Morgan Chen

Use Kara’s note to settle the wrapper. Pick or lightly edit one subject line + preview text so it stays developer-facing and bounded, then I can let Kestrel lock it. From: Kara <kara@kestrel-test.com> To: Morgan Chen <morgan@atlas-test.com> Cc: Sarah Kim <sarah@atlas-test.com> Date: Thu, May 2, 2024 9:08 AM PT Subject: Re: Mercury newsletter draft - three lines that may be too strong Morgan, Sarah — Per the last pass, I left the body copy, screenshot choice, and caption where we landed them. I only need the subject-line / preview-text wrapper before I queue the send. Options 1) Subject: What we’re learning from Mercury admin workflows Preview text: A practical look at invite source, resend behavior, and support handoff context in the current Mercury flow. 2) Subject: Inside one Mercury admin workflow Preview text: One current example of how Mercury makes invite source, resend behavior, and support context easier to read. 3) Subject: A smaller look at launch readiness Preview text: A tighter read on one Mercury admin path, using the current developer-facing example from the newsletter. My own bias: option 3 may still sound bigger than the approved copy even with “smaller” in front. If you want to stay fully inside the bounded lane, 1 or 2 feels safer. Happy to lock whichever lane you prefer and stop touching the top line after that. Kara Kestrel Marketing On Fri, Apr 26, 2024 at 11:41 AM Kara <kara@kestrel-test.com> wrote: > Morgan, Sarah — > > Sending the Mercury newsletter proof package for next week’s pass. This version keeps the current Figma screenshot from the source-workspace frame, uses the safer developer-facing headline, and stays off exact Evergreen / Acme pull-quotes.

Can you give me the #eng-releases version of Jordan’s auth replay fix note? It should say update and deletion replays are clean, archived-workspace metadata is preserved, and Honeycomb shows no JWT metadata mismatch. Cut the Clerk archived-user ordering sentence if it makes this sound like a broad auth rewrite health statement.

Support briefly pulled the stale internal_search_v2 retry snippet for Pinecone, but caught it before anything went out. Current quickstart wording stayed the source of truth for the reply.

For the Founders Fund working session logistics, write the shortest reply confirming we can add a Zoom backup link for the two possibly remote attendees. Keep the Embarcadero Tower meeting format unchanged and don’t change the Tava Kitchen headcount.

Acme’s May statement preview looks settled. Greg sent Jake the screenshot after the footer wording cleanup; Jake confirmed the visible total is coming from the Stripe-backed ledger path and now reads as a statement total. Greg does not need another finance explainer.

