# Atlas whole-corpus launch checklist

Status: human review required

Purpose

Search any platform in the corpus and show the documented first-mile path from account creation to first developer success. LLM APIs are one cohort inside the corpus, not the product boundary.

Per-platform review

- Identity resolved to the current official platform and organization.
- First-party source set reviewed, current, and sufficient for the selected route.
- Starting boundary is account creation unless an explicit comparable exception is approved.
- First-success boundary has an official label or observable terminal state.
- Atomic selected route includes every required action from account creation through first success.
- Account, app, OAuth or credential, scope, redirect URI, request, and verification fields are inventoried when documented.
- Decision nodes list documented options, selected option, and effect on the first-success route.
- External gates, unavoidable waits, automatic platform outcomes, and recovery states are attached to the exact route node.
- Every prerequisite, node, field, option, edge, gate, branch, wait, and first-success claim has accepted source evidence.
- Documented structural complexity profile is generated from the verified route.
- Compatible-domain comparison is shown only after at least three qualified peers pass the same gates.
- Potential friction is labeled as documented or potential, never as observed drop-off without behavioral evidence.
- Durable direct link verified in the deployed environment.
- Correction path opens with platform context, without automatic submission.
- Publication decision recorded.

Current state

- Corpus records: 237.
- Public routes: 1.
- Qualified comparison cohorts: 0.
- Routes with at least three qualified peers: 0.
- Render is the only currently public route.
- Remaining corpus records stay searchable as known platforms but non-public until the review gates above pass.

Required pilot evidence

- At least 20 representative people complete the route-understanding flow.
- At least 5 people copy a platform share link.
- At least 3 recipients open a shared link and view a second platform.
- At least 2 useful corrections are started.
- Route-accuracy and user-value protocols in `evaluation/product-validation-protocols.json` pass their predeclared thresholds.

Local implementation does not satisfy human review, production deployment, or pilot validation by itself.