Morgan Chen

Sarah checked the live Evergreen thread after Nadia’s read and kept the next customer reply under Sarah’s name. Nadia only got the short internal pattern note from the thread, not the external response lane.

Jake found two Friday support tickets that fit Nadia’s new admin-handoff bucket, but he kept the response route with the weekly support owner and marked them that way so the taxonomy doesn’t confuse live support ownership.

Rishi confirmed the next Tessl technical conversation is set for the following Tuesday, and he saved this week’s Atlas runbook artifacts before leaving for the weekend. No new process ask came with it.

Kibo picked up a foxtail in the outer fur of one paw at Lake Merritt. Jamie got it out before it irritated the skin, and he was walking normally again by afternoon.

I cleaned up the Friday customer-growth notes and moved the roadmap-ish section under examples-to-watch. Devon still flagged “repeatable motion” and “expansion path” as too sales-forecasty. Draft safer replacements that keep the recap close to signal buckets, evidence, caveats, and owner routes.

Mother’s Day settled into a short call: Maya found three photo options in an old text thread, I picked the least blurry one from my phone, and the call had Mom, Maya, me, Jamie, plus a brief Kibo cameo.

Anna’s Sunday refresh is the first Monday follow-up after Nadia’s taxonomy read. Prep the Monday readout so the metrics map to the buckets without making the taxonomy sound more certain than the evidence supports. From: Anna Martinez To: Morgan Chen, Devon Hayes, Nadia Singh Date: Sun, May 12, 2024 4:36 PM PT Subject: Mercury metrics refresh - Sunday check Morgan / Devon / Nadia - Sunday refresh below for the first Monday follow-up after Nadia’s taxonomy read. Core rows are effectively unchanged from the Friday cut; I added Nadia’s bucket column so the weekly read maps cleanly to the new customer-growth categories. | Row | Metric | Looker | Mixpanel | Nadia bucket | Status | Note | | --- | --- | ---: | ---: | --- | --- | --- | | 1 | activation_start | 83% | 83% | activation friction | matched | corrected activation definition still holding. | | 2 | admin_first_action | 62% | 61% | admin handoff / role confusion | matched within rounding | steady. | | 3 | invite_sent_week1 | 49% | 49% | admin handoff / role confusion | held | stable supporting-activity row. | | 4 | sync_enabled_week1 | 38% | 38% | activation friction | held | stable. | | 5 | retained_activity_week2 | 32% | 26% | activation friction | real weak signal still present | concentrated in workspaces with unresolved admin setup; not expansion evidence. | | 6 | repeat_admin_action_week3 | 18% | 18% | expansion stalls after first live sync | matched | clean this week; no active stall signal in this row. | Row 5 split carried forward after Friday’s taxonomy cleanup | Segment | Looker | Mixpanel | Bucket | Read | | --- | ---: | ---: | --- | --- | | All Mercury workspaces | 32% | 26% | activation friction | reference row above | | Workspaces with unresolved admin setup | 18% | 12% | activation friction | weakness remains concentrated here | | Workspaces with admin setup resolved by day 7 | 44% | 39% | activation friction | not showing the same drop | Notes for Monday - This remains the weekly directional internal cut, not the quarterly cohort / board layer. - I mapped retained_activity_week2 to activation friction, not expansion stalls. The Friday label correction carries through here. - This table only covers the activation / admin / post-sync behavior buckets. Procurement / security packaging and true product-gap reads still need account-level notes rather than inference from this cut alone. - Bucket labels here are for internal pattern capture; they do not change current account, commercial, or product ownership. - Anna

Nadia’s Monday questions are about routing defaults, not reopening the bucket list. From: Nadia Singh To: Morgan Chen Date: Sun, May 12, 2024 Subject: Monday follow-up after Friday's taxonomy read Morgan — I think the bucket set from Friday is right. The remaining ambiguity for me is about default routing, not about widening the read. 1) Admin handoff / role confusion Do you want a support-owner subroute when the row is a real repeated interpretation problem but the live response still belongs to weekly support? My worry is that otherwise the bucket starts to look like customer-growth owns the fix. 2) Procurement/security packaging What is the cleanest way to show repeated asks without turning that section into a sales forecast? I am leaning toward pattern + caveat + owner route only, with current external-safe wording boundaries kept explicit and no forecast language. 3) True product gaps Should the default be that a named product owner is present or has reviewed the row before it appears in the weekly read? I want to avoid a bucket of real issues that has no product route attached. If useful, I can mark up next Friday's template with those defaults rather than reopen the bucket list itself. — Nadia Draft concise answers for her. Keep the next read useful, but preserve Sarah/support, Devon/commercial, and product-owner lanes where they belong.

