not a lane).

Beacons served / violated

Interaction contract

Reads/writes: response windows (ADR-0009 — manager value denominated there), SA (inward channel, fidelity, audits), Liability Ledger (agenda riders, retention promises, dual-use bills), reputation (per-person, ADR-0010 — hires are reputation seeds who recruit juniors), replay/exploit-finder (assignment policies join the policy space).

Rejected alternatives

Consequences / open questions