What the sim has answered

Why the rest are missing

Counted per part, not per row, so a row missing two parts appears twice. The gear and inventory codes are `resolver.js`'s own, because it owns that classification for the site and the sim alike; the trip and verdict codes are derived at the dataset seam, which is the only place that can see the stat blocks and the ranking. A reason attributed to a generator is that generator's sentence quoted; the rest describe an absent key and say so.

Stage is the field worth reading. "never asked" means the solver was never given the row, "refused" means it was given it and declined with a diagnosis, "rejected" means it answered and the answer failed acceptance, and "declined, cause not recorded" means it was one of those and the generator wrote its reason to a console instead of a file. That last one is a gap in what the sim reports about itself: 0 of the parts below. It reads zero because the rows that carried it were never rows. Until 2026-08-29 this page manufactured three tier rows under every style tab, including the tabs nobody ever dressed, and handed the solver a question about a setup that does not exist. Those tabs are listed in the table without rows now, and this counter can only rise, which is the only direction that means anything.

• 57a hybrid row observed by the siteThis tab wears one style and carries the others as switches. The ranking compares single styles, so it was never asked about this combination.
• 34no bag was solved, so there was nothing to rank observed by the siteThe verdict is written beside a solved bag. This row has none, so the ranking was never asked about it.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 91 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 91 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 85 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 85 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 93 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 93 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 95 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 95 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 92 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 6noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 92 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 4noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 87 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 4noBuildToSolve never asked the generator’s own wordsthe gear solver refused this row (slayerLocked), so there was no loadout to build a bag around and this generator was never handed the row. Its words: needs 87 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 91 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 91 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 85 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 85 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 93 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 93 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 95 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 95 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 92 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 3slayerLocked never asked the generator’s own wordsneeds 92 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.
• 2slayerLocked never asked the generator’s own wordsneeds 87 Slayer to attack; a mid account has 80. No mid account can start this fight, so no loadout is solved for this tier. The max tier can.
• 2slayerLocked never asked the generator’s own wordsneeds 87 Slayer to attack; a budget account has 80. No budget account can start this fight, so no loadout is solved for this tier. The max tier can.

What the site withholds

A gap above is something the sim could not answer. This is something the site refuses to serve, which is a different claim and is the reason it has its own heading. Hand-authored loadout data is never to reach a browser: not the gear, not the bags, not the verdicts, not the gp figures. It is not deleted from the model, because the solvers read it and the audits rank against it, and deleting it would destroy the yardstick the sim measures its own progress by. It is barred at the one seam every page and every API response passes through.

This list is read out of that seam's own ledger, so a field barred later appears here on its own and a field the sim learns to derive leaves on its own. 785 per-slot gear overrides, 82 boss records whose hand-written bags are blocked from the sim as well as the site (nothing reads them, not even as a comparison), 237 typed style verdicts and 0 distinct typed item names were withheld from the payload this build. 11 verdicts survived, because the sim proved those.