Rishi’s note for Tuesday’s Tessl call is below. Discord DM — Morgan Chen ↔ Rishi Continued thread: Atlas coverage follow-up Sun May 12, 2024 10:26 AM Rishi: Quick heads-up before Tuesday’s Tessl call. Still no process ask from them. This looks like another technical conversation, not anything broader. What I expect to spend time on: - replay evidence: what I trust before I call recovery clean, especially the case where the source event is stable but derived state was only partially written - schema-change safety: checkpoint choice, lock/rollback posture, and how conservative I am about backfills under load - Atlas incidents I can describe without customer details: replay/backfill recovery patterns, verifier-seam weirdness, and how I separated customer-visible recovery from internal retry noise I’m keeping it high level on customer specifics and not bringing live account details into the conversation. Atlas runbook review is still separate and on my calendar. No new ask from me — just wanted you to have the shape before Tuesday. Give him brief boundary guidance on Atlas examples he can discuss at a technical level without exposing customer-specific details.

BART has a weekend alert for Monday morning single-tracking from Oakland into SF. I told Devon I’ll start Monday remote if the delay holds rather than risk missing the first customer-growth follow-up block.

Can you message Devon that the BART single-tracking alert is still active, so I’m starting remote from Oakland. My first in-person window should be treated as tentative, not late.

Anna and Nadia got through the first Monday pass on yesterday’s Mercury metrics refresh. Nadia is okay with the caveats for now: retained_activity_week2 stays tied to activation friction, and we’re not moving the Friday bucket. We also settled the routing defaults off that conversation: Sarah keeps continuity on active customer threads, Devon takes the commercial/procurement frame, Anna owns data caveats, Jake makes the product-priority calls, and anything urgent still routes from the support rotation doc.

Rishi’s Tessl technical with Anna Rao happened. His read was basically “still serious,” but there’s no offer yet — they told him they expect to decide later this week.

Kara says the bounded answer on the organic broad-ship question went out cleanly. They accepted the no-date version, stayed on the newsletter list, and didn’t pull us into roadmap or funding.

The Atlas runbook review with Rishi, Leo, and Jake surfaced the shape I was worried about: verifier/auth is written well enough for someone else to follow, but replay/backfill recovery and the workspace/org-invite migration steps are still too much “ask Rishi.”

Priya moved Evergreen’s accepted audit and SSO wording into the next Figma copy pass. Sarah, Devon, and Nadia now have enterprise-readiness language they can use without implying feature dates.

Marcus and Leo caught two Mercury launch-readiness tickets that were green in Linear but didn’t have Honeycomb evidence attached. Holding the #eng-releases note until the proof is linked and the claims check out.

Anna Rao called: Tessl is making Rishi a real database-internals/platform offer. Rishi says he’s leaning toward accepting but wants the weekend to sit with it, and he hasn’t given notice. I’m not starting a counteroffer circus. Please ask Rishi for transition impact by Atlas path, using the ownership map and the support/code-path notes he already wrote down, especially the areas that still depend on him. Also privately tell Devon that Anna’s intro may cost us someone we rely on more than I’d been admitting.

Trader Joe’s substituted in poultry-based dog treats. Jamie caught it before Kibo got any, and we kept the bag out of the house instead of pretending it was close enough.

Rishi still hasn’t given notice. He did send the Atlas impact pass I wanted, path by path: Leo needs to shadow the technical flows, Jake is where the product-priority calls land, and the support rotation doc should stay the customer-facing anchor.

Finance wrinkle caught before it escaped: the Stripe renewal recon export showed one customer with a duplicate entitlement extension from BillingOrchestrator. billing-service is corrected now, before the month-end Google Sheets finance view went out.

