{
  "schema": "sim-proof/1",
  "pr": 2587,
  "issue": 2528,
  "commit": "3cc2746af",
  "tier": "T3",
  "criteria": [
    {
      "id": "A1",
      "text": "Oracle test: over the corpus (offered ids, not-offered known places, unknown words, ambiguous words) the capability or the refusal, including its code and \"did you mean\" list, equals the old path's.",
      "status": "met",
      "evidence": [
        "`node --test apps/server/src/edge/boot/world-places-oracle.test.ts` -> pass, 0 differences: Eden 2026, about 2,200 cases at arrival and again after one walk, whole answers (ok, code, reason, did_you_mean, chosen, next) against the old path frozen in the test",
        "corpus: offered walk ids, offered other place verbs, not-offered known places by id for every place verb, place names and aliases, things' walk labels, unknown ids and words, ids of things and spots that are no place, ambiguous words (names two places share, thing labels that are a place's name or alias, labels two things share) as walk, stroll and 'walk:the …', and a cold sub-corpus with a tick before each case; the answers include place.ambiguous and capability.unknown (asserted)"
      ]
    },
    {
      "id": "A2",
      "text": "Milliseconds per act under explicitPlaceCapability at 50 agents fall by at least 60 % (M2's profile).",
      "status": "unmet",
      "evidence": [
        "https://github.com/soulsimulation/souls/actions/runs/43840052046",
        "https://github.com/soulsimulation/souls/actions/runs/43840059183",
        "`LOAD.md Under explicitPlaceCapability, 50 Souls, ms per act` -> base 32.7, 40.5 (mean 36.6); candidate 66.0, 34.2 (mean 50.1): +37 %, repeats overlap: no measurable gain, within noise",
        "Step 0: base mean 36.6 ms per act at 50 Souls, at or above 5, so the premise holds",
        "`parts under it, 50 Souls (base -> candidate)` -> offeredWalkObject 23.6, 36.0 -> 47.6, 9.6; resolvePlaceForVerb 8.1, 3.8 -> 4.1, 4.9; resolveBootPlace 6.1, 1.7 -> 1.8, 4.6; knownBootPlaces 2.1, 1.3 -> 0.7, 0.6; strictCheck 1.0, 0.7 -> 0.6, 0.6",
        "finding: under load the cost is offeredWalkObject's build of every walk offer (23.6 and 36.0 of 32.7 and 40.5 ms on base), which this PR does not remove; the local profile (one Soul, a kept focus build) put it near zero, so the design's step 3 was set aside on wrong evidence"
      ]
    },
    {
      "id": "A3",
      "text": "eval-scripted --battery m4-v1: same receipts and state hashes.",
      "status": "met",
      "evidence": [
        "https://github.com/soulsimulation/souls/actions/runs/43840064051",
        "https://github.com/soulsimulation/souls/actions/runs/43840067966",
        "`result.json per item: succeeded, why, endedBy, card outcome, state, closed` -> 12/12 the same, base eb00055f3 against candidate 3cc2746af, once whatson-interest's calendar id (minted per run: cal-er5i2f4g against cal-6toqlnml) is set aside as declared",
        "receipt-scored items name the same receipts and phases on both: eat:eat.e_2accdf4c started, play-music:play-music.e_18fdb81d started",
        "differences: admission retries only (world.busy, soul.busy), which refuse before the handler and write no receipt"
      ]
    },
    {
      "id": "A4",
      "text": "eval-load --souls 24,50 --act-sec 10 --profile, base against candidate, 2 repeats.",
      "status": "met",
      "evidence": [
        "https://github.com/soulsimulation/souls/actions/runs/43840052046",
        "https://github.com/soulsimulation/souls/actions/runs/43840059183",
        "`50 Souls, base r1, r2 -> candidate r1, r2` -> world's rest of handler 75.2, 82.3 -> 97.5, 75.1 ms per act; profile act.handler-other 30.9, 68.5 -> 39.5, 22.2; builds per act 1.32, 1.27 -> 1.26, 1.18; achieved 108/600, 110/600 -> 105/600, 109/600; acts in the window 38, 22 -> 23, 38",
        "`24 Souls, under explicitPlaceCapability` -> base 43.7, 37.5; candidate 46.5, 66.1 (noise)",
        "act latency not scored (bars): every arm's act calls queued about 19.5 s at the world (admission state, PR 2583 not merged); even arm only, eval-load has no uneven shape yet"
      ]
    }
  ],
  "media": [
    {
      "file": "doors-everywhere-desktop.mp4",
