Morgan Chen

Sofia closed the loop on the one-page customer-language addendum. It’s enough for her internal post-close read, and she’s not asking for a data-room refresh before her own board catch-up.

Board pre-read is out before noon. Devon did the last Monday pass, cut the over-explained caveat sentence, and kept the Mercury and Nadia paragraphs bounded.

Can you compare the invited-member handoff copy/support sequence and give me the smallest edits before this goes through truth-check? Invited-member handoff packet — Apr 8 review post Figma comments — Priya P1 — pending-member row Surface - Status label: Invite sent - Workspace chip: Treasury Ops - Helper text: Waiting on admin / first source to finish acceptance. Priya note - This still makes admin and first source sound like competing authorities. Can we make the source workspace clearer without making the row louder? P2 — acceptance screen header Surface - Top-left header may display the user's last active workspace. - Banner text: You've been invited to join Evergreen Bank - Body text: Accepting this invite adds you to the workspace selected by the invite. Priya note - Header can be true-to-session but confusing; copy should tell the user the invite is the source for membership, not the current header. P3 — helper-text candidate Candidate copy - You may see your last active workspace until acceptance completes. This invite adds you to Treasury Ops. Priya note - Maybe too specific for generic product copy; support can name the workspace. Support sequence draft — Jake Working use - Evergreen invited-member handoff / acceptance-header confusion Draft sequence 1. Verify the backend membership grant and the intended workspace before replying. 2. If the grant is attached to the intended workspace and only the header reflects the user's last active workspace, do not call it a permissions failure. 3. Support should say the header can reflect the user's last active workspace until acceptance completes. 4. Do not advise revoke/resend unless the membership record is absent or the grant is attached to the wrong workspace. 5. If the membership record is absent, escalate in the normal Mercury truth-check lane before customer wording goes out.

Sarah and Devon’s skeleton is below. Please trim and order it into the bounded pre-start packet HR can format; cut anything that starts to look like a strategy workstream. Subject: Nadia pre-start packet skeleton Date: Mon, Apr 8, 2024 From: Sarah Kim To: Morgan Chen Cc: Devon Hayes, HR Morgan — Below is a first pass at a bounded pre-start packet for Nadia. Goal is reading/context only before Apr 15, not a new strategy project. HR can format once you decide what actually belongs. Pre-start packet skeleton — Nadia Singh Purpose - Give Nadia the existing Mercury and Evergreen materials we already rely on. - Show the current post-raise owner handoffs before she inherits any live context. - Keep the packet light enough that it orients her without implying new customer promises, pre-start customer work, or a future-team design exercise. Section 1 — Mercury cohort package Include - existing retention/cohort read - current admin-controls learning - caveats from the post-raise operating reset Working note - This section should orient her to the evidence base we already trust rather than ask her to reopen every metric. - I would keep the reset caveats with it so the data read stays separated from release notes and enterprise-readiness packaging. Section 2 — Evergreen current-flow examples Include - invited-member handoff screenshots - admin empty-state examples - SSO/auditability wording notes - support-routing examples already in use Working note - This should stay in present-state language only. - The useful thing here is seeing how current product truth has been kept separate from later commercial/procurement packaging. Section 3 — Post-raise owner split Current owner handoffs to show explicitly - Mercury execution and launch evidence — Jake - Platform seams and release discipline — Leo Park - Activation/onboarding UX — Priya - Retention evidence — Anna Martinez - Enterprise-readiness commercial packaging — Devon Hayes with Sarah Kim - Evergreen external thread continuity / current-state materials — Sarah Kim - Board narrative and hiring sequence — Morgan Chen Working note - I would spell out the current handoffs across Mercury, Evergreen, support, product, and customer-growth-adjacent work so Nadia sees the existing boundaries before she changes anything. Section 4 — First operating questions Draft list 1. What repeatable customer signal exists already? 2. Where is Mercury evidence clean enough to reuse? 3. Which Evergreen examples are operating evidence versus roadmap pressure? 4. What handoffs are already working? 5. What should Nadia avoid changing before she has first-hand context? Add-on A — Customer-call list - Suggested early customer meetings - Note: could be useful if Morgan wants momentum before start Add-on B — Customer-growth recruiting debrief placeholder - Possible discussion of future team shape Sarah note - HR can format once Morgan decides what belongs in the pre-start packet. Devon note - This should stay light; I included the add-ons only so we can explicitly cut them if they create the wrong motion.