Monday read bundle: From: Nadia Singh To: Morgan Chen Cc: Sarah Kim, Devon Hayes Date: Fri, May 3, 2024 4:18 PM PT Subject: Monday customer-growth read — prep bundle v1 Morgan — Pasting the workable bundle for Monday so we are not pulling from four half-separate notes. Keeping it internal and pattern-capture only. Sarah stays live-thread continuity owner on Evergreen; Devon stays commercial/procurement caveat owner; nothing below changes current product facts or external routing. Two rows still need Devon’s caveat before we use them in a general internal read: Evergreen audit-language sensitivity and Acme visible-number trust. --- 1) First Evergreen scratchpad --- A. Admin-role clarity vs support wording - The recurring question is not the role label by itself; it is what state or authority the label is being asked to carry. - “Admin” becomes acceptable when resend/remove powers are visible enough on the same surface. - If those powers are not visible, reviewers start inferring too much from the role name alone. B. Audit-export labels as procurement-sensitive - Label choice moves into contract/reporting inference faster than current product truth. - “View audit events” is better than “Export audit log,” but it still triggers a CSV/download expectation. - Useful as a buyer-readiness wording signal, not as evidence of current audit-history capability. C. Source-workspace copy as a reusable example - This is the cleanest reusable example from the thread so far. - Better copy made the current product behavior legible without changing the backend behavior. - Once the support wording tightened, the conversation moved from “permissions risk” to “display clarity,” which is exactly the kind of reusable pattern we should keep. D. Stale visible admin state as a trust pattern - When a visible admin surface lags a real action, reviewers treat that lag as trust debt even when backend state is correct. - That makes it useful as a visible-state trust example. - It should not be described as wrong membership data or a permissions problem unless the underlying record is actually wrong. E. Sarah continuity details to keep separate from pattern notes - Exact Evergreen reviewer routing, present-state wording already committed in-thread, and account-specific follow-up sequencing stay with Sarah’s live-account notes. - Those are useful for continuity but should not be blended into a generalized pattern row. --- 2) Devon proof-versus-pressure grid --- | Example | Devon read | Useful internal use | Guardrail | | --- | --- | --- | --- | | Evergreen admin controls | Buyer-readiness friction around wording, visible state, and role clarity inside the current Mercury flow. | Valid repeated signal for the internal read. | Not proof of enterprise readiness, not a launch claim, not a reason to imply SSO/audit/admin-model commitments. | | Acme visible-number trust | Finance-confidence context from a live NDA-constrained thread. | Valid as narrow internal context only. | Not Mercury activation proof, not safe to externalize, and not raw material for written Mercury examples while the NDA-scope boundary is unresolved. | | Pinecone retry semantics | Support/process clarity on the current API v2 path and retry explanation lane. | Useful as a support/process example if we need a non-Evergreen comparison row. | Not enterprise-readiness proof and not a broader connector-market story. | Devon note from the side thread - Evergreen admin-control language can enter the read only with the caveat that wording sensitivity is not the same thing as present capability. - Acme visible-number trust should stay narrow enough that it does not start reading like cross-customer proof. --- 3) Nadia + Sarah walkthrough — Evergreen thread-history groups --- | Group | What went here | How to use it | | --- | --- | --- | | customer asked live | who can resend or remove a pending invite; what “pending” or “invite sent” implies; whether source-workspace/header mismatch is display confusion or permissions risk; whether audit wording implies export/reporting; why member count can look stale after cancel | Keeps the original customer question intact. Good for reconstructing what was actually being asked before we classify it. | | support wording reusable | backend membership grant stays tied to the workspace selected by the invite; header can briefly reflect the user’s last active workspace until acceptance completes; do not advise revoke/resend unless the membership record is absent or attached to the wrong workspace; stale member count after cancel is display refresh/admin-panel cache, not wrong membership data | Candidate source for internal wording and support macros. Reusable because it stays in current-product truth. | | procurement-sensitive phrase | “Export audit log”; “SSO required”; “enterprise-ready admin controls”; “compliance-ready” | Evidence that certain phrases pull the reader into contractual/policy inference faster than the product truth supports. | | metric needed before claim | repeated visible-state trust issues across more than one clean example; support volume specifically tied to resend/remove authority; retained_activity_week2 movement after admin copy changes; recurrence of stale member-count or two-tab state confusion outside isolated testing | Do not upgrade these into broad patterns until we have more than one clean example or a metric that stays stable long enough to trust. | | do not externalize | direct Acme finance-language phrasing; internal owner-lane notes; any wording that makes the CSV question sound like an existing artifact; anything that turns Evergreen’s present-state thread into regulated-bank proof; any wording that implies a standalone procurement/security packet already exists now | Internal only. This bucket is mostly about preventing leakage from a useful note into the wrong audience. | Walkthrough note - Sarah’s live-account continuity details stay separate from pattern capture. - The goal is not to make Evergreen “the owner story.” The goal is to harvest repeated interpretation problems without changing external routing. --- 4) Candidate rows for Monday’s internal customer-growth read --- | Repeated pattern | Evidence source | Concrete example | Proposed caveat | Current owner lane | Needs Devon caveat before use | | --- | --- | --- | --- | --- | --- | | Evergreen audit language as buyer-readiness friction | Evergreen wording follow-up; procurement matrix on Apr 29; Priya’s copy-pass review notes | “View audit events” lands better than “Export audit log,” but still prompts reviewers to ask whether a CSV/report exists behind the label. | Wording sensitivity is real, but it is not evidence of current audit-history capability or an export artifact. | Sarah live-thread continuity + Priya/Jake product-truth lane | Yes | | Evergreen two-tab member-count display as visible-state trust | Evergreen admin-testing follow-up on stale member count after cancel; internal reproduction comparing stale visible state across open admin views before refresh | Visible admin state that lags a real action gets read as trust trouble even when membership data is correct. | Bounded visible-state example only; not wrong membership data, not a permissions failure, and not a general reliability claim. | Sarah + Jake/Priya support/product lane | No | | Acme visible-number trust as finance-confidence context | Existing Acme thread and internal notes under the current NDA-scope constraint | When visible numbers need explanation for too long, finance confidence drops even if the underlying billing source is right. | Narrow internal context only; constrained by the live NDA-scope boundary; not Mercury activation proof and not safe for new written materials. | Sarah/Devon with existing NDA boundary | Yes | | Pinecone retry semantics as support/process clarity | Pinecone support/process notes on retry explanation and the current API v2 path | Teams read retry behavior as product reliability unless support/process wording makes the current path explicit. | Support/process clarity example only; not enterprise-readiness proof and not a broader connector claim. | Support + current product-facts lane | No | Ranking / cut guidance for Monday - If we keep the read tight, lead with Evergreen audit language and Evergreen visible-state trust. - Acme is usable only with the NDA and scope caveat kept explicit. - Pinecone is the first row I would cut if the read starts feeling too unified across unrelated evidence. Open question for Monday framing - Do we want the read to show one anchor pattern plus one secondary contrast row, or do we want to show the fuller four-row candidate set and be very explicit that the rows sit in different owner lanes? Can you consolidate this into the internal-read section we should actually use — surviving rows and caveats only. Keep it pattern-capture/internal, not a strategy memo and not a change to Sarah/Devon routing.

