The orchestrator's earnings

The seventh use of §19's path, and the one Baruchi named: why should an inspector profit from something a troop did? from their perspective it's just like any other actor in the system. Run 8's finding (above) was that the panel rule is the honesty gain and Ω's backing the capacity gain, and that the clock's liability is a stock the orchestrator cannot pay down — the troops' net at 60/40, 6,200–6,800 a run, went to the market's reviewers as uplift by 1.4.0's clause that nothing a troop earns reaches Ω, while the same Ω carried the rebates on about 2 a tick and ended near zero at 50–60% with liabilities of 20,000–26,000 and plants unfunded.

Source: HANDOFF.md l.1384–1389

The money was in the wrong pocket: the troops' net at 60/40 was 6,200–6,800 a run (the fund's uplift paid plus its closing balance less the surcharge received), and 1.4.0 §2.4 sent it to the market's reviewers as uplift by the clause that nothing a troop earns reaches $\Omega$, while the same $\Omega$ carried the rebates for the lateness those troops were summoned to prevent. The two-thirds cap at 60/40 halved the queue (94 → 44), lifted on-time from 0.60 to 0.90, cut mean lateness 38 → 8 and the liability 19,778 → 3,512 with false approvals unchanged and careful wealth within 3% — and, a figure the run-8 notes had not reported, tripled plants unfunded (81 → 272), because the state delivered more from the same $\Omega$.

Source: spec/agentic_economics_loop_spec_v2_9_0.md l.1272–1272

A troop is an actor of the state; what it earns is the state's, and the state's bills are paid from it. (1.7.0)

Source: spec/agentic_economics_loop_spec_v2_9_0.md l.32–32

The rules (§4.1, §10.4, §12.1, §12.2, §12.3, I23). Two accounts again: the duty fund is 1.3.0's uplift fund — the raised tax in and nothing else, its whole balance out as uplift at every batch, no reserve, no lock; the troops' wallet is a separate troop account — Ω's advances and every troop's earnings in; costs, locks, forfeits, burns, repayments out; at each batch the unlocked balance above R_T flows to Ω. The clause withdrawn: a troop is an actor of the state and its earnings are the state's; the guards that stand in its place (members first at every seat, the windows, cost at summons, no emission/uplift/ρ/κ to a troop, the state's share public) plus one reading, Ω's income by source. R_T is capital held: the two-period peak of costs purchased plus locks, every troop in the field; the 1.5.0 refused component withdrawn (its reason went with A_T); its purpose now the troops' in-flight capacity shielded from the liability queue. Ω pays its debts from its income: the 1.6.0 order unchanged and funded; the plant reserve's horizon ruled a plant's life. The cap ⌈2E_t/3⌉. ω₀ swept {5, 10, 15}% with r_s pinned at 1.

Source: HANDOFF.md l.1399–1408

520 runs, 20 seeds per cell, 20 actors, every final audit passed, engine 003a238.

Source: HANDOFF.md l.1441–1441

cell	plants unfunded	Ω's liability	Ω at the end (min, last period)	on-time share (2nd period on)	queue	reviews unassigned	troop deliveries	award summonses refused (case-ticks)	troop surplus to Ω	careful wealth (uplift)	FA (probationers')
1.6.0 control (merged fund, half cap)	81	19,778	94 (48)	0.47	94	7,746	1,353	13,049	—	4,956 (1,291)	13 (3.7)
accounts alone (half cap)	124	16,694	347 (314)	0.28	718	3,534	560	33,638	3,575	4,239 (528)	13 (3.7)
cap alone (merged fund)	272	3,512	47 (28)	0.83	44	3,424	1,723	1,979	—	4,808 (1,367)	13 (3.4)
both (1.7.0)	0.7	317	4,352 (4,010)	0.89	42	3,449	1,719	607	6,029	4,250 (795)	14 (3.1)

Source: agentic_sim/results/studies/1.7.0-earnings-notes-for-reviewer.md l.11–16

Both together do what Hypothesis 1 says: plants unfunded zero in 17 of 20 seeds (2, 4 and 8 in three; one seed tipped), the liability a few hundred, Ω above 4,000 through the last period. The accounts alone are worse than the control: under the merged fund the troops were financed by the surcharge whatever Ω owed; under the two accounts every summons beyond the capital held is Ω's advance, and Ω pays debts older than a period first — once the half cap lets lateness accrue, the debts take Ω's income, the state cannot summon, and the next period is later still. R_T as capital held (461) shields the in-flight cohort and nothing more; the cap is what keeps the loop from closing.