Rishi accepted Tessl’s database-internals/platform offer and gave notice. His last Scaffold working day is June 21, 2024; he’s taking a short break after that and then joining Anna Rao’s team at Tessl. So the Atlas work is now a real handoff plan through June 21: finish AT-03 verifier/auth, write AT-04 replay/backfill recovery, write AT-05 workspace/org-invite migration, have Leo shadow the technical paths, keep Jake on product-priority calls, and keep customer-facing ownership in the support rotation doc. Also, any Atlas API v2 docs/search support should go through the current GraphQL API v2 path, not directly to Rishi.

AWS cost alert was real but contained: a Mercury canary was pushing too much Honeycomb trace export volume because debug sampling got left high. Jordan and Marcus traced it and dropped the sampling rate back down before it touched finance reporting.

Good first test of the new Atlas routing: one support item still went straight to Rishi, but we moved it back through the support rotation doc. Leo handled the first technical read on the GraphQL API v2 path, and Jake made the priority call.

Warm Oakland morning, so Jamie and I took Kibo around Lake Merritt earlier than usual. We’re keeping the rest of the holiday quiet — Blue Bottle stop, then home, not turning it into another errand loop.

Sarah heard back from Evergreen: procurement/security have passed the examples pack around internally and want a next-step review window. We don’t have a July slot or agenda locked yet.

Please schedule the Evergreen enterprise-readiness review for July 11, 2024, 10:00–11:00 AM Pacific. Scaffold side is Sarah, Devon, Nadia, and me. Keep the calendar description date-free and centered on current-state evidence, SSO/SAML evaluation framing, admin-side audit history, admin vs. billing-owner separation, and what a standalone procurement/security packet would need to include. Role boundaries: Sarah keeps the customer thread, Devon owns the commercial/procurement framing without feature-date promises, and Nadia listens for repeatable customer-growth patterns.

Good, Greg confirmed Acme is fine with Monday being live-only under the NDA scope. They’re not expecting fresh Mercury examples, screenshots, packet excerpts, or a sanitized one-pager before the call.

Sarah and I had the live-only Acme NDA-scope discussion with Greg Shipman and Acme’s product lead. Greg backed off the broadest version of Acme’s claim, but the remaining disputed-materials question is now two categories for counsel to review, not solved. For now the boundary holds: Acme can use the public generic Mercury page and we can keep talking live, but we should not send fresh Mercury examples, screenshots, packet excerpts, customer-specific briefing material, sanitized one-pagers, or other written Mercury material until the narrowed issue is resolved. Please keep treating the pause as legal-scope driven, not as us cooling on Acme.

Pinecone caught a retry example that still sounded like the old API v1 route. Sarah and Jake fixed the partner guidance so it stays in the current Go connector world: GraphQL API v2 plus JWT.

Please save this in Kibo’s sitter notes so future sitters don’t improvise on food or treats. After-Visit Summary — Kibo Visit date: June 7, 2024 Location: Oakland Patient: Kibo Present with: Morgan Chen and Jamie Visit type: Routine check Reason for visit Routine check-in for Kibo before summer weather settles in. Summary from today Kibo is stable overall at this routine visit. The main items discussed were: - Mild weight creep. - Some fatigue in warmer weather. - Keeping his food routine simple and measured. - Preserving the strict no-poultry food rule. - Shifting walks earlier on hot days. - Planning a routine dental follow-up in fall 2024. Food and treats Keep Kibo’s food boring and consistent. Do not improvise with new foods or rich treats. Measure his regular food rather than free-pouring or adding extras. Avoid extra snacks unless Morgan and Jamie have specifically set them out for Kibo. Hard no: Kibo must not receive poultry-based food or treats. For Kibo, poultry includes chicken, turkey, duck, poultry meal, chicken meal, or similar label ingredients. If a label contains chicken, turkey, duck, poultry meal, chicken meal, or anything similar, do not give it to him. Skip it and ask Morgan instead of guessing. Warm-weather activity Kibo has had some fatigue in warmer weather. On hot days, move walks earlier when it is cooler and keep the rest of the day lower-key rather than adding extra activity in the heat. Weight note Kibo has mild weight creep. The plan is to keep portions measured, keep food consistent, and avoid added treats or substitutions. Dental follow-up Plan a routine dental follow-up in fall 2024. Caregiver/sitter note If someone else is watching Kibo, give them the food rule exactly as written: No poultry-based treats or food for Kibo. Chicken meal counts as poultry. If a label has chicken, turkey, duck, poultry meal, chicken meal, or similar, skip it and ask Morgan instead of guessing. Plan until next routine follow-up - Keep food simple, consistent, and measured. - No poultry-based food or treats under any circumstances. - Move walks earlier on hot days. - Keep warmer days quiet if Kibo seems tired from the heat. - Schedule routine dental follow-up for fall 2024.

