The price of work

Finding [S1–S11, C-2.0]. Every run was priced by fiat at $\beta=20$ from a bottomless wallet; the selection rule's price term never moved, §4.6's execution-cost operation was never called, and nobody delegated. So the credit was a score: the driver's backing stood at $10^7$ and $v_t$ moved from 1.0000 to 0.9995 in a thousand ticks. "The state is broke" (runs 8–9) was never the loop breaking but a cash rule imposed on $\Omega$ as a stand-in for servers; the two accounts, $R_T$, the plant reserve and the order of payment existed to ration that cash. Scarcity had no signal but a queue (run 11's hard spike: 20–40 ticks in line, unspoken to). Baruchi's frame: decouple the servers from the economy; let printing make money worth less; give the money something to balance against.

Source: spec/agentic_economics_loop_spec_v2_9_0.md l.1298–1298

What it is. The first major version: a change of approach to money, on Baruchi's frame ("the servers needs to be decoupled from the economy, otherwise we contaminate it"; "maybe it is time we introduce floating prices"; "if I print more money then money is worth less"; "the actors have nothing to buy").

Source: HANDOFF.md l.1628–1630

2.0.0 lets the price float: requesters with finite wallets (issuance 10 a tick each, wallets 200) choose a ceiling and lock the all-in maximum at posting (a root's first lock; the nominals and the fee rate freeze there; I29); members bid; the price level P_t — the trailing mean of every awarded price, P_0 = 20 before the first — indexes every nominal the rules fix (f, G^V, D, φ, K, r_s, c_T, the plant pots) in place of v_0/v_t, frozen per case; effort costs stay nominal as the anchor; there is no backing (sources: issuance, emission, creation; sinks: the burns, retirement, departure); Ω is never broke — it creates for a payment the rules require and retires the deficit from every inflow, both public (I28) — and A_T, R_T, the plant reserve, Ω's order, the troop account and the liability stock are retired; the state delivers at cost (c^D_T = B̄) under a capacity N_T = 40 outside the ledger (a delivery holds two places until its inspection is seated; a refusal for capacity is recorded and told); B̄ counts members' awards only; a request below B̄ is flagged and its requester may raise the ceiling; delegation waits for 2.1.0 (it needs costs that differ across actors).

Source: HANDOFF.md l.1631–1639

What changed. Demand: requesters with wallets and a declared budget policy, funded by issuance; the pay-in locked at posting, a root's first lock; the ceiling the requester's, raised when flagged below market (§2.1, §4.1, §5.1). Supply: members bid at or above their reservation; the state at $\bar B$ from members' awards only, and at cost (§5.1, §7.7). Money: the price level $P_t$ indexes every nominal the rules fix, frozen per case; effort costs nominal as the anchor; no backing — sources issuance, emission, creation, sinks the burns, retirement, departure (§4.4–4.6). The state: never broke — creation for a required payment, retirement from every inflow, the deficit and $\Omega$'s net position public; $A_T$, $R_T$, the plant reserve, the order, the troop account and the liability stock retired (§10.4). Servers: $N_T$ outside the ledger, the capacity case recorded, two places per delivery (§12.3). Delegation deferred to 2.1.0 with heterogeneous costs.

Source: spec/agentic_economics_loop_spec_v2_9_0.md l.1300–1300

580 runs, 20 seeds a cell, 1,000 ticks unless stated, every per-tick check and final full audit passed. Fixed-price controls: runs 10–11 (1.7.1-summons/, 1.7.1-riots/), at P_t = P_0. Everything below is read from the second period on and, where it is wealth, in real terms (W · P_0/P_t).

Source: agentic_sim/results/studies/2.0.0-price-notes-for-reviewer.md l.3–3

Findings. The credit is worth less: P_t 31.6 / 34.1 / 37.4 / 37.3 at 50–80% with bidding members (the ceiling control 21–24: every bid the customer's ceiling, a third of a wallet the bids drain to 55), the cap β_max = 40 under any pressure. Set by the free members' margins, not by cost: at 60/40 two of eight members deliver, awarded every free tick, margin 3.0 for eight hundred ticks, bid 29, the state at B̄ behind them; the adaptive tail with twenty members free is the one place the price falls — to 23. Inflation is real and lands on holders: a 2,000 shock (9% of M) lifts P_t 15% in a window, the stampers' real wealth −13%, the honest's unchanged; it does not come back and Ω retires nothing of it. The state at cost is below cost: a fifth of its deliveries fail and lose B̄ + G, nothing offsets it, Ω's unlocked balance is zero from tick ~400 in every cell and the deficit grows 10–20 a tick without bound (3,700 at 50/50 to 59,700 in the flood); the 1.x-margin control runs the identical path with the deficit at zero — Hypothesis 7 fails, Hypothesis 6 inverts. The capacity case: under N_T 40 the crowd fills the field in a window (14,349 request-ticks refused, on time 0.84, rebates 16,000), the flood keeps it full (on time 0.33, rebates 6,300 a window all created, the deficit 59,700, the requesters hoarding 3,000 each, M 99,000); unbounded, 182 troops at the peak and nobody waits. Honesty is price-neutral: FA 10.7 / 10.4 / 12.7 / 14.7 (run 10: 11 / 14 / 15 / 18), the lagging auditor 0 at every fixed split, stampers' real wealth a tenth of the honest's.