All-hands stayed bounded. Nadia attended but did not speak; I answered the pre-submitted questions live: customer-growth work is internal for now, Mercury support evidence should be useful without becoming launch theater, the Kestrel newsletter is marketing hygiene, and Atlas continuity work is normal coverage discipline.

Write me a short note to Rishi asking for the migration-history details before we draft the support-safe archived-label example. In the shadow exercise, Leo could gather the source event and verifier output, but still had to ask where the archived org label came from. That’s the next Atlas runbook gap.

Jordan posted the narrowed auth replay fix note in #eng-releases. It covered update and deletion replay behavior, preserved archived-workspace metadata, and the Honeycomb evidence — no broader auth rewrite decision baked into it.

Kestrel’s maintenance-only Mercury newsletter is done with the approved bounded wording. Kara’s first-week read to Sarah was a few modest product-interest replies and no confused “funding announcement” response, so Sarah logged it as a small marketing hygiene win and left the public copy lane alone.

First warm afternoon of the week and Kibo dragged through the pre-dinner walk. Jamie moved the longer loop later; Kibo perked up after water.

Devon found a tiny HN thread off the Mercury newsletter drifting into “enterprise admin pivot?” speculation. A couple comments corrected it on their own, so we left it alone.

Maya stopped by with the printed appointment pages Mom wanted. Kibo barked through the handoff; I spent five minutes checking the pages were complete.

Tried an early Lake Merritt loop with Jamie and Kibo, but a charity run blocked our usual path. We cut across side streets, skipped the long loop, and got home before my cleanup block.

Sarah saw two late Evergreen audit-label comments sitting in the queue. She’s holding the live response for Monday and only forwarded Nadia a short redacted pattern note for the internal read prep.

Rishi picked one of Anna Rao’s windows for the next Tessl technical conversation and sent me a quick heads-up before the week starts again. Our Atlas runbook review stayed on Monday’s calendar.

