Opportunities: Wiki Knowledge Applicable to Localz — Incremental Pass, 2026-08-09
Scope: prior passes (08-07, 08-08) covered Airbnb's shared-core pattern, incident-response escalation, STRIDE, the distributed job scheduler, and search autocomplete. This pass checks Architecture/Fundamentals/infrastructure-as-code.md (new this window) against Localz's documented stack state.
Infrastructure as Code's "common pitfall" section is, almost word for word, a discrepancy already flagged in Localz.md
Localz.md (line ~312-319) documents a live contradiction: Localz-Required-Stack.md and Anti-Lockin-Pattern-for-Startup-Stack.md both recommend Clerk as the MVP auth choice behind an adapter interface, but Development-Setup-Resources-and-Capacity.md documents an actual running dev environment using Keycloak instead — described in Localz.md's own words as "a live discrepancy between the documented 'recommended' stack and the stack apparently already stood up in dev." This window's Infrastructure as Code note names this exact failure pattern directly: "allowing manual changes to slip in alongside an IaC-managed environment... the infrastructure now differs from what the state file believes is true... the team starts distrusting plan output entirely." Localz doesn't have a Terraform state file yet, but the underlying problem is identical at the documentation layer — the "recommended stack" docs are the desired-state definition, and the actual dev environment has drifted from it without a corresponding update, which is precisely the note's Principal Engineer Lens question: "can I reproduce this system's state from a known definition?" Right now, for Localz's auth layer, the answer is no — the definition and the running system disagree, and nobody has recorded which one is authoritative.
This resolves cleanly as a five-minute documentation fix, not a re-architecture: either (a) update Localz-Required-Stack.md to reflect that Keycloak is the actual current decision, or (b) tear down the Keycloak dev setup and stand up Clerk per the documented recommendation — either is fine, but leaving the drift unresolved is the specific state the IaC note warns compounds ("the team starts distrusting plan output entirely") once more decisions get built on top of an ambiguous foundation (e.g., the Auth Adapter interface's sync_user_profile() behavior likely differs meaningfully between Clerk and Keycloak).
Not yet applicable
- Consistent hashing, mutual TLS, service mesh — Localz's team size and current infrastructure scale don't yet justify any of these; consistent with prior passes' assessment for service mesh specifically.
- Team Topologies and Conway's Law — a genuine org-scale intervention; not applicable while Localz is a single-developer project.
- Data storage tiering — premature before Localz has production data volume to tier.