Clerk’s deprecation notice hit that older session helper, but the only remaining use is one internal TypeScript test harness. Jordan and Marcus checked it against the completed API v2 auth rewrite and it’s not production-impacting, so this is just test cleanup.

Please tell HR to keep the customer-growth org shape quiet for now. Nadia’s 60-day read is that the cadence is clearly useful, but still too manual; she wants more reliable account-note inputs and cleaner source material before we add GTM headcount, not a junior team to manage. Devon and I are not opening RevOps, CS, SDR, or field-sales lanes in Q3. We’ll revisit after Nadia has two more cadence cycles and the July Evergreen review gives us better evidence.

Atlas got a small but useful live test today. The low-sev issue did not boomerang to Rishi by default: support rotation had the customer-facing owner, Leo took the first technical pass, Jake made the sequencing call, and Rishi was just there to correct if needed.

Please turn Sofia’s email and my working notes into the plain operating-read section for the Q2 board update. Bullets are fine; keep it factual and don’t make it sound like we’re accelerating hiring, fully covered on Atlas, or declaring Mercury launch risk cleared. From: Sofia Alvarez To: Morgan Chen Date: Tue, Jun 18, 2024, 7:42 AM Subject: Q2 board update — plain operating read before the board note Morgan — before you turn the Q2 board update into polished board language, can you send me the plain operating read on three things? I’m not looking for the investor-story version yet. I want the factual version you are using internally, including the uncomfortable parts and what evidence changed your confidence. 1) Hiring pace after the Series B - Did the Series B actually change the pace or shape of hiring, or did you mostly keep the same evidence-led posture? - I’m especially interested in whether the customer-growth hire created a new org/hiring pull, and whether you are opening more GTM or Mercury engineering lanes in Q3. - Please distinguish between “we hired the planned leader” and “we are accelerating headcount because we raised.” 2) Atlas risk after Rishi’s departure - Now that Rishi has accepted Tessl, what is the real Atlas continuity risk? - What is covered by Leo/Jake/support rotation, and what still depends too much on Rishi? - If there has been any live test of the interim split, include that rather than only the transition plan. - Please be explicit about dates — I know his last day has not happened yet. 3) Mercury release discipline and confidence - Has the Mercury discipline actually improved release confidence, or are people just marking things green earlier? - I’d like the raw operating evidence: release checks, observability evidence, incidents/anomalies, and what the team did when the evidence was missing. - Please keep this separate from the larger retention/customer-growth story unless the evidence really connects. If it is easier, send bullets under those headings. I’ll read it as operating evidence, not as proposed board copy. Sofia --- Morgan Chen working notes for Sofia request Created: Jun 18, 2024, 9:06 AM Last edited: Jun 18, 2024, 10:18 AM Purpose: raw Q2 operating evidence for Sofia before drafting the board note. These are factual inputs only, not board language. Cutoff: Jun 18, 2024. Rishi’s final Scaffold working day is scheduled for Jun 21, 2024. Atlas handoff is active and not complete. 1) Hiring pace after the Series B — raw evidence Series B / operating posture - Feb 22 Series B closed with Northstar as lead. Board milestones after the close have stayed centered on defensible retention evidence and repeatable enterprise expansion from Mercury/Evergreen learning, not on immediately launching a Series C narrative. - Post-raise operating ownership was set without a broad hiring acceleration: Jake runs Mercury execution and launch evidence; Leo Park owns platform seams and release discipline; Priya owns activation/onboarding UX; Anna Martinez owns retention evidence; Devon Hayes owns enterprise-readiness commercial packaging with Sarah Kim; Morgan holds board narrative and hiring sequence. - Post-raise hiring sequence decision: next new leadership search was customer-growth/GTM leader focused on repeatable enterprise expansion from Mercury and Evergreen learning. The second Mercury engineering req was not reopened; later hiring remained evidence-led. - Existing Mercury engineering headcount plan still has the second Mercury engineering hire on hold. Revisit trigger is external launch-readiness feedback from design-partner conversations/trials, not an arbitrary calendar date. Decision rule still depends on whether blockers look like senior platform/architecture seams, implementation throughput, or manageable current-team follow-through. Customer-growth leader - Apr 15 Nadia Singh started as Head of Customer Growth. Reporting line: Morgan. First-90-day scope: build repeatable customer-growth motion around Mercury activation, Evergreen-style enterprise-readiness signals, expansion failure modes, and handoffs with Sarah Kim and Devon Hayes. - Nadia’s role was not defined as bespoke enterprise roadmap brokerage or a generic sales kickoff function. - Initial cadence during first 90 days stayed internal: Sarah brings live-account continuity