Nadia’s Monday agenda draft is here. Tighten it for Friday so the read stays on signal taxonomy and owner routes, not a strategy memo. I’m fine cutting backup examples hard if needed. Friday May 10 customer-growth read — Monday agenda draft Date: Mon, May 6, 2024 Owner: Nadia Singh Audience: Morgan Chen, Sarah Kim, Devon Hayes, Jake, Anna Martinez Working goal - Keep the read internal, evidence-bounded, and useful for operating choices. - Turn current Mercury + Evergreen learning into repeatable pattern rows, not a sales forecast and not a strategy memo. - Keep an owner route on every row so customer-growth does not become the place every uncomfortable customer issue goes to die. Proposed live agenda - 1) Frame the read: what it is / is not — 3 min - internal pattern capture - no external wording changes in-room - no roadmap negotiation in-room - 2) Metrics anchor from Mercury — 8 min - activation reminder: a workspace is activated only when real_source_connected or first_live_sync_completed happens within 7 days - sample_import_completed and invite_sent are supporting signals only - directional note on retained_activity_week2 and unresolved admin setup - 3) Pattern rows — 20 min - work through 3-4 rows only - one repeated pattern per row - 4) Owner-route check — 8 min - ask whether the live route is support, product-truth, packaging/commercial, or customer-growth synthesis - 5) Parking lot — 6 min - capture anything that sounds like a next move rather than signal Working row format | evidence source | repeated pattern | concrete example | caveat | owner route | parking lot | | --- | --- | --- | --- | --- | --- | | metric cut, support example, live-account note, or packaging note | the smallest reusable pattern | paraphrase only; no direct quotes | why not to over-read it | who owns the live lane or next check | only if discussion drifts into action | Column notes - Evidence source: be explicit about where the row came from. - Repeated pattern: keep it narrow enough that it could survive beyond one account. - Concrete example: paraphrase only; do not paste customer wording. - Caveat: say the limit out loud so the row stays useful. - Owner route: who owns the live thread, live response, or product truth. - Parking lot: only for follow-on ideas that would otherwise turn the read into a task list. Candidate rows for Friday - Mercury activation friction - admin_first_action appears, but real source connection or first live sync does not happen inside day 7 - Evergreen admin-control clarity - role label gets over-interpreted when resend/remove powers are not visible on the same surface - Evergreen audit-language sensitivity - audit labels invite procurement-style inference faster than current product truth supports - Boundary case, only if time: Acme visible-number trust - finance-confidence context, not core Mercury activation evidence - Boundary case, only if time: Pinecone retry semantics - support/process clarity, not enterprise-readiness proof Unsafe for the read - Exact customer quotes - Any roadmap language, ship-date framing, or package line that sounds like a commitment - Summary labels such as enterprise-ready, compliance-ready, or expansion path - Blended rows that collapse product, support, procurement, and activation into one story - A next-move column in the main read; if it shows up at all, it should stay explicitly parked Read-safe rules - Use present-state language only. - If a row needs a product promise to sound coherent, it should not be in the Friday read yet. - Keep Sarah as live Evergreen continuity owner. - Keep Devon as commercial/procurement caveat owner. - Keep exact customer routing and copied thread language out of the read. Open choices before Friday - One anchor row plus one contrast row, or a short set of 3-4 rows with stronger caveats - Whether Acme and Pinecone stay backup only unless we explicitly need non-Evergreen comparison

Sarah pulled the late Evergreen comments. Can you draft the customer-safe reply so we cleanly separate current product-copy wording, support explanation, and procurement-sensitive language? Email thread excerpt Subject: Re: Mercury admin wording follow-up From: Evergreen admin testing team To: Sarah Kim Date: Sun, May 5, 2024 6:41 PM PT Sarah — Two late wording reactions from the admin/procurement side after the matrix below. 1) "View activity history" - Acceptable if it stays clearly present-state and does not imply a CSV, downloadable export, or report artifact exists somewhere behind the label. - The issue is not the word "activity" by itself; the issue is whether reviewers read the label as a path to a file/export surface that does not currently exist. 2) "Configured by workspace admin" - This tests better for the SSO surface than "required." - It reads as a workspace-level configuration state rather than bank-wide enforcement that is already in force. 3) "SSO required" - This still sounds like bank policy enforcement is already complete here. - Our reviewers do not read it as "a workspace admin can configure this"; they read it as "the bank already requires this." No timing ask in this note. We just wanted the late label reactions in before the next copy pass hardens. Thanks, Evergreen admin testing team Quoted context below On Mon, Apr 29, 2024 at 9:18 AM PT, Evergreen admin testing team wrote: Procurement wording matrix | Surface / phrase under review | What landed better | Remaining concern | | --- | --- | --- | | `View audit events` | Clearer than `Export audit log`; reads more like a present-state view than a contractual reporting promise. | It immediately raises the question of whether a CSV or downloadable report exists somewhere behind the label. | | `Export audit log` | None from our side. | Reads like a contractual reporting feature, not just a label on a current admin surface. Legal/procurement read it as if a report artifact already exists. | | `SSO required` | None from our side. | Still sounds like policy enforcement is already complete here, rather than something configured at the workspace-admin level. | | SSO wording framed as workspace-admin configuration | Better direction than `SSO required`. | The more the copy can make clear that this is about workspace-admin configuration rather than bank-wide enforcement, the less it gets read as a policy claim. |

