Owner guidance 2026-08-29 for governor review 10476 (PR #11), finding by finding:
1. Explicit tier rank/tie-break for equal Mining gates. pickaxe_tiers
declared bronze and iron both at Mining 1 / Attack 1 while the
level-1 worked example called bronze canonical and carried it (27
slots) — internally contradictory under the entry's own
highest-unlocked acquisition policy. pickaxe_tiers now carries an
explicit rank (bronze 1 .. rune 6) and the selection rule resolves
equal Mining gates by rank: Mining 1 deterministically selects IRON
(7-tick roll, rank 2) over bronze (rank 1), and Attack 1 EQUIPS it
(iron attack gate = 1) -> 28 ore slots. Stated at every policy
layer (summary, accounts.requires_other, best_practice.equipment,
prepare state, economy.supplies, a14 equip-tool gate, pickaxe_tiers
head comment). An owned-only bronze run is documented as an operator
choice, never the entry's selection rule.
2. Level-1 lifecycle/rate inputs recomputed. The level-1 worked
example now uses the equipped iron pickaxe: 28 ore slots per drop
load, iron cadence [4200, 6000] ms (midpoint 5.1 s), and the
copper/tin rate re-derived from that cadence: 1.6 rolls x 5.1 + 0.5
drop + 0.39 break = 9.05 s/ore -> 398 ores/hr -> ~6,960 XP/hr (pin
6,900), replacing the bronze-midpoint 360 ores/hr / 6,300 XP/hr.
Updated in mine_sites (se-varrock-copper / se-varrock-tin),
economy comment block, profit_drivers, worktree example, and risk
notes; no other rate changes (fast-xp iron 28,000 and banked rates
untouched).
3. Recomputed 1-Attack/1-Mining fixture + capacity proof. Schema:
pickaxe_tiers gains required rank (1..6); pickaxe_examples gain
required expected_slots (28 equipped / 27 carried). Validator:
select_pickaxe() ties by rank; every pickaxe_examples row is
recomputed including expected_slots, the 1-Attack/41-Mining
rune-carry fixture is preserved, and the 1-Attack/1-Mining row is
now mandatory (selects iron, equipped, 28). Committed negative
fixtures f18-f22: bronze-at-Mining-1, carried-at-Attack-1, wrong
slots, missing 1/1 fixture, and a tier missing rank are all
rejected (22 selftests pass).
Preserved: complete GE tool acquisition loop, structured copper/tin
below-15 fallback, fast-xp/banked-income modes, dynamic 27/28
inventory handling, declared banked targets, respawn/world-hop model,
and all prior rate gates. version 6 -> 7, updated 2026-08-29.
Gate: python3 tools/validate.py -> 2 entries, 0 errors, 1 pre-existing
F2P WARN (low-alch-f2p researched, untouched); --selftest -> 22 PASS;
git diff --check clean.
Governor bounce 5 (2026-08-28) repaired finding by finding:
1. Tool purchase is now executable end to end: new a10 ge_open before any
ge_buy; a12 wait-ge-fill is BOUNDED (one 3-6 min fill window) and on
no-fill/partial-fill TIMEOUT cancels the resting offer -> a13
collects/refunds -> fail closed via recover (3 loops -> failed); a resting
offer is never re-placed indefinitely. Actions renumbered a11..a17.
2. min_level 1 copper/tin below-15 path now has STRUCTURED fallback rows
(mine_sites se-varrock-copper / se-varrock-tin, level 1, 17.5 xp, safe
East Varrock) added to mining_modes.fast-xp.sites; level-1 and level-14
selections are pinned in a new mining_selections block and recomputed by a
new select_fast_site() validator gate (+ 3 committed selftest fixtures).
3. a2 no longer advertises silver/coal/gold/mithril banked targets: it
restricts banked-income to the declared rows (iron @ se-varrock-bank,
adamantite @ al-kharid-south-adamantite); superheat-coal technique is
marked informational-only. New ge_open/ge_offers_done/ge_collect
validator gate (+ 1 fixture). Modeled below-15 rate 6,300 XP/hr (was
the unbacked '~8,800' prose) recomputed from the bronze cadence midpoint.
Validator: 2 entries, 0 errors, 1 pre-existing WARN; --selftest 17 PASS;
git diff --check clean.
- Buy-limit reset NEVER cancels an offer: quote_age_policy now aborts
only on a stale scanned edge or explicit strategy reallocation; the
4h window is a capacity clock (limit_ledger), not an offer deadline.
Removed every lapse-as-cancel trigger from best_practice, whens,
quote_age_policy, a5/a8, worked example, risk, and the a6 verdict.
- Completed offers pass through the collection box: new canonical
sub-states buy-filled-awaiting-collect / sell-filled-awaiting-collect;
a full buy cannot sell and a full sell cannot close before ge_collect
(scheduler dispatches collect for all four *-awaiting-collect states).
- Works the arrowtips example side-aware: a 4,000/10,000 partial BUY
cancel returns 4,000 units + refunded CASH for the unfilled 6,000 —
never 'refunded units'; the phantom 10,000-unit withdrawal is gone.
