{
  "schemaVersion": 1,
  "project": {
    "name": "Possible",
    "url": "https://possible.sh/",
    "repository": "https://github.com/fraylabs/possible",
    "package": "https://www.npmjs.com/package/@fraylabs/possible",
    "judgingPage": "https://possible.sh/judging/"
  },
  "recordedComparison": {
    "name": "Robot Snake: /goal control versus Possible",
    "question": "What operational knowledge does a non-expert's rough robot-snake request cause the agent to include?",
    "protocol": {
      "model": "gpt-5.6-sol",
      "controlPrompt": "/goal I want to make a robot snake",
      "onlyFollowUp": "Simulation first. I do not have a fixed budget or access to a 3D printer.",
      "controlEnvironment": "Empty Git repository, fresh CODEX_HOME, and no Possible, Robot Prototype, robotics, CAD, or MuJoCo skill.",
      "evaluationContract": "The Robot Prototype Outcome Pack contract existed before the control run."
    },
    "observedResults": {
      "goalControl": "A browser simulator, firmware handoff, hardware planning, and 18 passing tests.",
      "possible": "CAD, URDF/SRDF, MuJoCo simulation, autonomous obstacle-avoidance evidence, Rerun telemetry, 12 tests, 186 interface checks, and three defects caught and repaired by fresh verification."
    },
    "interpretation": {
      "goal": "/goal provides dynamic pursuit and adapts execution as evidence changes.",
      "possible": "Possible provides the reviewed outcome contract: workstreams, safeguards, interfaces, evidence, and completion conditions.",
      "together": "Possible defines the multidisciplinary completion target; /goal can sustain and adapt its execution."
    },
    "evidence": [
      {
        "label": "Control protocol and complete human input",
        "url": "https://github.com/fraylabs/possible/blob/main/apps/web/public/demo/robot-snake/CONTROL-RUN.md"
      },
      {
        "label": "Preserved control artifacts",
        "url": "https://possible.sh/demo/robot-snake/control/"
      },
      {
        "label": "Possible artifact manifest",
        "url": "https://possible.sh/demo/robot-snake/manifest.json"
      },
      {
        "label": "Possible completion report",
        "url": "https://github.com/fraylabs/possible/blob/main/apps/web/public/demo/robot-snake/evidence/outcome-receipt.md"
      }
    ]
  },
  "judgingCriteria": [
    {
      "criterion": "Technological Implementation",
      "claim": "Typed Outcome Packs coordinate execution and verification.",
      "implementationFact": "The compiler converts manifests into skill installs, owned workstreams, approval gates, and completion requirements.",
      "significance": "One contract governs the run from preparation through verification.",
      "evidence": {
        "label": "Outcome Pack compiler source",
        "url": "https://github.com/fraylabs/possible/blob/main/packages/packs/src/compiler.ts"
      }
    },
    {
      "criterion": "Design",
      "claim": "Each demo presents the outcome beside its proof.",
      "implementationFact": "The demo gallery exposes four visual outcomes and their preserved evidence.",
      "significance": "Judges can inspect the work without reading agent logs first.",
      "evidence": {
        "label": "Demo gallery",
        "url": "https://possible.sh/demo"
      }
    },
    {
      "criterion": "Potential Impact",
      "claim": "Possible supplies work a novice did not know to request.",
      "implementationFact": "The Robot Prototype pack covers mechanical design, simulation, control, telemetry, safety, and review.",
      "significance": "One rough request can start multidisciplinary work outside existing expertise.",
      "evidence": {
        "label": "Robot Prototype pack",
        "url": "https://github.com/fraylabs/possible/blob/main/packages/packs/src/robot-prototype.ts"
      }
    },
    {
      "criterion": "Quality of the Idea",
      "claim": "Outcome Packs make operational judgment reusable.",
      "implementationFact": "The Robot Snake run begins with an ambition and ends with an inspectable completion report.",
      "significance": "The system transfers more than a single capability or instruction.",
      "evidence": {
        "label": "Robot Snake completion report",
        "url": "https://github.com/fraylabs/possible/blob/main/apps/web/public/demo/robot-snake/evidence/outcome-receipt.md"
      }
    }
  ],
  "guidedEvidenceTrail": [
    {
      "step": 1,
      "stage": "Intake",
      "fact": "The confirmed product brief records facts, constraints, required outputs and acceptance checks.",
      "url": "https://github.com/fraylabs/possible/blob/main/apps/web/public/demo/still/PRODUCT-BRIEF.md"
    },
    {
      "step": 2,
      "stage": "Compiled workstreams",
      "fact": "The generated run prompt assigns site, film and CAD ownership before execution.",
      "url": "https://github.com/fraylabs/possible/blob/main/apps/web/public/demo/still/CODEX-THREAD.md#run-prompt"
    },
    {
      "step": 3,
      "stage": "Verification failure",
      "fact": "The first browser trace preserves the integrated site's asset-path 404s.",
      "url": "https://github.com/fraylabs/possible/blob/main/apps/web/public/demo/still/verification/browser-results-initial-failure.json"
    },
    {
      "step": 4,
      "stage": "Repair",
      "fact": "The fresh-review receipt records the relative-base fix and mandatory rerun.",
      "url": "https://github.com/fraylabs/possible/blob/main/apps/web/public/demo/still/evidence/final-receipt.md#material-failure-found-and-repaired"
    },
    {
      "step": 5,
      "stage": "Passing completion",
      "fact": "The outcome receipt reports the post-repair browser pass, 58/58 artifact audit and remaining limitations.",
      "url": "https://github.com/fraylabs/possible/blob/main/apps/web/public/demo/still/OUTCOME-RECEIPT.md"
    }
  ]
}