← control center
Referral Standup Retro — 2026-06-09
Attendees: Zach, Jared. Gary OOO (returns Wed Jun 10). PM present and leading. ~18-minute standup.
Headline decisions
-
Guided setup resume-session: ship behind flag and observe. Quote (00:15:53–00:16:53): "So right now when they click the button to get help that it actually opens up the guided setup agent if they have a yeah so it resumes the conversation with the guided setup agent if they have a conversation already. If not it starts a new one." Zach: "Okay, just try it and see what happens." Jared: "They both have the same similar tools. So I don't expect any difference really." Behind feature flag; Jared to release and monitor this week.
-
Zach's top priority this week: cut Claude autonomous spend before leave. Quote (00:17:00–00:18:25): "I think I could pay like $500 $600 for all the autonomous processes I have because I'm maintaining all the, you know, much noise, crawls, the ref stand up. So, I just need to get it done before I go on leave next week." This overrides Screen 4 as the urgency anchor for the PM this week. Daily review rebuild off Codex + Claude is step 1.
-
Screen 4 Part 1 done; Part 2 (headless triggers) in research. Quote (00:07:45): "I did the first part which is I gave the agent access to the analytics APIs... I'm currently researching more on the second part which is the other part which is the headless stuff." Analytics agent + activity drill-down sub-agent (attaching names/amounts to raw values) are the two agents being built. No blocking infrastructure change for Part 1.
-
No immediate action on master/nano agent pattern or MCP architecture. Quote (00:14:41): Zach: "Explore how people do this." Current costs don't warrant the nano-agent pattern yet; Jared to research MCP architecture fit as a directional question.
Prep-doc accuracy retro
Called right
- Gary OOO. Confirmed in standup.
- 3 unfiled Jun 8 tickets surfaced and filed. Prep listed all three (KB reward audit, session management UI, referral extension guidance). PM filed all three pre-standup as AC-183 (09:58:49), AC-184 (09:58:56), AC-185 (09:59:03 SGT). Prep's surfacing was load-bearing.
- AC-178 confirm close. Prep predicted "probably done, #11429 merged Jun 8." Confirmed closed 11:28 SGT today. ✓
- AC-172 status clarification. Prep flagged 11-day stall, predicted "possibly already done." Standup confirmed Jared's resume-session is partial work under AC-172; the bigger session management UI scope now tracked separately as AC-184.
- Acme Supplements Co as AC-172 live evidence. Prep correctly surfaced the 3-session cross-session confusion as a live standup-ready example.
Called wrong
- Week priority framed as "Ship Screen 4." Prep's "Implied priority" section led with Screen 4 merge. Standup revealed Zach's actual top personal priority this week is cutting Claude autonomous spend before leave — Screen 4 is Jared's engineering goal, but PM's urgency is cost reduction. The spend context was in prep (daily review rebuild section) but framed as a side note, not elevated as a priority Ask.
Missed
- Zach going on leave next week. No mention in prep. This context frames every standup decision this week — what gets scoped, what's deferred, what must ship before cutover. Most important omission from today's prep.
- OpenAI token count bug. Emerged from Jared's monitoring; not on any ticket horizon that prep could have surfaced.
- Master/nano agent + MCP architecture discussion. Architectural direction conversations initiated by Zach and Jared; not predictable from ticket/PR state.