• ○presets withheld hand authored9 hand-authored style.tier loadouts, 11 worn slots each. A typed best-in-slot kit per style and tier. The largest single block of authored gear in the payload.What would fill it: gen-gear.js / solvedgear.js, which answers 486 of the 582 site rows today. The 96 it declines carry gen-gear.js's own sentence in `gearRefusedBy`.
• ○wildyPresets withheld hand authored9 hand-authored Wilderness risk kits (high / mid / rag) per style. Same shape as `presets`, selected by `wildyTier` at a risky encounter.What would fill it: The loss dial Shane ruled on 2026-09-02 ("Two dials ... the loss dial should also be on bosses outside of the wild"). Not implemented; gen-gear.js still refuses all 10 risky encounters outright via RISK_REFUSED.
• ○tierSwaps withheld hand authoreda per-tier item substitution table (3 tiers). A typed "at this tier, use that item instead" map. It exists only to turn an authored template into a tiered one.What would fill it: gen-builds.js solves each tier separately, so a solved bag is returned WHOLE and skips the swap entirely. Nothing on the page reached this.
• ○ammoFor withheld hand authoreda typed default arrow / bolt / dart / javelin per class per tier. A loadout choice, typed. resolver.js's applyAmmo reads it to fill the ammo slot.What would fill it: THE ROUTE OUT HAS BEEN TAKEN, 2026-09-16, and the old sentence is kept below it because the record of how a bar closed is what this ledger is for. It read: "A generated ammo-compatibility field on ITEM_POOL. resolver.js already says so at its own refusal: 'If a weapon that SHOULD be viable starts losing this way, the fix is a generated ammo-compatibility field on the pool, not a list here.'" That field is what ammochoice.js reads, and `BEST_AMMO` in this file is the answer it gives.
• ○noAmmoWeapons withheld hand authored3 typed weapon names that need no ammunition. The old reason, kept for the record: "A typed claim about three weapons. It is a fact about the game and a source could state it, so the preference exception does not cover it." The first sentence is what was false. A source DOES state it, and data.js had already read the sentences.What would fill it: THE ROUTE OUT HAS BEEN TAKEN, and this field is kept rather than deleted because the record of how a bar closed is what the ledger is for. It used to read "The same generated pool field as `ammoFor`", which pointed at a field nobody was building and made a mechanic that already existed look years away. What filled it was `NO_AMMO_PATTERNS`, which had been in data.js the whole time, materialised over ITEM_POOL in this file. `ammoFor` still waits on that generated pool field; this no longer does.
• ○overhealFood withheld hand authoredthe typed name 'Anglerfish'. A consumable choice, typed. resolver.js unshifts it into an authored bag.What would fill it: gen-builds.js, which already chooses food per row. Reached only from buildInventory, which returns [] without authoredLoadouts.
• ○comboFood withheld hand authoredthe typed name 'Cooked karambwan'. A consumable choice, typed.What would fill it: gen-builds.js. Same closed path as `overhealFood`.
• ○comboFoodWildy withheld hand authoredthe typed name 'Blighted karambwan'. A consumable choice, typed, plus a typed rule that the Wilderness wants the blighted form.What would fill it: gen-builds.js, once the loss dial exists.
• ○wildyTier withheld hand authoreda typed max/mid/budget to high/mid/rag mapping. Its only purpose is to index `wildyPresets`, which is barred above. It is a naming of risk bands nobody derived.What would fill it: The loss dial. Until then there are no risk bands to name.
• ○bosses[].inv withheld hand authoredthe hand-authored inventory per boss and style. 73 distinct item names.. A typed bag. This is the field Shane's "nothing hand authored" rule names first.What would fill it: gen-builds.js / solvedinv.js, which answers 614 rows, plus `bagRefusedBy` for the rest.
• ○bosses[].styles[style][tier].base withheld hand authoreda preset id per boss, style and tier. 582 of them.. Names one of the barred `presets`. A pointer at authored gear is authored gear.What would fill it: solvedgear.js. The tier KEY is kept as an empty object so the row still exists; see `keptShape` below for why that distinction is load-bearing.
• ○bosses[].styles[style][tier].set withheld hand authoredper-slot hand-authored gear overrides. 564 sets, 785 slot overrides.. Typed worn gear, slot by slot. The single largest authored loadout surface after the presets themselves.What would fill it: solvedgear.js for 486 rows; `gearRefusedBy` states the generator's own reason for the other 147 keys it declined.
• ○bosses[].styles[style].v withheld hand authoreda typed best / good / ok / no verdict per boss and style. 226 authored, 11 generated by gen-verdicts.js and KEPT.. A recommendation about which style to bring, typed. dataset.js has said since the style ranking landed that these are "the largest hand-authored surface left on this site" and that Shane called them untrustworthy.What would fill it: gen-verdicts.js (11 rows today, sourced from BOAT_COMBAT) and `styleRank`, which is the model's computed per-tier ranking and already decides which style the page opens on. app.js has read `computedVerdict` and never `boss.styles[s].v` since the ranking landed, so the site rendered none of this.
• ○bosses[].hybrid withheld hand authoredthe typed `hybrid` record on 18 bosses: `styles` (which styles a fight combines, in a typed order), `mode` (on tormented-demons) and `why` (a sentence).. A loadout decision, typed. `styles` decided whether a switch tab existed, which styles it carried, and through its ORDER which style was worn at three bosses. `mode` decided the tab shape. That is what a player WEARS and CARRIES, which the 2026-09-02 bar keeps off the site.What would fill it: Nothing needs to: the stamp answers every question this field used to. What the stamp still cannot answer is its own declared refusal, STYLE_AXIS_REFUSALS.hybridAxis in styleaxis.js (a requirement gated on an unwired delve dial, a worn style with no published evidence yet, and unpriced switch ticks).
• ○bosses[].gp withheld hand authoreda typed gp-per-hour or gp-per-kill string on 15 bosses. A numeric claim about the game, typed, rendered as a chip on the boss list and the boss header. No source in the model states it.What would fill it: audit-gp.js prices drop tables per tier, and Shane's GP DRIVER dial decides what a player actually picks up. Neither reaches the payload today.
• ○bosses[].noFill withheld hand authoreda typed "do not top up the spare slots" flag on 4 bosses. A loadout policy, typed. resolver.js's fillPolicy and addOverhealAndCombo read it.What would fill it: Nothing needs to. A solved bag is returned WHOLE and skips the spare fill entirely, so the policy has had no effect on a rendered bag since solvedinv.js landed. The invnote sentence it drove is corrected in app.js in the same change: it was describing a composer that no longer runs.
• ○WILDY_TIER_LABEL withheld hand authoredtyped display labels for the three tier buttons at a risky encounter: 'Geared', 'Mid risk', 'Rag'. Three names for risk bands no source states. The tiers are account states (Shane, 2026-08-12), so the same three labels apply at every encounter.What would fill it: Nothing needs to: resolver.js tierLabel returns Max / Mid / Budget and no longer reads the table, which is what makes this bar safe where the 2026-09-03 one was not.
• ○bosses[].pk withheld hand authoreda typed "this content is PvP-risky" flag on 2 bosses (zombie-pirates, revenants). A claim about the encounter, typed. It is a fact a source can state, so the preference exception does not reach it.What would fill it: FILLED, 2026-09-03. The served record carries a `pk` again and its VALUE is now computed: barAuthored() drops the typed flag and re-derives it from the model's own WILDERNESS_LEVEL, which gen-wildernesslevel.js reads off the encounter and location pages with a quote per row. The two agree exactly, 10 and 10 with none on either side alone, and PK_MERGE publishes that comparison on /healthz.