Rishi sent the missing archived-label details for the Atlas shadow exercise. Discord DM — Morgan Chen ↔ Rishi Continued thread: Atlas coverage follow-up Mon May 6, 2024 9:14 AM Rishi: Promised migration-history details below. This is the piece Leo was missing in the shadow exercise. Atlas archived-label / migration-history note Status: working note for runbook cleanup, not final text Goal - Let the weekly support owner verify whether a workspace / org-invite issue is just old migration naming versus a real wrong-target problem. - Keep raw migration labels out of support/customer wording. - Keep missing-checkpoint and manual-replay decisions in the Atlas-owner lane rather than turning them into a support default. 1) Internal verification fields that are actually useful | Field / cue | What it tells us | Support/customer safe? | | --- | --- | --- | | current workspace id | Authoritative current target workspace | Safe only after translation into current workspace name; do not paste raw ids customer-facing | | membership record workspace id | Where the accepted membership grant will actually land | Safe only as a translated statement like "the membership record is attached to the intended workspace" | | invite target workspace id | The workspace the original invite was issued against | Internal verification only | | alias-history / rename record | Whether the current workspace used to carry an older display label during migration | Internal verification only | | `archived_org_label` / prior org label text | Old display label that may still appear in migration history or older incident notes | Never expose to support or customers | | acceptance event id / verifier output | Whether the accept flow hit the expected record path | Internal verification only | | last confirmed checkpoint | Whether replay/backfill evidence is complete enough to reason about recovery | Internal verification only | 2) The label Leo was asking about - The field to avoid exposing is the archived pre-migration org label stored in alias history (`archived_org_label`). - It is useful only to prove lineage when the old invite trail references a label that no longer exists on the product surface. - Support and customer copy should use the current workspace name only. Do not paste the archived label into a support thread, Zendesk note, or customer email. - If the archived label and the current workspace resolve to the same lineage, the support-safe summary is just: "we verified the invite and membership record are tied to the intended current workspace." 3) What counts as enough evidence for the migration-history check - Compare invite target workspace id to membership record workspace id. - Use alias-history only to explain an apparent name mismatch internally. - If verifier output exists, use it to confirm the accept path hit the same target workspace lineage. - Do not let the old label become the explanation itself. It is an internal cross-check, not customer-facing language. 4) Where the weekly support owner should stop - If the last confirmed checkpoint is absent, partial, or inconsistent with verifier output or customer-visible state, stop there. - Do not ask for manual replay. - Do not tell the customer replay will be run. - Do not keep digging for old labels in side threads as if that resolves the replay question. What to gather before escalating instead - current workspace name plus internal workspace id - membership record status / target workspace - invite target workspace - source event id if one exists - verifier output or accept-event trace if present - current customer-visible state - whether alias-history shows only a rename/display-label issue versus a true target mismatch Runbook line I would use for now - "If checkpoint evidence is missing or partial, the weekly support owner stops after collecting the verification fields above and escalates to Atlas owner review; support should not request or promise manual replay from the customer thread." 5) Lane reminder - Weekly support owner in the support rotation doc keeps thread ownership and the immediate next step. - Leo can use the fields above for a first-pass verifier/auth boundary read. - Jake only comes in if something turns into a time-sensitive sequencing or cut decision, not for migration-history debugging. If this helps, the support-safe archived-label example should end at "current workspace verified" and never quote the old label itself. Turn this into a support-safe runbook example. The example should explain how to verify the current workspace without exposing raw migration labels or making manual replay sound like the support default.

internal_search_v2 is returning the replacement API v2 retry snippet now: same-cursor GraphQL retry, the pre-resolver 401 mutation, and the unsupported token-rotation-header language. The stale v1-style cursor snippet didn’t show up in the test lookup.