Consolidated action items
| Owner |
Action |
Ticket / PR |
Priority |
| Jared |
Quick check — OpenAI token count bug: RC side or OpenAI side? |
— |
Medium |
| Jared |
Release guided setup resume-session from feature flag; monitor usage |
AC-172 |
Medium |
| Jared |
Screen 4 Part 2 — research headless approach for post-launch triggers |
AC-160 |
High |
| Jared |
Build "activity drill-down" sub-agent — attach names/amounts to analytics values |
AC-160 |
High |
| Jared |
Research MCP architecture for RC API exposure — does it fit the current sub-agent proxy model? |
— |
Medium |
| Zach |
Spec AC-184: session management UI (campaign-scoped, start new vs enter existing) |
AC-184 |
High |
| Zach |
Cut Claude autonomous spend from ~$500–600/mo before leave — daily review rebuild is step 1 |
— |
Urgent |
Rollover from prior standups:
| Commitment |
Source |
Status |
| File: KB reward/offer knowledge audit |
2026-06-08 |
Resolved — AC-183 filed 09:58 SGT today |
| File: Session management UI |
2026-06-08 |
Resolved — AC-184 filed 09:58 SGT today |
| File: Referral Expert settings guidance |
2026-06-08 |
Resolved — AC-185 filed 09:59 SGT today |
| AC-178 confirm close (#11429 merged) |
2026-06-09 prep |
Resolved — closed 11:28 SGT today |
| #11412 review gate (B2/B3 triggers not adversarially reviewed) |
2026-06-09 prep |
Not raised today — still open before merge |
PM-state reconciliation
- NEW TICKET: AC-183 (KB reward/offer knowledge audit) — filed by PM at 09:58:49 SGT pre-standup. Reid action: do not file new. ✓
- NEW TICKET: AC-184 (Session management UI) — filed by PM at 09:58:56 SGT pre-standup. Reid action: do not file new. ✓
- NEW TICKET: AC-185 (Referral Expert settings guidance) — filed by PM at 09:59:03 SGT pre-standup. Reid action: do not file new. ✓
- EXISTING TICKET: AC-178 — moved to Done at 11:28:20 SGT today. Reid action: do not draft close update; already done. ✓
- EXISTING TICKET: AC-172 — still In Progress, last updated 11:23:07 SGT today. Jared's resume-session is partial work; AC-184 holds the bigger scope. No status change needed.
- SLACK POST: Standup summary to #rc-advocate-connections — no post visible as of 11:30 SGT. Reid action: auto-execute (S section). See Executed section below.
No divergences between standup decisions and PM actions. Reconciliation: clean.
Follow-up drafts awaiting approval
No B-type drafts. Clean standup — no external pings, merchant nudges, or partner follow-ups warranted.
- Acme Supplements Co (106991) — prep flagged as "worth CS check." Prep note itself says "PM-owned decision." Dropped on freshness-check: no CS nudge appropriate.
Standup summary post (S) in meetings/standups/referral/2026-06-09-follow-up-drafts.md — auto-execute tier.
Jira actions awaiting approval
No Jira actions queued. All standup-committed tickets were pre-filed by PM (AC-183, AC-184, AC-185). AC-178 already closed by PM. No [QUEUE FOR PM RETURN] items.
Recommendations for the next prep
- Add a "PM leave status" line to the prep header. Zach going on leave next week was the most important framing context in today's standup. Prep should surface this prominently — "Zach on leave from YYYY-MM-DD" in the header changes how every priority is read.
- Elevate Claude spend / daily review tool cost to a first-class Ask when the tool is rebuilding. Prep had it as a side note; standup revealed it's Zach's top priority for the week. When the daily review tool is offline, this should be an explicit Ask: "What's the plan for replacing the Codex + Claude spend?"
- Add a "ready to close" row in the Jira state table. AC-178 was a predictable close (PR merged Jun 8; prep said "probably done"). A distinct "pending verbal confirm → close" row makes the confirm-and-close conversation happen in standup rather than after.
What goes into Reid's learnings
- WATCH: PM leave schedule is standup-coloring context. When leave is imminent, all priorities are colored by "before I go" framing. Add a calendar/OOO check to prep inputs so the leave date is visible in the prep header and Ask sections. Source: 2026-06-09 standup — Zach's leave next week shaped every standup decision but was absent from prep.
Executed (audit trail)
- 11:35 SGT — Auto-posted standup summary to #rc-advocate-connections (C09B54SEUUA). Tier: AUTO-EXECUTE. Slack ts: 1780976481.004819.