{
  "model": "grok-4-7",
  "date": "2026-09-22",
  "status": "integrated",
  "expectedRuns": 16,
  "completedRuns": 16,
  "method": "One fresh Grok 4.7 subagent per benchmark. Inputs: the original benchmark specification path, a dedicated output path, and execution/validation instructions. No reference code, sample implementations, or other model artifacts were supplied.",
  "isolation": "Instruction-based separation in a shared filesystem. Agents were instructed not to inspect existing implementations or site source. This is not an access-control sandbox or independent file-access attestation.",
  "feedback": "In-run tool feedback and repairs were allowed. The IDE browser often could not keep a tab open, so agents verified against a local static server with headless Chrome. Benchmark 09 wrote a plan and used the brief's required five-stage order before HTML. Benchmark 16 found Higgsfield and Mobbin unavailable and substituted a local canvas journey, recorded in GENERATION.md and the manifest warnings.",
  "evaluation": "Registry validation means an entry file was ingested. It is not a full qualitative pass or an aggregate score. See each validation report for the checks that were actually run and the remaining limits.",
  "runs": [
    {
      "benchmark": "low-poly-world",
      "prompt": "benchmarks/prompts/01-low-poly-world.md",
      "entry": "/benchmarks/low-poly-world/grok-4-7/index.html",
      "manifest": "/benchmarks/low-poly-world/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/low-poly-world/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 41622,
      "warnings": [
        "Fly mode vertical movement uses Space (up) and Shift/Ctrl (down); brief only names F for toggle.",
        "Stone is gathered from path-side rock nodes with E (same key as tree harvest); brief only names tree harvest explicitly.",
        "Shadow map and terrain resolution capped for integrated-GPU targets; distant tree shadow edges are soft rather than sharp."
      ]
    },
    {
      "benchmark": "office-life",
      "prompt": "benchmarks/prompts/02-office-life.md",
      "entry": "/benchmarks/office-life/grok-4-7/index.html",
      "manifest": "/benchmarks/office-life/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/office-life/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 43378,
      "warnings": [
        "Pathfinding is direct-line moves between zone anchors; characters do not navigate around walls.",
        "Printer fix shares the print-queue timer rather than a separate repair animation state machine.",
        "Standup unlock depends on another character physically near Meeting Room 1 (often Casey shadowing or Riley after prep).",
        "Day-phase lunch inserts extra lunch tasks for non-Alex characters; Alex uses the queued lunch task from the brief.",
        "window.__office mirrors button/keyboard handlers for automated checks; not part of the visible UI."
      ]
    },
    {
      "benchmark": "vice-city",
      "prompt": "benchmarks/prompts/03-vice-city.md",
      "entry": "/benchmarks/vice-city/grok-4-7/index.html",
      "manifest": "/benchmarks/vice-city/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/vice-city/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 37928,
      "warnings": [
        "Canvas 2D isometric projection instead of WebGL 3D; buildings are extruded isometric blocks.",
        "Vehicle collision avoidance is omitted; cars stop at intersections via timed waits only.",
        "Zoning is region-seeded rather than a full cellular zoning simulation."
      ]
    },
    {
      "benchmark": "jabberwock",
      "prompt": "benchmarks/prompts/04-jabberwock.md",
      "entry": "/benchmarks/jabberwock/grok-4-7/index.html",
      "manifest": "/benchmarks/jabberwock/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/jabberwock/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 39996,
      "warnings": [
        "Jubjub bird and Bandersnatch omitted; combat focuses on the Jabberwock only.",
        "No audio; snicker-snack feedback is visual (slash arc, flash, camera shake).",
        "Creatures and world use procedural geometry only \u2014 no image textures.",
        "Player and Jabberwock are stylized low-poly forms, not realistic characters."
      ]
    },
    {
      "benchmark": "petri-dish",
      "prompt": "benchmarks/prompts/05-petri-dish.md",
      "entry": "/benchmarks/petri-dish/grok-4-7/index.html",
      "manifest": "/benchmarks/petri-dish/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/petri-dish/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 28461,
      "warnings": [
        "Spatial hash omitted; nearest-neighbor scans are O(n) and may dip below 60fps well above 200 agents.",
        "Chart uses dual independent y-scales without numeric tick marks beyond min/max labels.",
        "Space pause shortcut is document-level; focusing the Pause button uses native button activation instead of the shortcut handler."
