Measured harness ledgerPublic result
Grok 4.6ECHO Main Menu — Grok 4.6 xhigh
Recreate the pinned ECHO-style cinematic main menu reference as an interactive browser experience.
xhigh reasoningHeadline result
- Workflow cost
- $2.71
- Wall-clock
- 1h 09m 03s wall-clock
- Processed tokens
- 20.04M processed
- Record state
- verified_exact_metrics_runtime_replay_qualified_protocol_deviation_no_blind_evaluation
Public summary
Grok 4.6 xhigh verified_exact_metrics_runtime_replay_qualified_protocol_deviation_no_blind_evaluation ledger: 1h 09m 03s wall-clock, 20.04M processed, and $2.71 Exact final-segment CLI receipt plus deterministic reconstruction from the native per-inference ledger for the interrupted segments.
Run identity and stack
- Result ID: echo-main-menu-browser-grok-4.6-xhigh
- Technical model: grok-4.6
- Provider: xAI via grok-sub
- Stack: xAI via grok-sub
- Stack: Technical model/configuration: grok-4.6
- Stack: Browser recreation
- Stack: Pinned visual references
- Stack: Harness v1 ECHO prompt
Primary artifact integrity
- Kind: interactive-browser-title-menu
- Path: artifacts/echo-main-menu-browser-grok-4.6-xhigh/source/index.html
- SHA-256: b2a46457eabf789114150415c6b37ade55f622ba8a2cf1646d47783f1b242052
Validation evidence
- Result: FUNCTIONAL_PASS_WITH_QUALIFICATION
Recorded caveats
- The run is not one runner process: a local one-hour watchdog interrupted active work, and the preserved Grok session was continued until it stopped itself.
- Wall-clock time includes model reasoning, tool execution, asset generation, installation, rendering, visual iteration, and verification; it is not model-only compute time.
- Reasoning tokens are reported separately but are already included in provider-accounted output tokens, so they are not added again to the total.
- The 2.712231 USD total combines one exactly recorded final segment with ledger reconstruction for the interrupted segments using the CLI's per-call rates and long-context rule.
- Independent replay verifies runtime behavior on one browser and machine, not visual-fidelity scoring or cross-device performance.
- A trustworthy numeric FPS value was not captured; continuous motion was verified through changing live frames instead.
- The initial runner imposed a 3600-second local watchdog and exited while Grok was still working. The preserved conversation and workspace were continued. A short intermediate continuation was interrupted only to remove the practical cutoff; the final continuation ran until Grok returned end_turn.
Visible evidence gaps
- blind visual-fidelity evaluation record
- trustworthy local FPS measurement
Public result only
This result keeps its public summary and evidence, but it does not currently have a matching Builder test with prompts, projects, Harness workflows or skills.