and customer-thread notes; Devon brings commercial/procurement packaging context; Nadia turns repeated Mercury/Evergreen patterns into a weekly customer-growth read. - Apr 29 Nadia was deliberately looped into the Evergreen Bank thread only for customer-growth pattern capture and repeatability work. Sarah remained live customer-thread continuity owner; Devon remained commercial/procurement owner. Nadia’s involvement did not change product facts, create a new shipped-product owner story, or authorize roadmap promises. Customer-growth evidence quality / org pull - May customer-growth read classified Mercury/Evergreen signals into buckets: activation friction; admin handoff/role confusion; procurement/security packaging; expansion stalls after first live sync; true product gaps. Each bucket is supposed to keep an owner route attached so customer-growth does not absorb every customer-adjacent issue by default. - Jun 14 roughly-60-day read with Nadia: cadence is clearly useful but still manual. - Jun 14 Nadia asked for more reliable account-note inputs and cleaner source material before any new GTM headcount, not a junior team to manage immediately. - Jun 14 Morgan and Devon decided not to open a RevOps, CS, SDR, or field-sales hiring lane in Q3. - Jun 14 HR was told to keep the customer-growth org shape quiet until Nadia has two more cadence cycles and the July Evergreen review produces better evidence. Current hiring facts as of Jun 18 - One planned leadership hire completed: Nadia as Head of Customer Growth. - No second Mercury engineering req reopened. - No Q3 RevOps, CS, SDR, or field-sales lane opened. - Customer-growth cadence is producing useful pattern reads, but the input system is still manual and not yet clean enough to justify another GTM layer. - Hiring sequence remains held by Morgan with Devon/HR drafting or routing only inside the small reset group unless Morgan reviews. Open evidence gaps / caveats for hiring pace - Nadia’s weekly read is useful but depends on account-note quality from Sarah/Devon/live threads. - Need two more Nadia cadence cycles plus July Evergreen review before changing the customer-growth org shape. - Mercury/Evergreen expansion repeatability evidence is still being built; not enough to treat customer-growth as a scaled team yet. 2) Atlas risk after Rishi’s departure — raw evidence Pre-notice coverage map - Rishi had already written down Atlas ownership map and support/code paths that still depend on him before the Tessl offer became concrete. - Existing routing rule before notice: urgent Atlas issues land in support with the named weekly owner and an immediate next step first. Rishi pulled only for verifier or PR-1187 issues, replay/backfill recovery and idempotency, legacy workspace/org-invite migration issues, or Atlas-specific replay/verifier behavior in deploy or observability triage. - Leo Park can provide first-pass seam read for auth/platform-boundary issues. - Jake is fallback for time-sensitive Atlas product-priority, sequencing, and cut decisions if Rishi is unavailable. - API v2 docs/search support should use the current GraphQL API v2 path rather than becoming a direct route to Rishi. May runbook review - May 15 separate Atlas runbook review happened with Rishi, Leo, and Jake. - Confirmed verifier/auth coverage was readable. - Confirmed replay/backfill recovery and workspace/org-invite migration steps still depended too much on Rishi’s memory. - The biggest named Atlas knowledge pockets remained AT-04 replay/backfill and AT-05 workspace/org-invite migration history. Tessl offer / notice sequence - May 17 Anna Rao called Morgan after Tessl decided to make Rishi a concrete database-internals/platform offer. - May 17 Rishi told Morgan he was leaning toward accepting but wanted the weekend to sit with it. - May 17 Morgan did not launch a counteroffer circus; she asked Rishi for transition impact by Atlas path and privately told Devon the intro might cost Scaffold someone it still relied on more than Morgan had admitted. - May 20 Rishi had not yet given notice but gave Morgan a path-by-path Atlas transition impact note. The note identified Leo’s shadowing needs, Jake’s decision points, and the support rotation doc as the customer-facing anchor. - May 22 Rishi accepted Tessl’s database-internals/platform offer and gave Morgan notice. - May 22 final Scaffold working day set for Jun 21, 2024, with a short break before joining Anna Rao’s Tessl team. Active transition plan after notice - Through Jun 21 transition plan is: - finish AT-03 verifier/auth note; - write AT-04 replay/backfill recovery steps; - write AT-05 workspace/org-invite migration notes; - have Leo shadow the technical paths; - keep Jake on product-priority decisions; - keep customer-facing ownership in the support rotation doc; - route Atlas API v2 docs/search support through current GraphQL API v2 path rather than directly to Rishi. Live routing evidence since notice - May 24 Atlas support item initially routed straight to Rishi after his notice was redirected through the support rotation doc. - May 24 Leo handled the first technical read on the GraphQL API v2 path. - May 24 Jake made the priority call. - Jun 17 low-severity Atlas issue during Rishi’s final week served as a live test of the interim split. - Jun 17 support rotation doc