Sarah at Founders Fund sent final logistics for the Embarcadero Tower session: two people remote, Zoom backup link is already in the invite, room unchanged, and HR confirmed the Tava Kitchen headcount still holds.

Jamie caught that Kibo’s care refill was down to the last dose. I picked up the replacement near home between late calls, so we’re covered.

Anna’s special Mercury cut is below. I need readout language for Nadia that keeps the real week-two activation/admin-setup weakness separate from rows that are just instrumentation cleanup or now clean. From: Anna Martinez To: Morgan Chen, Devon Hayes, Nadia Singh Date: Tue, May 7, 2024 2:14 PM PT Subject: Mercury metrics cut for Nadia Friday read Morgan / Devon / Nadia - I pulled a tighter Mercury cut for Nadia’s Friday customer-growth read so we can isolate the recurring week-two weakness from the rows that are now clean. I kept the usual row names for continuity. | Row | Metric | Looker | Mixpanel | Status | Note | | --- | --- | ---: | ---: | --- | --- | | 1 | activation_start | 83% | 83% | matched | held after the prior cleanup; no new caveat. | | 2 | admin_first_action | 62% | 61% | matched within rounding | steady; no caveat needed. | | 3 | invite_sent_week1 | 49% | 49% | held | stable supporting-activity row. | | 4 | sync_enabled_week1 | 38% | 38% | held | stable. | | 5 | retained_activity_week2 | 32% | 26% | real weak signal still present | concentrated in teams with unresolved admin setup; not a measurement issue. | | 6 | repeat_admin_action_week3 | 18% | 18% | matched | workspace-filter cleanup is holding. | Quick split on row 5 (retained_activity_week2) | Segment | Looker | Mixpanel | Read | | --- | ---: | ---: | --- | | All Mercury workspaces | 32% | 26% | reference row above | | Workspaces with unresolved admin setup | 18% | 12% | this is where the weakness is concentrated | | Workspaces with admin setup resolved by day 7 | 44% | 39% | not showing the same drop | Framing for Friday - This is still the weekly directional internal layer, not the quarterly cohort / board layer. - activation_start here still uses the corrected definition: real_source_connected or first_live_sync_completed within 7 days. sample_import_completed is excluded; invite_sent is supporting activity, not activation by itself. - For Friday’s read, I would keep row 5 as activation / admin-setup friction. Row 6 is now clean and should stay a watch row, not a second problem. - Anna

Priya’s microreview landed on the safer Evergreen copy direction: “View activity history” for the current prototype, no CSV/export language on the surface, and SSO helper text pointed at workspace-admin configuration instead of bank-policy enforcement.

Jake retested the two-tab member-count case after the focus-change cleanup. The canceling tab refreshes immediately, the second open tab refreshes on focus, and Leo is treating what remains as a display-refresh caveat, not a release blocker.

Kara has the first organic reply bundle from the Mercury newsletter and one person asked the broad-ship question. Draft a bounded answer she can use that doesn’t turn the newsletter into launch or roadmap language.

Anna Rao moved Rishi’s next Tessl technical conversation to the following Tuesday because Tessl’s engineering panel changed. Rishi is keeping the Atlas runbook review work on this week’s calendar.

