Consolidated findings register
AI simulationv0.3
~110 findings from 12 role runs (S0–S8), consolidated by theme. Individual files carry the
full entries; this is the classification. Applied = fixed in keystone on readability
(commits 8bbe8ee, db7da1b, 2c88669, d5c977b). Candidate = drafted in
NORMATIVE-CANDIDATES.md, awaiting the maintainer’s decision. Deferred = recorded in
keystone’s BACKLOG/Roadmap with evidence. By design = defensible scope choice, recorded.
Themes (multiple independent role hits)
| # | Theme | Hits (roles) | Disposition |
|---|---|---|---|
| T1 | Conditions have no machinery — every real decision was “with conditions”; no field, store, tracking or escalation existed; 3 of 11 acceptance conditions silently died (one → the 02:07 page, one → the un-drilled successor) | S2b-7, S3a-8, S4-9/10, S5C-1/4, S6-4, S7-4, S8-4 (6 roles) | Applied (form fields, FSD-PRO “accepted with conditions” + tracking language) + Candidate N3 (the tracking shall) |
| T2 | Continuity proven once, never re-tested; no owner-transition trigger — the standard re-tests backups, never people; nothing fires between “owner named” (GV-2) and “leaver’s access revoked” (SA-2) | S4-6, S7-2/3/5/12/13, S8-5/6 (3 roles, incl. the successor’s “one change” verdict) | Applied (SAC decay note, cross-training sufficiency prompt) + Candidate N1 (owner-transition shall — the headline) |
| T3 | No service levels, no reporting to customers — only normative SLA mention is withdrawing them at retirement; a customer-audit request bounced through four people | S5-SDM-1/4, S6-5/6, S5-PM-4, S3b-5, S8-2 (4 roles) | Applied (support-model publication prompts) + Candidate N2 (the BACKLOG’s own SLM/SR thin shalls, promoted) |
| T4 | Decision rationale has no surviving home — RAIDD dies at delivery; the standing decision log existed only because an approver invented DEC7 | S3a-1/2, S4-7, S7-1 (3 roles) | Applied (RAIDD transfer note, per-service-docs store incl. “HLD carrying decisions”, evaluation-evidence home) + Candidate N5 (SAC Documentation row wording) |
| T5 | What a signature commits is stated nowhere — three approvers independently invented the answer | S2b-1/2, S3b-1, S5C-5 (2 roles, 3 gates) | Applied (commitment statements per gate in FSD-PRO + form preambles) |
| T6 | Role gaps: sponsor, Operator-in-process, first line, design authority, product/value owner, review independence, approver risk tiers | S0-2/3, S1-3, S2b-3, S3a-3/4, S5-SDM-3, S5-SD-1, S5-PM-1, S8-7 (7 roles) | Applied (Operator defined; escalation-line + alternate-approver stand-up prompts; ratify-above-author) + Deferred (5 new calls on the Tier 3 roles guidepost — the standard’s own chosen vehicle) |
| T7 | Documents: existence mandated, content not; SOP circular; docs homeless; operator-executability unasked; bespoke-code maintainability unasked | S4-1/2/3, S5-SD-4/5, S7-8 (3 roles) | Applied (SOP fixed; per-service-docs store; dated operator acceptance) + Deferred (content-guidance appendix → BACKLOG, to ride with the v0.4 worked example) |
| T8 | Portfolio/WIP promise has no mechanism at the gate | S2b-4, S3b-2, S5-PM-2/3 (3 roles) | Applied (decide-with-register-open; enhancement queue named in Information Stores) |
| T9 | Ongoing people-cost invisible — run-cost read as cash everywhere | S3b-3, S4-8 (2 roles) | Applied (operating-effort line on FRM-02 §7.1; FRM-03 cost row “cash and operating effort”) |
| T10 | GV-5 review: no per-service scope, no inputs, no independence, no missed-review tripwire — sat behind the entire failure curve, then failed silently itself | S8-1/2/3/7, S3b-5 (2 roles) | Applied (Implementation Guide step 7 agenda + hold-to-date) + Candidate N4 (note-level per-service scope) |
Smaller items — all applied
Status enum vs §8 outcomes; per-dependency §6 due diligence; related-demands-may-share-a-brief; forms point at Definitions; Solution Owner confirmed at sign-off; raise-the-risk-record hint; Quickstart timing honesty; design-principles-or-security-baseline reword; ratified-with-this- decision row; T-shirt anchors; adoption record named; no-central-severity-scheme line; monitoring/support-hours coherence prompt; Owner-column semantics; cite-don’t-copy; FSD-RR-7 “not retire — routed to the front door” phrasing → left for the maintainer (edits the Requirements Note column; see NORMATIVE-CANDIDATES.md postscript).
By design (defensible, now recorded)
- No CRM / customer-satisfaction measurement (declared non-goal; PM findings accept the boundary, the miss is the reporting half → N2).
- Verbal per-team promises, operational folklore — no framework captures these (S7’s “just life” column).
- The build itself hands to FSD-CH — the seam held (S4-5, praise).
The praise column (what demonstrably worked)
The SAC baseline mechanism (5 roles; “the most valuable hour of the workshop”, “made the decision near-mechanical”); FSD-RR-6 incident profile (the desk’s best tool — two mechanical severity calls, a pre-outage Sev2); the tested-restore and end-to-end-alert evidence bars (each caught a real defect pre-live); FSD-SC-4’s exit-path question (paid out in full — the 2027 supplier exit was a config swap because 2026’s due diligence forced DEC2); the gate chain (decidable, defensible, “heard and funded” delivered); the four Gate 1 outcomes + PoC box; the driver rule protecting risk cases; §4 cross-process prompts; the builder rule; the scope test; “don’t design anything yet”; Definitions; the seven-step stand-up; the “material change” review arm; and — the succession verdict — a record set that carried a no-handover takeover “far better than any real-world inheritance”, credited to the standard.