named the customer-facing owner. - Jun 17 Leo handled the first technical pass. - Jun 17 Jake made the sequencing decision. - Jun 17 Rishi stayed available for correction. Current Atlas risk read as of Jun 18 - Rishi has accepted Tessl offer and is in notice period; final day has not happened yet. - The interim split has been exercised on at least one low-severity issue, not a severe outage or high-pressure customer escalation. - Covered enough to route: support rotation doc for customer-facing owner; Leo for first technical pass on platform/auth/API v2 seams; Jake for sequencing/product-priority decisions. - Still exposed: AT-04 replay/backfill recovery and idempotency history; AT-05 workspace/org-invite migration history; any Atlas-specific replay/verifier behavior in deploy or observability triage where written steps are incomplete. - Rishi remains available for correction through Jun 21, but relying on that availability is not a completed handoff. 3) Mercury release discipline and confidence — raw evidence Ownership / release discipline setup - Post-Series-B operating ownership assigns Jake to Mercury execution and launch evidence. - Leo Park owns platform seams and release discipline. - Priya owns activation/onboarding UX. - Anna Martinez owns retention evidence. - Devon Hayes owns enterprise-readiness commercial packaging with Sarah Kim. - Morgan holds board narrative and hiring sequence. Release-readiness evidence behavior - May 16 Marcus and Leo found two Mercury launch-readiness tickets marked green in Linear without linked Honeycomb evidence. - May 16 they held the #eng-releases note until the Honeycomb evidence was attached and the claims could be verified. - The useful operating signal from May 16 is that green Linear status was not treated as sufficient when evidence links were missing. - The gap exposed on May 16 is that ticket status could move ahead of observability proof unless Leo/Marcus checked it. Observability / infrastructure hygiene evidence - May 23 AWS cost anomaly pointed to elevated Honeycomb trace export volume from a Mercury canary. - May 23 Jordan and Marcus traced the cause to debug sampling left high. - May 23 Jordan and Marcus brought the sampling rate back down before it affected finance reporting. - This was not a customer-facing Mercury incident in the notes available here; it was an observability/cost hygiene issue from canary trace export volume. Auth/API cleanup evidence relevant to release confidence - Jun 11 Clerk posted a deprecation notice for an older session helper used in one internal TypeScript test harness. - Jun 11 Jordan and Marcus confirmed the completed API v2 auth rewrite was not production-impacted. - Jun 11 change scope stayed limited to test cleanup. Release-risk readout artifact available to Leo/team - Existing document: “Mercury release-risk readout — Leo template.” - Template tells team to keep readouts short and structured, not broad platform memos. - Required fields in the template: top risks; evidence source; severity/confidence; next verification; explicit exclusions. - Evidence sources named in template: dogfood account, staging pass, screenshot, log, bug-bash finding, Figma state/frame, or checklist row. - Template explicitly separates repeated evidence from one-off demos/screenshots and calls out where evidence is still thin. - Template excludes abstract platform recommendations, broad architecture memos, rewrite plans, and generic platform wishlists unless there is direct Mercury launch impact. Customer-growth / retention boundary for Mercury evidence - Nadia’s customer-growth cadence uses Mercury/Evergreen patterns, but release confidence evidence should stay separate from broader retention story unless directly connected. - Current Mercury/Evergreen signal buckets from Nadia’s read: activation friction; admin handoff/role confusion; procurement/security packaging; expansion stalls after first live sync; true product gaps. - Owner routing remains important so release discipline does not become the catch-all for every customer-growth issue. Current Mercury discipline facts as of Jun 18 - Evidence checks caught at least two launch-readiness tickets where status was green without linked Honeycomb proof. - #eng-releases note was held until evidence was attached and verified. - Honeycomb sampling anomaly was detected and corrected before finance reporting impact. - Clerk deprecation notice was checked against production API v2 auth rewrite and scoped to internal TypeScript test cleanup. - Release discipline evidence available here is operational and artifact-based: Linear status, Honeycomb evidence links/traces, #eng-releases hold, AWS anomaly, canary sampling correction, API v2 auth impact check. Open evidence gaps / caveats for Mercury release discipline - May 16 showed discipline working at review time, but also showed the workflow can still mark tickets green before evidence is linked. - May 23 showed observability coverage caught the Honeycomb trace-volume issue, but also showed debug sampling could be left high in a canary. - Jun 11 auth check was cleanup/non-production impact, not proof of a broader Mercury launch pass. - No single Q2 artifact in these notes says “all Mercury launch risks cleared.” The evidence is a set of operating checks and corrections through Jun 18.