Acme billing dashboard is fixed: Jake corrected the display cache, Greg’s finance view now shows a single April total, and the Stripe ledger was single-charge the whole time.

First replay after the workspace-claim mapper patch: user-update passed, deletion replay still has one normalized log line emitting a missing workspace claim.

Family-call plan moved to Sunday afternoon. Mom wanted both of us on a weekend call, Maya can’t do Saturday, and I’m keeping it off the work calendar.

Small Kibo drama by Lake Merritt: the leash clip snapped, Jamie had the spare in the bag, and the walk kept going.

Mercury truth-check comments — Apr 9 on Jake's Evergreen support note Comment 1 - On draft line: This may be a permissions mismatch. - Returned comment: Do not use permissions mismatch unless the backend grant is wrong. In the reproduced case, the backend grant stayed attached to the intended workspace. Comment 2 - On draft line: The user landed in the wrong workspace. - Returned comment: More accurate: the acceptance header can reflect the user's last active workspace until acceptance completes. That header symptom is not the same as the membership grant. Comment 3 - On draft instruction: Customer may revoke and resend if worried. - Returned comment: Avoid revoke/resend unless the membership record is absent or attached to the wrong workspace; otherwise resending adds noise. Comment 4 - Suggested customer-safe wording: We checked the invite path and the membership grant remains tied to the workspace selected by the invite. The header can briefly show the user's last active workspace before acceptance completes, so support should verify the membership record rather than treating the header alone as a permissions issue. Comment 5 - Support-sequence note: Keep backend verification first, then customer explanation; no product roadmap language. Turn these into the concise support note Jake can use — accurate, customer-safe, and not overly backend-y.

Tuesday Figma pass: Priya kept the calmer pending-member copy, tightened the “waiting on admin” helper text, and swapped the overloaded “first source” label for plainer source-workspace wording for the next pass.

Leo found one recovered-state live-sync wording diff in the support explanation. The replay didn’t show a platform/state mismatch, so he’s keeping it out of today’s product changes while invited-member handoff work continues.

Pinecone sent the retry transcript. I need bounded technical answers against the current API v2 GraphQL/JWT behavior; don’t promise token rotation or new connector semantics. Subject: Re: Pinecone connector verification docs wording Date: Tue, Apr 9, 2024 From: Pinecone To: Morgan Chen Hi — Following up on the API v2 retry questions, we ran a short connector-side spike against the current GraphQL/JWT path and captured the cases below. Sharing the transcript so we can make our retry behavior line up with the supported path. Case 1 — paginated GraphQL retry after 401 - Request: query workspaceNodes(first: 50, after: cursor_page_1) - Attempt 1: JWT expired during the first attempt - Response 1: 401 token expired; no page returned - Client action: refreshed credentials and replayed the same page-two query - Response 2: returned node IDs [node_881, node_882, node_883, node_884] and cursor cursor_page_2_done - Pinecone note: same IDs as prior controlled run, no duplicate node observed Case 2 — expired JWT before write - Mutation: upsertConnectorCheckpoint(workspace: wrk_test, seq: 42) - Request auth state: expired JWT - Response: 401 before resolver - Follow-up read: checkpoint remained seq 41 - Pinecone note: appears no side effect before auth Case 3 — attempted token-rotation header - Request included header: X-Scaffold-Token-Rotation: rotate-if-valid - Token type: workspace-scoped token - Response: ignored header and processed the normal read path - Pinecone note: current docs do not mention this header; is in-place rotation supported or should we reconnect? If it helps, we can send raw request/response logs, but the three cases above are the ones we need to align on for the connector spike. Thanks, Pinecone Attached internal scope note - Answer only from the current API v2 quickstart and supported connector/auth model. - Quickstart reference: https://docs.atlas-test.com/api/v2/graphql-quickstart - Do not promise a new token-rotation endpoint, SSO/admin expansion, or broader connector-header behavior.

