Team Loop — Cross-Model Routing & Comparison Console

Validated Evidence Automated check pass rates, wall-clock seconds, and token telemetry.

Subjective Scoring Reviewer-calibrated 5-dimension rubric totals (editable per model).

Warnings & Defects Non-fatal accessibility warnings or interaction defects flagged in QA.

Failure States Timeouts or unvalidated runs excluded from comparison candidates.

1. Seed Submissions & Side-by-Side Candidate Selector

Four seeded benchmark runs. Failed runs remain preserved in underlying telemetry but cannot be selected as comparison candidates.

Left Comparison Slot (Candidate A) [Primary Anchor]

Keyboard: Tab + Enter or Alt+S

Right Comparison Slot (Candidate B) [Challenger]

Record why a model justifies its latency/token cost or why its defects block production routing, then download a self-contained JSON audit record.

Routing Recommendation Verdict

Decision Notes (Evidence synthesis, trade-offs & routing rationale)

Ready to export local JSON record.

Live Decision Payload Preview (JSON)