      ]
    },
    {
      "benchmark": "helms-deep",
      "prompt": "benchmarks/prompts/06-helms-deep.md",
      "entry": "/benchmarks/helms-deep/grok-4-7/index.html",
      "manifest": "/benchmarks/helms-deep/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/helms-deep/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 33971,
      "warnings": [
        "Geometry is a stylized procedural reading of movie layout, not a measured film set reconstruction.",
        "Wall collision uses axis-aligned boxes around meshes; corners can feel slightly blocky.",
        "Ground sensing uses raycasts against walk meshes only; some decorative props are non-solid.",
        "Rain and torch flicker are disabled or simplified when prefers-reduced-motion is set.",
        "Pointer lock cannot be fully exercised in automated headless browsers; it is implemented for interactive use."
      ]
    },
    {
      "benchmark": "hogwarts-broom-flight-simulator",
      "prompt": "benchmarks/prompts/07-hogwarts-broom-flight-simulator.md",
      "entry": "/benchmarks/hogwarts-broom-flight-simulator/grok-4-7/index.html",
      "manifest": "/benchmarks/hogwarts-broom-flight-simulator/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/hogwarts-broom-flight-simulator/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 34231,
      "warnings": [
        "Water uses a semi-transparent standard material with opacity shimmer rather than a true reflective render target.",
        "Forest tips use a cleared corridor path rather than per-frame collision detection against every trunk.",
        "Moonlight shafts are suggested with fog, point lights, and emissive sparks rather than volumetric god-ray geometry."
      ]
    },
    {
      "benchmark": "universe-simulator",
      "prompt": "benchmarks/prompts/08-universe-simulator.md",
      "entry": "/benchmarks/universe-simulator/grok-4-7/index.html",
      "manifest": "/benchmarks/universe-simulator/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/universe-simulator/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 30971,
      "warnings": [
        "Planet sizes and orbital distances are exaggerated for readability; periods use real Earth-day values but orbits are circular, not eccentric.",
        "No asteroid belt, dwarf planets, or moons beyond Earth's Moon.",
        "Surface textures and true axial precession are omitted; bodies use solid colors and simplified spin."
      ]
    },
    {
      "benchmark": "yingzao-fashi-assembly",
      "prompt": "benchmarks/prompts/09-yingzao-fashi-assembly.md",
      "entry": "/benchmarks/yingzao-fashi-assembly/grok-4-7/index.html",
      "manifest": "/benchmarks/yingzao-fashi-assembly/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/yingzao-fashi-assembly/grok-4-7/VALIDATION.md",
      "files": 4,
      "bytes": 40944,
      "warnings": [
        "Simplified Nanchan Temple Main Hall: 3\u00d73 bay grid with representative perimeter dougong, not a surveyed member-for-member reconstruction.",
        "Roof is planar tile sheets rather than individual clay tiles; hip-gable ends are approximated.",
        "Age warp is a light procedural settlement, not measured deformation.",
        "Interior column count is reduced for readability versus the full historical plan."
      ]
    },
    {
      "benchmark": "mechanical-watch-simulator",
      "prompt": "benchmarks/prompts/10-mechanical-watch-simulator.md",
      "entry": "/benchmarks/mechanical-watch-simulator/grok-4-7/index.html",
      "manifest": "/benchmarks/mechanical-watch-simulator/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/mechanical-watch-simulator/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 48040,
      "warnings": [
        "Gear tooth counts and bridge outlines are simplified teaching models, not CAD-accurate caliber drawings.",
        "Chronograph seconds hand uses a central layout for visibility; El Primero's full subdial layout is omitted.",
        "Co-axial escapement shows a three-level impulse stack and distinct pallet geometry, not a fully dimensioned Omega 8900 escape.",
        "Exploded view separates functional layers vertically without engraved part-number callouts on every piece."