Please edit the remaining overclaim in this Kestrel excerpt and leave the developer-facing example intact. From: Kara <kara@kestrel-test.com> To: Morgan Chen <morgan@atlas-test.com> Cc: Sarah Kim <sarah@atlas-test.com> Date: Tue, Apr 9, 2024 10:14 AM PT Subject: Re: Mercury newsletter draft - three lines that may be too strong Morgan, Sarah - I took the note from last week and swapped in one developer-facing example so the piece is less abstract. Updated excerpt below. A practical Mercury example A developer or support lead should be able to see who resent an invite, which workspace the invite came from, and whether sync recovered before support escalates. That kind of trace keeps the conversation grounded in what happened instead of forcing teams to guess from a screenshot. Enterprise-grade controls are arriving in Q3 so teams can treat these admin paths as launch-ready across larger accounts. I think the example is more concrete now, but I left one stronger Q3 sentence in case you want the piece to land with more confidence. Kara

Devon cleaned up the April finance forecast after the backfill spike fell out of the AWS run-rate view. The stale appendix cell no longer makes the board materials look like infra spend jumped.

Rishi’s rough bullets are pasted here. Before he turns them into runbook text, flag what’s missing or in the wrong lane — especially support wording vs implementation detail. Discord DM - Morgan Chen <-> Rishi Continued thread: Atlas coverage follow-up Tue Apr 9, 2024 8:21 AM Rishi: Promised rough bullets for the two runbook holes. Pasting as-is before I polish. AT-04 replay/backfill behavior - Replay jobs are safe to rerun when the source event ID is stable. - Backfill should start from the last confirmed checkpoint, not from the first failed customer-visible row. - Support line: tell customers replay is automatic after transient failures. - Implementation note: replay uses verifier seam from PR-1187 and can show duplicate internal attempts while emitting one customer-visible state. - Missing? Need example from API v2 docs/search support path. AT-05 workspace/org-invite migration history - Old org-invite records can point at workspace IDs that were renamed during migration. - Do not assume the invite email header is authoritative. - Customer-facing note: ask weekly support owner to verify membership record before escalating to Atlas. - Implementation note: migration stored prior org label for audit trail, but support should not expose raw migration label. - Need to mention Jake only for time-sensitive product priority? Not sure. These are intentionally rough; I know they still mix support language and implementation notes.

BART held Jamie long enough to miss the normal dog-walk window, so I took Kibo out between calls and left the late work call camera-off. Very glamorous ops day.

Trader Joe’s briefly showed a duplicate pending charge. Jamie got the reversal today, and the actual grocery order was unaffected.

Devon and Sarah have the lighter Nadia packet basically in place. Please do one final pass on the five operating questions so they feel like orientation, not homework before Apr 15. Subject: Re: Nadia pre-start packet skeleton Date: Wed, Apr 10, 2024 From: Devon Hayes To: Morgan Chen, Sarah Kim Cc: HR Quick markup below. I cut the pieces that create pre-start motion or make it sound like Nadia should start doing customer work before Apr 15. Nadia pre-start reading packet — markup Keep Section 1 — Mercury cohort package - existing retention/cohort read - current admin-controls learning - caveats from the post-raise operating reset Section 2 — Evergreen current-flow examples - current-flow examples already in use - invited-member handoff screenshots - admin empty-state examples - SSO/auditability wording notes - support-routing examples already in use Section 3 — Post-raise owner split - current owner handoffs for Mercury, Evergreen, support, product, and customer-growth-adjacent work Section 4 — First operating questions 1. What customer evidence is already strong enough to reuse without new calls? 2. Where are Mercury and Evergreen signals telling us the same thing, and where are they different? 3. Which owner handoffs should stay untouched during the first two weeks? 4. What support or product examples are useful evidence, not immediate asks? 5. What would you want to learn before proposing any customer-growth motion? Cut - Customer-call list - Customer-growth recruiting debrief placeholder Devon note - Cut customer-call list and recruiting placeholder. These questions are the only thing that still risks sounding like homework. --- From: Sarah Kim To: Morgan Chen, Devon Hayes Cc: HR Date: Wed, Apr 10, 2024 Works for me. If Morgan is comfortable with the wording, HR can format this lighter version.

