Conversion model: public Lab + methodology → inbound submit-for-eval → scoped private evaluation → branded report.
Public intake: marketing site #contact — company, artifact type, timeline, context. 48-hour acknowledgment SLA on every submission.
Operational detail (ack copy, reply conversion, scope sheets, survival lane) stays in founder-held materials — not linked from this room.
What the customer gets
- Numeric report with defined pass/fail thresholds
- Shown failure cases (not only passes)
- Episode counts, seeds, variance / confidence intervals where applicable
- Explicit sim-only limitations stated alongside results
- Optional tier comparison (mild / moderate / severe) for policy runs
- Branded delivery — engagement metadata + methodology appendix wrapping the technical findings
Motions
| Motion | Scope | Typical inputs |
|---|---|---|
| Policy stress evaluation | Pillar 1 — adversarial sim tiers on agreed tasks | Policy checkpoint or agreed baseline controller; task list |
| World-model physics QA | Pillar 2 — consistency checkers on generated video / trajectories | Video clips or trajectory dumps; tracking assumptions documented |
| Public-artifact paid audit (survival) | Pillar 2 — fixed-scope ≤6 public clips, ≤5 business days, short branded report | Public clips + provenance/permission; frozen detectors; fee gated on payment path |
| Continuous scoring | Phase 5 roadmap — release-pipeline integration | API / CI hook; versioned artifacts |
Engagement shape
- Intake — artifact type, timeline, company (site form; 48h ack)
- Scope call — tasks, episode budget, success criteria (scope sheet locked)
- Run — reproducible config logged in core (customer-private under
results/private/) - Report — branded markdown + engagement.json (findings + failure cases + limitations + methodology appendix)
- Optional follow-on — re-run on next policy version or expanded task set
Format dry-runs (not customers): three reference reports from public evidence — DP-REF-001 Lift, DP-REF-002 Stack, DP-REF-003 CogVideoX — validate the delivery template before live partner artifacts arrive.
Demand signal (async diligence): every site-form inbound and scope call is logged on the design-partner pipeline (≥10 conversations or ≥3 serious threads — strengthens pitch; does not block soft raise).
Public methods: Methodology report. Pricing posture: Pricing hypotheses. Product framing: Product overview. Specs: Core specs.