Still hand authored, and still shipping

The honest debt, and the part of this page to read first. Each of these was measured before it was left in, and each says what barring it would have cost: an empty list here is the goal state. Nothing is quietly kept because removing it was awkward.

• ●bosses[].req kbd Wilderness depth still served hand authoredthe typed "Level 42 Wilderness" in the kbd requirements, the one Wilderness depth on the site the model does not derive. The model is RIGHT to have no row for it, so there is nothing to substitute. gen-wildernesslevel.js reads the depth an encounter RESIDES at, and the King Black Dragon lair is not in the Wilderness; only the walk to the lever is. The authored sentence says exactly that ("the lair itself is not the Wilderness"), which is a fact the reader needs and which no row in WILDERNESS_LEVEL states. Stripping the number would leave a depthless warning about a walk, which is worse than a typed 42.Could the sim derive it? NOT FROM THE ENCOUNTER PAGE, which is why the table has no row. It would need the depth of the ROUTE rather than of the fight, a distinct fact the model does not hold for any encounter. Published as WILDERNESS_DEPTH_MERGE.typedWithNoModelRow so the one exception is counted rather than assumed.
• ●bosses[].cmb still served hand authoreda typed combat level on all 79 bosses. Not loadout data, and it is rendered in the prerendered boss page, the boss header and llms-full.txt. Barring it blanks a stat line on 79 crawlable pages for a field outside the ruling's stated scope.Could the sim derive it? NO. dataset.js has recorded since the hitpoints merge that BOSS_STATS has 32 distinct keys and not one of them is a combat level, which is exactly why `hp` could be merged from the sim and `cmb` could not.
• ●untradeableSource still served hand authoreda typed "how to get one" hint for 64 items with no GE listing. It is a descriptive hint rather than a loadout choice, and data.js already classifies it as prose that stays: "UNTRADEABLE_SOURCE is the remaining hand-authored surface here. It is a `how to get one` hint rather than a claim the sim computes on, and no source in the model states it, so it stays until something does."Could the sim derive it? NO source in the model states it today.
• ●bosses[].hp fallbacks still served hand authored4 bosses keep an authored hitpoints figure. Each is a declared refusal already, not an oversight: three raid stat blocks are stand-ins for a whole encounter and the Gemstone Crab's hitpoints are infinite on its own page. Carried in HP_KEEPS_AUTHORED with the reason per boss.Could the sim derive it? NO, and the sim says so rather than guessing.

Every row

Worst first. Each block is one boss; each square is one style tab at one tier, split into 4 bars in the order worn gear, inventory, kills per trip, style verdict. A filled bar is an answer from the model, an empty one is a gap. A style tab with no squares at all is one the author gave a verdict and no setup at any tier: it is a tab, not three rows, and the solvers were never handed it.