Sarah moved Nadia’s tentative Mercury/Evergreen context hold off the quarterly all-hands placeholder. It’s still tentative while the packet gets finalized.

The deletion replay issue is narrower now. Jordan traced the remaining mismatch to Scaffold’s normalization path dropping archived workspace metadata; live create and update paths are still clean in the replay read.

After the updated header copy, Evergreen walked through the invited-member acceptance screen. They still dislike one source-workspace phrase, but the conversation stayed about display clarity instead of permission risk.

From: Board To: Morgan Chen, Devon Hayes Date: Wed, Apr 10, 2024 9:06 AM PT Subject: Re: board pre-read Morgan, Devon - Thanks for sending the pre-read. Four follow-up questions so we can keep the discussion tight: Q1: How much sample size sits behind the week-two retained-activity drop, and should we treat it as signal or noise? Q2: Is Evergreen blocking launch, or is Evergreen informing support/product readiness in the normal course? Q3: How bounded is Nadia's start really if the company is already talking about customer-growth motion? Q4: Does the pricing/package caveat change the B-round narrative you shared after the close, or is it still a product/packaging caveat rather than fundraising risk? Board Morgan margin note: Answer within the pre-read; do not rewrite the note or turn this into a new memo. Please prep concise answers for these four board follow-ups. Keep them inside the pre-read framing — no board-note rewrite and no new memo.

Jake renamed the low-risk Mercury cleanup ticket in Linear because the old title made it look release-blocking. Leo confirmed the new title matches the actual risk and doesn’t require a code change.

Mom’s phone/form thing got unstuck. Maya handled the call with her while I decoded the error screenshot between meetings; switching browsers opened it.

This morning’s Lake Merritt loop became a half-loop. Rain picked up, Kibo refused the back half, and Jamie met us near home with a towel.

Invited-member handoff pass is closed. Priya and Jake finished the copy/support sequence: calmer pending-member surface, tighter waiting-on-admin helper text, and source-workspace wording instead of the overloaded first-source label. Jake’s note went through the normal Mercury truth-check lane. Leo’s Thursday replay across the Evergreen scenarios matched backend workspace grants to UI state with no platform/state mismatch. I’m closing the special support-priority watch back into normal support monitoring. Live-sync trust stays stable as a bounded green watch item from the Feb 14 timeout-recovery follow-up, and the tiny recovered-state wording diff is support-explanation only if it stays tiny and isolated. No sync-engine, retry-policy, auth/admin-role, or enterprise-readiness scope in current Mercury release language.

Sofia debriefed after her internal board chat: the customer-pull evidence landed as credible, and there isn’t another diligence artifact ask this week.

Kara accepted the edit removing the remaining “enterprise-grade controls arriving in Q3” line. The developer-facing example is still in the newsletter draft.

Can you draft a short correction request to Blueline asking them to remove the emergency-rate surcharge? Keep it factual and small; I don’t want to turn the whole repair into a dispute. From: Blueline Plumbing To: Morgan Chen <morgan@atlas-test.com> Date: Thu, Apr 11, 2024 9:38 AM PT Subject: Invoice for Apr 4 service visit Hi Morgan, Attached is the invoice for the Apr 4 repair visit. Invoice summary Service date: Apr 4, 2024 Service window: 12:00-2:00 PM Location: Oakland | Line item | Amount | | --- | ---: | | Service visit, weekday shifted window, Apr 4, 12:00-2:00 PM | $185 | | Parts and standard repair labor | $240 | | Emergency-rate surcharge | $95 | | Subtotal | $520 | Thank you, Blueline Plumbing Morgan note: The visit was the Thursday weekday window they moved us into; it was not same-day emergency service. I want the surcharge removed, not a whole fight about the repair.

