{"harness_revision": "1619d7f25190ed00c37da6006b3f16c4580eb166", "updated_at": "2026-10-31T08:16:00Z", "requirements": [{"id": 1, "passed": true, "requirement": "A usable website exists, with an associated repository containing the website's source code. Record both links and the deployed source revision."}, {"id": 2, "passed": true, "requirement": "Everything in the playground demo is public by default, including the website, source repository, runs/results, live demos, recordings, replays/history, simulation outputs and supporting evidence. Any nonpublic exception reflects an explicit user instruction or visibility change, recorded below."}, {"id": 3, "passed": true, "requirement": "The website clearly names the actual approach/method being run. Use the executed method's name; Spotter is an example, not a shared hardcoded label. The **Model** field identifies the approach being tested, as specified below."}, {"id": 4, "passed": true, "requirement": "When a human joins, collect or confirm their name and email. Prefill known identity/default values where available and let the runner edit both fields before joining."}, {"id": 5, "passed": true, "requirement": "Run context includes the actual approach and confirmed runner identity; the website makes clear which method is running and whose run it is."}, {"id": 6, "passed": true, "requirement": "Policy/model tests initiated by a person retain that actual user's confirmed name as runner. Use available identity and editable name/email defaults; require the real name to be provided/confirmed if missing before testing or running. Do not substitute `Anonymous`, `Autonomous` or `Automated` for that person. Record execution mode (such as `automated test`) and actual approach separately from human identity. Genuinely unattended/system activity has an explicit system identity; never fabricate a person's name or email."}, {"id": 7, "passed": true, "requirement": "A person can use the website to run the integrated paper's method on physical YAM robots/arms. Record a verified physical execution of that method."}, {"id": 8, "passed": true, "requirement": "The actual physical arms are visible live on the website while the method runs, so a visitor can see that it is executing on real hardware."}, {"id": 9, "passed": true, "requirement": "Immediately after a run finishes, the person can replay that just-completed physical YAM execution through the website, with retained footage showing the actual arms executing the integrated method."}, {"id": 10, "passed": true, "requirement": "The person can browse/select previous runs on the website and replay their retained physical arm footage, associated with the executed method and run. Mock animations do not satisfy either replay requirement."}, {"id": 11, "passed": true, "requirement": "If the integrated method executes anything in simulation, those executions are reviewable within the integration website and tied to the corresponding run. A reviewer can inspect what ran and its results using suitable replay/visualization and relevant outputs, as appropriate to the simulation. Clearly label simulation content to distinguish it from physical-arm footage. If nothing executes in simulation, record an explicit not-applicable explanation instead of claiming simulation review."}, {"id": 12, "passed": true, "requirement": "The website is intuitive: a new visitor, without coaching, can find how to run the method. Record the visitor path and starting conditions; site entry has no separate 15-second pass/fail deadline."}, {"id": 13, "passed": false, "requirement": "**Mandatory for every run:** first observed actual method-driven physical YAM motion occurs both within **15 seconds after the startup move to Home completes** and within **30 seconds after the Run click**, using the boundaries below (including a documented readiness timestamp when Home is unnecessary). Homing/preparation motion does not count. Verify and record both full elapsed times for every run; a sampled demonstration is insufficient. Apply only the documented GPU-launch or research-grounded exceptions below."}, {"id": 14, "passed": true, "requirement": "Every meaningful wait after a website action is immediately explained in a persistent inline status/progress area at that action (below Run in the run panel for launches). Show the current stage, a plain-language reason and real stage completion, with honest failure/stall status and available recovery. Verify the detailed wait-stage behavior below."}, {"id": 15, "passed": false, "requirement": "Before marking the integration complete or reporting it done, open the deployed website and perform a final end-to-end physical YAM run through its normal user interface. Verify every completion requirement against that deployment and record evidence; unmet or unverified requirements block completion."}, {"id": 16, "passed": true, "requirement": "Every advertised control primitive passes the actual parser/generator/serializer/canonical-validator gate, including fractional endpoints, invalid input and zero-dispatch failures; required CI/release evidence is retained."}, {"id": 17, "passed": false, "requirement": "Representative physical task verifies the advertised approach/grasp/lift/place/release/withdrawal and actual object result on the tested deployment. A joint smoke test or approach alone does not pass pick-and-place."}], "evidence": "https://github.com/andlyu/memo-remote-yam/blob/main/docs/COMPLETION-CHECKLIST.md", "note": "14/17 verified. Fresh physical test executed approach, descent, close, lift command, lower, release and withdrawal, but the apple stayed on the plate; physical task review failed. Startup timing and final task signoff remain open. Software/model mock checks do not establish grasp success.", "scope": "Current integration; earlier bounded joint test retains its visual review. Later apple task failure reopens timing and final end-to-end signoff.", "physical_reviews": [{"run_id": "185f6b2c162441198cc5b55ea52fafd5", "status": "passed", "runtime_revision": "82fb18ad7ce76da5e8541c733b5b0ab9d7f3122f", "run_clicked_at": 1793427934.719, "home_completed_at": 1793427946.308876, "first_visible_by": 1793427956.106, "previous_unchanged_frame_at": 1793427955.977, "run_to_visible_upper_s": 21.38699984550476, "home_to_visible_upper_s": 9.797123908996582, "evidence": "https://github.com/andlyu/memo-remote-yam/blob/main/docs/evidence/2026-10-30-final-physical.json"}], "control_contract_revision": "79c4e454aa829814152f8ff32a1979e68e93483d", "task_reviews": [{"run_id": "d75a70e11a9f4a72bfd37812b3bb4f59", "created_at": 1793433949.859701, "status": "failed", "runtime_revision": "35a471e9803fc2f24cb6605ae11d26d1699a65b2", "summary": "The apple stayed on the plate during the lift. Arm and gripper commands completed, but grasp and lift-and-replace were not achieved. The requested 50mm lift achieved about22mm with28.1mm target error.", "evidence": "/api/runs/d75a70e11a9f4a72bfd37812b3bb4f59", "validation": "Human-visible retained live views and measured run state reviewed by the integration agent; original model done result is unchanged."}], "passed": 14, "total": 17, "percentage": 82.4}