      "kind": "video",
      "seconds": 32.8,
      "viewport": "desktop 1280x720",
      "uncut": true,
      "realSpeed": true,
      "caption": "Eden 2026 at 3cc2746af, recorded by `clip --kit doors-everywhere` on souls-timed (run 43840108997): Clara, about 55 m from the dome, acts walk:eden-dome by id at 0:05, runs there through the closed door and stands inside at 0:30; a walk to a place by id behaves as before.",
      "start": "arrival",
      "url": "https://assets.eden.art/proofs/pr-2587/doors-everywhere-desktop.mp4",
      "preview": "https://assets.eden.art/proofs/pr-2587/doors-everywhere-desktop.preview.gif",
      "previewStartSec": 0
    }
  ],
  "files": [
    {
      "file": "LOAD-base.md",
      "caption": "eval-load base, agent/2528-meter-base eb00055f3 (run 43840052046): 24 and 50 Souls, 2 repeats, --profile --menu-stats",
      "url": "https://assets.eden.art/proofs/pr-2587/LOAD-base.md"
    },
    {
      "file": "LOAD-cand.md",
      "caption": "eval-load candidate, 3cc2746af (run 43840059183): 24 and 50 Souls, 2 repeats, --profile --menu-stats",
      "url": "https://assets.eden.art/proofs/pr-2587/LOAD-cand.md"
    }
  ],
  "checks": [
    {
      "cmd": "npx tsc --noEmit -p apps/server",
      "result": "pass",
      "note": ""
    },
    {
      "cmd": "node --test apps/server/src/edge/boot/world-places-oracle.test.ts tools/playtest/eval/prof-buckets.test.mjs",
      "result": "pass",
      "note": "3/3"
    },
    {
      "cmd": "node --test apps/server/src/w1-legacy-compatibility.test.ts apps/server/src/edge-ledger-strict.test.ts apps/server/src/edge-ledger-act-on-demand-id.test.ts",
      "result": "pass",
      "note": "run by the reviewer agent"
    },
    {
      "cmd": "node .github/scripts/guards.mjs scrub --base origin/agent/2527-walk-bind-once",
      "result": "pass",
      "note": "0 problems"
    },
    {
      "cmd": "node tools/changelog/compile.mjs --check",
      "result": "pass",
      "note": ""
    }
  ],
  "failedAttempts": [
    {
      "what": "steps 1 and 2 alone (lazy menu, known places once)",
      "result": "no measurable change locally on a far place by id: the strict check is skipped for a known place and the known places are cheap"
    },
    {
      "what": "the exact-id narrowing instead of the design's step 3",
      "result": "-30 % locally (20.9 -> 14.5 ms per call), no measurable gain under load: the load's cost is offeredWalkObject's walk build, which the local single-Soul profile hid"
    }
  ],
  "externalities": [
    {
      "what": "refusal path",
      "result": "strictCheck reads the world's known ids twice (without and with the menu)",
      "ok": true
    },
    {
      "what": "offered ids, words, callers passing an array",
      "result": "unchanged",
      "ok": true
    }
  ],
  "playtest": {
    "scenario": "eval-load --souls 24,50 --act-sec 10 --duration-sec 120 --profile --menu-stats on souls-timed, base then candidate, 2 repeats per arm; even arm only (eval-load has no uneven shape yet). eval-scripted m4-v1 (A3, one run per arm) is an equivalence check, in A3's evidence.",
    "arms": [
      {
        "name": "base agent/2528-meter-base eb00055f3, eval-load 24 and 50 Souls",
        "repeats": 2,
        "passed": 2,
        "url": "https://github.com/soulsimulation/souls/actions/runs/43840052046"
      },
      {
        "name": "candidate 3cc2746af, eval-load 24 and 50 Souls",
        "repeats": 2,
        "passed": 2,
        "url": "https://github.com/soulsimulation/souls/actions/runs/43840059183"
      }
    ]
  },
  "review": {
    "by": "reviewer agent",
    "verdict": "no correctness defect; every changed function returns the same result for every caller by construction; notes applied in 3cc2746af (cold menus, known non-place ids, asserted refusal codes, a doc comment moved)"
  },
  "limits": [
    "A2 unmet: no measurable gain under load; the remaining cost is offeredWalkObject's build of every walk offer (the design's step 3, not in this PR) and, locally, walkOnDemand's route search",
    "act latency not scored: every arm queued about 19.5 s per act (admission state, PR 2583)",
    "even arm only: eval-load has no uneven shape yet"
  ]
}