Greg asked if Acme’s invoice export had the same duplicate-row issue as the dashboard. Jake checked the Stripe-backed ledger path and answered that the export wasn’t affected.

HR’s formatted draft briefly called it “Customer Growth launch plan,” and Sarah corrected it to “Pre-start reading packet.” Sarah and Devon finished the lightweight version I wanted: existing Mercury cohort package, Evergreen current-flow examples, post-raise owner split, and a short list of first operating questions. The packet frames Nadia as starting from existing customer evidence and current owner handoffs, not as a pre-start strategy project. HR is holding the bounded packet for her Apr 15 start.

Pinecone has the bounded API v2 retry answers now. They said it’s enough for the connector spike and didn’t push for a token-rotation promise.

Auth replay patch is green in staging: update and deletion replay cases both passed, and Jordan’s Honeycomb traces no longer show the replay path producing the workspace-claim mismatch.

Devon and I sent the board follow-up answers. We kept the week-two activity weakness visible, framed Evergreen as operating evidence, and left the pre-read alone.

I owe Jamie one: I left my card at Blue Bottle after coffee, and he picked it up while walking Kibo. It’s back in my wallet before the weekend.

Kibo picked up a burr in his paw after the Lake Merritt walk. Jamie got it out at home, and Kibo was walking normally again by dinner.

Anna’s Sunday refresh is below. Please turn it into a short Monday internal readout that separates the real weak signal from the instrumentation row that’s still under check. From: Anna Martinez To: Morgan Chen, Devon Hayes Date: Sun, Apr 14, 2024 4:27 PM PT Subject: Mercury metrics refresh - Sunday check Morgan / Devon - Sunday refresh below. Clean rows are still holding after the cleanup pass. | Row | Metric | Looker | Mixpanel | Status | Note | | --- | --- | ---: | ---: | --- | --- | | 1 | activation_start | 82% | 82% | matched after cleanup | event-name fix still holding. | | 2 | admin_first_action | 61% | 60% | matched within rounding | no caveat needed. | | 3 | invite_sent_week1 | 48% | 48% | matched | stable. | | 4 | sync_enabled_week1 | 37% | 37% | matched | dedupe still clean. | | 5 | retained_activity_week2 | 31% | 24% | real weak signal still present | lower activity for teams with unresolved admin setup; not instrumentation. | | 6 | repeat_admin_action_week3 | 18% | 16% | narrowed but unresolved | workspace filter fix reduced gap, still not fully cleared. | For Monday, please keep row 5 as the real product/activation weakness and row 6 as instrumentation still under check. - Anna

Sarah has Nadia’s packet and welcome Zoom details queued for tomorrow morning. Nadia replied with a quick thanks and isn’t asking for more pre-start material tonight.

Sunday family call half-worked: Mom’s connection dropped about twenty minutes in, but Maya texted me the missing update afterward. At least it isn’t rolling into Monday.

Nadia started today at Scaffold as Head of Customer Growth, reporting to me. I framed her first 90 days as building a repeatable customer-growth motion from Mercury activation evidence, Evergreen-style enterprise-readiness signals, expansion failure modes, and the Sarah/Devon handoffs — not brokering bespoke roadmap exceptions or running a generic sales kickoff. Draft concise answers to these margin questions in that lane; don’t turn it into a broader strategy assignment. From: Nadia Singh To: Morgan Chen Date: Mon, Apr 15, 2024 Subject: Questions on the pre-start packet I left a few margin comments in the reading packet and copied them below. Section 1 — Mercury cohort package [Comment] Which Mercury activation evidence is already clean enough to reuse internally? Section 2 — Evergreen current-flow examples [Comment] Should Evergreen’s admin-control examples be treated primarily as buyer-readiness evidence or product-quality evidence? Section 3 — Post-raise owner split [Comment] Where should Sarah’s live-account continuity notes stop, so I do not accidentally become the account owner? [Comment] How does Devon want pricing/package caveats represented when we talk about customer pull? Section 4 — First operating questions [Comment] Which handoffs should remain untouched during my first two weeks?