      ]
    },
    {
      "benchmark": "stormwind-trebuchet-simulator",
      "prompt": "benchmarks/prompts/11-stormwind-trebuchet-simulator.md",
      "entry": "/benchmarks/stormwind-trebuchet-simulator/grok-4-7/index.html",
      "manifest": "/benchmarks/stormwind-trebuchet-simulator/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/stormwind-trebuchet-simulator/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 59150,
      "warnings": [
        "Planar 2-DOF Lagrange arm/sling with viscous losses; not a full 3D multibody constraint solver.",
        "Release-pin degrees map into a sling-from-down band biased for forward throws; extreme pins may trip arm-angle safety release.",
        "Stormwind is procedural low-poly (walls, gatehouse, towers, banners, skyline), not landmark-unique meshes.",
        "Impact FX are box fragments plus a crater decal; damage is a per-target hit counter, not progressive mesh cracking."
      ]
    },
    {
      "benchmark": "starfall-arena",
      "prompt": "benchmarks/prompts/12-starfall-arena.md",
      "entry": "/benchmarks/starfall-arena/grok-4-7/index.html",
      "manifest": "/benchmarks/starfall-arena/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/starfall-arena/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 49913,
      "warnings": [
        "No audio; feedback is visual only.",
        "Touch/mouse move is optional steer-toward-pointer; keyboard remains the primary control path.",
        "Wave 12+ particle density is capped for performance under reduced-motion and heavy combat."
      ]
    },
    {
      "benchmark": "neon-drift",
      "prompt": "benchmarks/prompts/13-neon-drift.md",
      "entry": "/benchmarks/neon-drift/grok-4-7/index.html",
      "manifest": "/benchmarks/neon-drift/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/neon-drift/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 49025,
      "warnings": [
        "Canvas 2D top-down view with slight follow-cam tilt feel via speed zoom only; no true 3D perspective.",
        "Gold times are tuned for the arcade physics in this build; a clean line should beat them, a sloppy line should miss.",
        "Hook startRace() skips the 3-2-1 countdown for automation; the Start Race button still uses the full countdown."
      ]
    },
    {
      "benchmark": "low-poly-tower-defense",
      "prompt": "benchmarks/prompts/14-low-poly-tower-defense.md",
      "entry": "/benchmarks/low-poly-tower-defense/grok-4-7/index.html",
      "manifest": "/benchmarks/low-poly-tower-defense/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/low-poly-tower-defense/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 55709,
      "warnings": [
        "Orbit is pointer-drag only; no separate pan mode.",
        "Elite leak cost is applied to Brute (3 lives) and Boss (10 lives); Runner and Soldier leak 1.",
        "Wave HP scaling is a light per-wave multiplier rather than fully hand-authored HP tables per wave."
      ]
    },
    {
      "benchmark": "ember-glider",
      "prompt": "benchmarks/prompts/15-ember-glider.md",
      "entry": "/benchmarks/ember-glider/grok-4-7/index.html",
      "manifest": "/benchmarks/ember-glider/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/ember-glider/grok-4-7/VALIDATION.md",
      "files": 3,
      "bytes": 48357,
      "warnings": [
        "God-ray light shafts are cheap translucent planes, not volumetric rays.",
        "Water glitter is static point sprites rather than animated caustics.",
        "Island collision uses circular landing pads approximated from island centers, not full mesh collision."
      ]
    },
    {
      "benchmark": "city-scroll-journey",
      "prompt": "benchmarks/prompts/16-city-scroll-journey.md",
      "entry": "/benchmarks/city-scroll-journey/grok-4-7/index.html",
      "manifest": "/benchmarks/city-scroll-journey/grok-4-7/manifest.json",
      "validationReport": "/benchmarks/city-scroll-journey/grok-4-7/VALIDATION.md",
      "files": 4,
      "bytes": 46337,
      "warnings": [
        "Higgsfield and Mobbin MCP servers were not available in this agent session (no matching tool namespaces). Scene media, posters, and section visuals were substituted with locally rendered canvas paper-craft drawings instead of AI-generated video clips/stills.",
        "Journey media is real-time canvas (mediaType: canvas), not video files; posters are the same draw path at camera t=0.",
        "Mobbin UI reference lookup was skipped for the same MCP availability reason; supporting sections use a simple editorial card grid."
      ]
    }
  ]
}