Source: HANDOFF.md l.1449–1455

stampers	plants unfunded	Ω's liability	Ω at the end	on-time (2+)	lateness	queue	troop deliveries (refused case-ticks)	troop surplus / fee income	careful wealth (uplift)	stampers	FA	lag kind
50%	138 (159)	20,643 (18,675)	59 (56)	0.25 (0.42)	90 (30)	609 (129)	500 (23,497) vs 1,169 (8,377)	2,743 / 2,418	4,056 (484) vs 4,432 (1,006)	611 (681)	11 (13)	0
60%	0.7 (272)	317 (3,512)	4,352 (47)	0.89 (0.83)	6.8 (8.3)	42 (44)	1,719 (607) vs 1,723 (1,979)	6,029 / 2,993	4,250 (795) vs 4,808 (1,367)	651 (715)	14 (13)	0.05
70%	0.1 (19)	0 (776)	7,967 (856)	0.97 (0.95)	4.0 (5.2)	41 (41)	1,859 (95) vs 1,861 (648)	7,191 / 2,996	4,659 (1,054) vs 5,653 (2,038)	736 (804)	14 (14)	0
80%	3.6 (1.0)	226 (1,409)	9,215 (1,135)	0.95 (0.93)	5.0 (7.0)	40 (40)	2,132 (1,642) vs 2,075 (1,714)	8,528 / 2,996	5,190 (1,539) vs 6,881 (3,231)	810 (955)	16 (17)	0

Source: agentic_sim/results/studies/1.7.0-earnings-notes-for-reviewer.md l.22–27

Tipping (1.7.0 against the 1.6.0 two-thirds control): at 60–80% solvent and on time — liability 317 / 0 / 226, Ω 4,352 / 7,967 / 9,215, plants unfunded 0.7 / 0.1 / 3.6, on-time 0.89 / 0.97 / 0.95 — with the state's share unchanged within noise (1,719 vs 1,723; 1,859 vs 1,861; 2,132 vs 2,075) while refusals fall (Hypothesis 3 holds as the rc.2 review framed it); the honest 12 / 18 / 25% poorer (uplift down two fifths to a half) and five to six times richer than a stamper (Hypothesis 4); false approvals 11 / 14 / 14 / 16 against 13 / 13 / 14 / 17, the lag kind zero. At 50/50 the market collapses — queue 609 (control 129), the state delivering 500 against 1,169, on-time 0.25, liability 20,643 — by the same mechanism

Source: HANDOFF.md l.1457–1463

The take: at 60/40, 10% clears everything — liability 0, plants unfunded 0, Ω 7,836, careful wealth 5,066 (above the control: the surcharge doubles too); at 50/50 fees do not carry the clock at any take swept (liability 20,643 / 13,169 / 7,637, queue 609 / 342 / 178 at 5 / 10 / 15%).

Source: HANDOFF.md l.1469–1471

The engine reads 1.4.0 §12.3's "no member free to commit" per tick — one free member blocks every award summons that tick — so at 50/50 the state is blocked on 79% of ticks with 289 requests waiting and 2.0 members free (60/40: 54%, 14 waiting, 1.5 free, but the members are all committed often enough that the state summons 21 deliveries per 10 ticks); the summonses come in bursts when none is free (the first at tick 288, 38 at once; 70% of all troop deliveries in bursts of twenty or more), the queue grows a request a tick between them, and the queue's age is the lateness.

Source: HANDOFF.md l.1493–1498

The collapse is the stall; the two accounts decide how the bursts are financed.

Source: HANDOFF.md l.1503–1504

(2) T_susp ∈ {0, 50, 100}. At 0 a suspended stamper is a probationer at the next rotation (460–570 suspensions a run);

Source: HANDOFF.md l.1506–1507

$T_{\text{susp}}\in\{0,50,100\}$ moved who is free and repaired neither split; honesty paid for lowering it at both.

Source: spec/agentic_economics_loop_spec_v2_9_0.md l.1286–1286

For the next decision, in order: (1) the troops' working capital at 50/50 — advances before old debts under stress, or a working-capital target above R_T, or the cannot-serve protocol that bounds the debt at the source (the take does not do it: 15% at 50/50 leaves 7,600 owed); (2) the take at 60/40 as 10% if plants unfunded must be exactly zero in every seed; (3) the liability stock that never clears once honest; (4) the honest 12–25% poorer — Baruchi's call; (5) drift at m_max 2, false approvals 80 against 64, read with Q14 first.

Source: HANDOFF.md l.1481–1485

No value is right at both splits and none repairs 50/50; the lever is the same one the stall reads the other way. The item for the decision is the summons test, before the working capital or the protocol.

Source: HANDOFF.md l.1512–1514