- conservation_violations() is now side-aware (buy: quoted =
filled + still_open + cancelled_unfilled with acquired items = filled
only, cash = spent + still_committed + refunded; sell: quoted =
sold + still_open + returned_unsold) with slot_phase_violations()
phase gates and committed negative fixtures: phantom refunded buy
units, a full buy sold before collection, a full sell closed before
proceeds collection. Existing tax/scheduler/limit fixtures preserved.
- grammar (ge_cancel/ge_collect), schema sub_states minItems, entry
version 5 + sources updated accordingly.
Governor bounce 4 + owner guidance 2026-08-28 (structured, provable, full lifecycle):
1. site model is data, not prose: new mine_sites table (ore, gates, rock_count,
respawn_seconds, rolls_per_ore, modeled_ores_per_hour, world_hop, selectable,
inventory_policy, access, safety, route). Adamantite @ Al Kharid south is now
respawn-BOUND (2 rocks, 240 s respawn): modeled 120 ores/hr ONLY with world
hopping (wiki F2P adamantite guide) = 70,080 gp/hr; no-hop ceiling ~30/hr =
17,520; the 839 ores/hr iron-cadence fantasy is gone and a cadence-ceiling
gate rejects it. Runite is selectable:false (Wilderness PK risk, ~12 min
respawn) — refused, not hand-waved. New hop state + hop_world action (a14)
implement the world-hop policy; long-respawn selectable sites REQUIRE
world_hop + a hop_world action (validator).
2. fast-xp success model: rolls_per_ore 1.33 (61+, derived from the wiki 52,500
benchmark: 1,500 ores/hr / 2,000 3-tick rolls/hr); rates recomputed with
success rolls + respawn math (3 rocks x 3.59 s = 10.8 s > 5.4 s respawn ->
zero idle, pinned) + drop/bank + breaks: xp_per_hour 28,000 (was 32,400 with
a flat 0.3 s and 1 ore/click); 52,500 stays a labeled external benchmark.
3. sub-29 fast-xp preserved: South-east Varrock DROP site (safe, same mode) —
the mode never silently switches to banking.
4. tool prep executable end to end: check-tool (has_item) -> withdraw-tool
(owned branch, bank_inventory) or buy-tool + wait-ge-fill (ge_offers_done) +
collect-tool (ge_collect, verify) -> equip-tool via menu_option Wield with
the Attack gate; recovery on purchase/equip failure (3 -> failed).
5. VALIDATOR PROVES THE CONTRACT (no phrase-search): pickaxe_tiers must match
the canonical F2P table exactly (Mining=use gate, Attack=equip gate,
roll_ticks); pickaxe_examples recomputed per row (1-Attack/41-Mining must
select + CARRY rune; 40-Attack/70-Mining equips; 5-Attack/30-Mining carries
mithril); both modes declared with fixed inventory policies; mode refs must
resolve to declared selectable policy-matching sites; Wilderness selectable
-> error; modeled ores/hr <= 3600/(rolls_per_ore x 2.7 s); xp_per_hour
within 2% of the modeled fast-xp rate. 13 committed --selftest fixtures
(negative cases included) wired into every gate run.
Evidence: validate 2 entries 0 errors 1 pre-existing WARN; --selftest 13 PASS;
synthetic 3-way merge main 79cab78 (incl. catalog) + head clean + gate green;
git diff --check clean. version 4 -> 5.
Owner lifecycle addendum (2026-08-28) + governor bounce 4 (lifecycle contract):
- state machine now covers the ENTIRE buy -> sell -> collect flow with no
stranded state: partial fills tracked per slot (filled_qty ledger), 4h-lapse
buys go to a new cancel_requote state (ge_cancel: units return to inventory,
cash refunds, slot freed) instead of collect, cancelled units still sell at
the high edge, lapse sells re-quote or lower rather than strand
- action grammar + schema extended ONCE with the cancellation mutation
ge_cancel (was missing; ge_clear/ge_offers_done are predicates only) and
used explicitly in the new a8 cancel-requote action
- concrete runtime state: minimum per-slot capital table (buy x 4h limit),
members 8-slot vs F2P 3-slot paths, per-slot offer phase + quantities,
buy-limit window, inventory/collection reconciliation in a9
- breaks connected to an explicit safe/resume path: resting offers persist
server-side + per-slot ledger survive the pause; resume = re-open GE,
re-check offers, continue current phase; never break mid-collect/mid-place
- rates qualified against the full lifecycle: 149,500 gp/h is named the 4h
buy-limit CEILING; realized range = ceiling x fill_factor (~0.5-0.9) x
duty_cycle (~0.90-0.95) ~ 96,300 gp/h midpoint; gp_per_hour_low 74,750
- pinned worked example: capital 22.5M -> 8 slot quotes -> arrowtips partial
fill (4,000/10,000 tracked) -> iron bar full fill sell (90,000 net) ->
cosmic no-fill ge_cancel + requote -> arrowtips partial matures at lapse
(cancel, sell 4,000, 68,000 net) -> first break after 6-10 rotations ->
resume -> stop conditions
- version 2 -> 3; evidence: validate 2 entries 0 errors 1 pre-existing WARN,
selftest 4 PASS (fractional-tax fail, ge non-margin pass, floored pass,
above-cap pass), synthetic merge main 79cab78 + head clean + gate green,
git diff --check clean
