standup
Standup — 2026-06-02
**ADR-0025 shipped.** Free tier gets Claude Haiku, Pro gets Claude Sonnet — the assignment has lived in `TIERS` as an undocumented implementation detail since t
Daily standup for 2026-06-02. What shipped:
ADR-0025 shipped. Free tier gets Claude Haiku, Pro gets Claude Sonnet — the assignment has lived in TIERS as an undocumented implementation detail since the V1 cycle began. #160 pins it. The math: ~$1.07/Free user/month holds under three preconditions, one of which (infra at $0.50) is flagged optimistic — realistic GCP stack lands closer to $1.30, which shifts the break-even conversion floor from 5% to ~6%. The ADR writes a mandatory reopen trigger: 90 days post-GA, if conversion sits below 4% AND the Haiku-vs-Sonnet eval gap exceeds 20 points, a cost-vs-quality analysis goes to Blair with a specific lever proposal.
Dead code removed. #159 deleted the legacy /api/v1/internal/sync cron route — unscheduled, no callers, but publicly reachable because Cloud Run’s invoker is allUsers. Three follow-on PRs cleaned up stale references and corrected language in ADR-0022 that described the exposure window inaccurately (#161, #162, #163).
Infrastructure and observability fixes also landed: Cloud Run env var drift codified (#157), audit-vs-analytics sink scope corrected (#155).
On the KB side, the Executive Summary moat paragraph now scopes cross-pollination by version — global brain in V1, per-customer brains in V1.5, cross-pollination in V2+ — per ADR-0014 (#10). The standup word-count floor dropped from 60 to 10 so quiet days stop producing workflow failures instead of files (#9).
Notable. The deleted sync route (#159) was publicly reachable the entire time it existed — no auth dependency, allUsers invoker. It wasn’t scheduled or called, but it was there. ADR-0022’s OIDC pattern is specifically designed to prevent the next one.
Window: 2026-05-28T13:00
→ 2026-06-02T13:00.366Z · Sources: memberintel @ 5f1b71c, memberpress-intel @ b54a095