Good release-discipline signal from Mercury: Leo kept the admin-side audit-history item out of the broader cut because the Honeycomb traces still weren’t convincing. Jake accepted the slip instead of trying to push it through #eng-releases anyway.

Please tell the team today’s Atlas handoff is good enough to operate, not proof that we’ve replaced Rishi’s depth. Rishi’s final Scaffold day is today. He’s leaving for Tessl to join Anna Rao’s database-internals/platform team after delivering the PR-1187 verifier note, a short AT-04 replay/backfill recovery runbook, and the cleaned AT-05 workspace/org-invite migration note. After today he is no longer Scaffold’s Atlas owner or the default Atlas customer-thread contact. Interim split: Leo owns first-pass technical work for verifier/auth, replay/backfill, and auth/platform-boundary reads; Jake owns Atlas product-priority, sequencing, and cut decisions; urgent customer-facing Atlas issues still start with the support rotation doc, the named weekly owner, and an immediate next step first.

Q2 board update is done. I was pretty direct about the tradeoffs: Nadia’s customer-growth cadence is working, but it is not a team-scale signal yet; Leo’s release-risk reviews are making Mercury more disciplined; and Rishi leaving for Tessl exposed a real Atlas gap that the interim split can manage but not erase. I told the board we are not opening another broad leadership search, field-sales build, RevOps/CS lane, or second Mercury engineering req in Q3. The next quarter is for Nadia’s evidence, Evergreen’s July review, Mercury release discipline, and Atlas transition stability. Hiring after that still needs evidence, not just the Series B close as the trigger.

Small finance cleanup: the late-June AWS forecast is back under last month’s run-rate now that the Honeycomb sampling correction is in. Devon and I left the Google Sheets finance view as-is — no new infra-cost exception needed.

Jamie got out early enough that we kept tonight simple: usual Lemongrass order, a short sit near Lake Merritt with Kibo while it was still light, then home before the lake got crowded. Nice to have the week end as a normal Oakland evening instead of another work spillover night.