Source: HANDOFF.md l.1686–1697

	50/50	60/40	70/30	80/20
P_t end, bidding (ceiling control)	31.6 (23.2)	34.1 (23.6)	37.4 (23.7)	37.3 (20.9)
members' mean price / troops' (bidding)	28.1 / 31.2	28.6 / 33.0	28.1 / 34.8	26.2 / 35.1
margin m at the end (bidding)	2.90	3.06	3.24	3.21
member awards / troop deliveries (bidding)	1,398 / 528	1,195 / 670	992 / 812	769 / 1,025
state's share of deliveries, bidding (ceiling)	27% (22%)	36% (37%)	45% (49%)	57% (65%)
posted (bidding / ceiling)	1,933 / 2,330	1,871 / 2,289	1,808 / 2,262	1,799 / 2,169
requesters' wallet at the end (bidding / ceiling)	91 / 59	77 / 60	70 / 53	68 / 283
deficit at the end (bidding / ceiling)	3,731 / 2,531	5,623 / 5,595	7,386 / 7,018	10,126 / 8,867
on time (bidding / ceiling)	0.988 / 1.00	0.986 / 1.00	0.986 / 1.00	0.993 / 1.00
careful real wealth (bidding / ceiling)	1,840 / 2,494	1,897 / 2,714	1,938 / 3,103	2,263 / 4,012
stamper real wealth (bidding / ceiling)	233 / 272	229 / 292	220 / 316	246 / 408
false approvals, 2.0.0 (run 10)	10.7 (11)	10.4 (14)	12.7 (15)	14.7 (18)
suspensions; lagging-auditor kind	72; 0	86; 0	100; 0	113; 0

Source: agentic_sim/results/studies/2.0.0-price-notes-for-reviewer.md l.13–27

What ran (results/studies/2.0.0-supply/, 400 runs, 20 seeds a cell, 1,000 ticks, issuance 60 a tick, all audits passed;

Source: HANDOFF.md l.1716–1716

Findings. Supply takes the price to cost, not to a plateau: P_t 22.5 / 11.6 / 10.9 at 20 / 30 / 40 honest (cost 8, bids 8(1 + m)), margins 1.9 / 0.36 / 0.23, members free 9.6 / 18.7 / 28.2. The strategy's ratchet has no interior equilibrium: in every cell where it runs, a free member's awards per tick converge on margin_down/(idle_k · margin_up) = 0.278 (0.270 at 20 honest, 0.273 at 60/40 at every cap, 0.25–0.27 across run 12) — the price rises until the requesters buy one task per 3.6 ticks per free member, and where the demand model cannot reach that rate (30 and 40 honest: K_open holds posting at 2.8 a tick) the margins decay to the floor.

Source: HANDOFF.md l.1723–1727

The cap was not what run 12 read: β_max 60 gives the same 34.1 as 40 at 60/40 — the wallet rule (a third of a 35 wallet, raised to B̄) binds first; 30 binds at 28.8; with market-aware requesters the cap is the price level to the unit (39.6 / 59.5). The state's pricing never lowers the price: at B̄ it never undercuts a member; where the requesters' ceilings sit at B̄ (the wallet rule with K_open lifted) it delivers 39–75% at 17–27 and the deficit runs to 6,000–14,000 — the cheapest market in the study is the one the state pays for.

Source: HANDOFF.md l.1729–1733

Honesty unmoved. Classification: supply is the trigger and the cure (the cliff between 20 and 30 honest); the strategy's constant places the cliff; the state's pricing sets the price only where the requesters' ceilings are B̄, at a loss; the cap is the price level only when the requesters bid above the market.

Source: HANDOFF.md l.1735–1737

	20 honest	30 honest	40 honest	60/40 (8 members; cap study, β_max 40)
P_t at the end / mean from period 2 / window max	22.5 / 24.4 / 28.0	11.6 / 12.6 / 14.0	10.9 / 11.6 / 12.3	34.1 / 33.1 / 34.1
margin m at the end	1.91	0.36	0.23	3.06
members free (mean from period 2; minimum)	9.6 (4.3)	18.7 (12.5)	28.2 (22.3)	3.2 (0.15)
posts a tick; posts refused by K_open	2.64; 2,981	2.79; 3,210	2.79; 3,206	1.87; 1,270
awards per free member-tick (the ratchet's constant: 0.278)	0.270	0.149	0.099	0.273
state's share of deliveries	0.1%	0	0	36%
requesters' wallet at the end	163	3,246	3,514	77
honest real wealth per actor; false approvals	1,446; 14.2	830; 10.5	589; 7.4	1,897; 10.4
deficit at the end	0	0	0	5,623

Source: agentic_sim/results/studies/2.0.0-supply-notes-for-reviewer.md l.13–23

For the decision (notes §8): (1) the state's cost rule — at cost loses the failure rate; candidates c^D_T = B̄(1 − b), the fee, or the margin as a reading; (2) the ceiling cap is the price level — the market has no anchor below the customers' maximum but the free members' margins and the state never undercuts; (3) the rebate under a refusal for capacity prints; (4) P_t at the period boundary would keep the three index-moved reviewer tests passing; (5) delegation (2.1.0) — the money is in the members' hoards.

Source: HANDOFF.md l.1702–1705

The reviewer's reply (October 19). Run 12 accepted as recorded; the raise-policy change accepted as a strategy's reading of its own declaration, the first wave discarded. The three rule fixes will be 2.1.0 and do not depend on the price question — the state at cost in expectation (c^D_T = (1 − b_T) B̄ − b_T G, b_T the trailing fail rate of the state's deliveries), creation only for what returns (summons capital, plants) with rebates from income and the bounded liability stock, a capacity refusal told, P_t updating at period boundaries. Before the draft: four evaluations under the rules in force, so the oligopoly can be classified — supply, strategy, or the state's pricing — run 13. Delegation and entry go to 2.2.0 together.

Source: HANDOFF.md l.1707–1712