Nadia’s v0 taxonomy draft is here. Please review it for bucket clarity and owner-route gaps before Friday; I don’t want the open blanks to turn into customer-growth owning everything by default. v0 customer-growth signal taxonomy for Friday May 10 read Date: Wed, May 8, 2024 Owner: Nadia Singh Working note - Owner route blanks are intentionally open for Friday discussion. - Use paraphrased examples only; exact customer quotes stay out. - Keep account continuity, commercial caveats, and product-truth lanes distinct. | Bucket | Working definition | Fits when | Does not include | Current example rows | Owner route | | --- | --- | --- | --- | --- | --- | | Activation friction | Where a workspace stalls before activation or immediately around activation. For this read, activated means real_source_connected or first_live_sync_completed within 7 days. | admin activity appears, but no real source connection or first live sync lands inside day 7; setup interest is visible but the account never reaches a real value moment | sample_import_completed by itself; invite_sent by itself; procurement or pricing asks that happen before live use | admin_first_action without first_live_sync_completed; team invited but source never connected; workspace looks in motion without becoming truly active | ____ | | Admin handoff and role confusion | Interpretation problems around roles, authority, visible admin state, resend/remove actions, and source-workspace wording. | the user is asking who can do what; a role name is carrying too much implied power; visible admin state is correct in backend but confusing on screen | confirmed backend permission bug; commercial/package dispute; broad enterprise-readiness claim | resend/remove pending invite authority unclear; source-workspace copy read as permissions risk instead of display clarity; stale member count after cancel reads as trust trouble | ____ | | Procurement/security packaging | Questions where current product truth is being translated into buyer, security, or procurement requirements. | wording choices trigger review beyond current capability; an account asks for procurement framing, admin model explanation, or security-language precision | roadmap promises; date commitments; account-specific exception requests; generic ARR or forecast discussion | audit wording implying export/reporting; SSO timing question; admin vs billing separation language; request for standalone procurement packet | ____ | | Expansion stalls after first live sync | Accounts that activated but do not broaden after the first live sync or first visible value moment. | first_live_sync_completed occurs, but follow-through does not broaden; initial enthusiasm outruns implementation confidence; no second admin motion appears | pre-activation setup blockers; support hygiene that happens before first live sync; packaging-only asks before real use | first live sync lands but no wider team adoption follows; champion is positive but setup confidence does not spread; usage remains one-threaded after first value | ____ | | True product gaps | Repeated issues that are not fixable by wording, support, or packaging alone and need a product-owner lane. | the same issue survives more than one clean example; support workaround is not durable; current-product wording still leaves a real capability gap | copy confusion; one account's procurement vocabulary; instrumentation caveat presented as product truth | audit-history capability gap if it persists beyond wording; admin/billing separation if current workflow remains ambiguous after copy/support fixes; repeated setup gap with no durable workaround | ____ | Boundary-case examples - Acme visible-number trust - Useful as finance-confidence context. - Keep internal only under the current NDA-scope boundary. - Not core Mercury activation evidence. - Not safe for new written Mercury examples. - Pinecone retry semantics - Useful as support/process clarity on the current API v2 path. - Not enterprise-readiness proof. - Not core Mercury activation evidence. - Do not let it turn into a broader connector story. Row-writing rules - One repeated pattern per row. - Paraphrase only; no exact customer quotes. - If a row needs roadmap language to feel important, it belongs out of band. - If the row is really a task list item, move it out of the taxonomy and into parking lot. Open blanks to settle Friday - Does admin handoff need a support subroute when the signal is real but the live reply still belongs to support? - When does procurement/security packaging stay a caveat bucket versus become a true product-gap input? - What minimum evidence should exist before a row moves from example to repeated pattern?

Draft the short scheduling note for the Friday read: start it ten minutes later so Jake can make it after the QA blocker review, but keep the tone clear that release triage still gets priority and isn’t being demoted.

Evergreen’s SSO wording question is now concrete. Email thread Subject: Re: Mercury admin wording follow-up From: Evergreen admin testing team To: Sarah Kim Date: Wed, May 8, 2024 11:17 AM PT Sarah — One concrete question from today’s admin walk-through. We now have two SSO surfaces using different wording, and reviewers are reading them differently. Question - When the admin-settings row says "SSO required," should we read that as bank policy is already enforced? - Or is the intended meaning that a workspace admin still has to configure enforcement for this workspace? Attached screenshot 1 — SSO toggle copy - Page: Admin settings > Security > Single sign-on - Section label: Single sign-on - Setting label: Require SSO sign-in - Helper text under the setting: Configured by workspace admin - Toggle state shown in the screenshot: Off Attached screenshot 2 — admin-settings row still using stronger wording - Page: Admin settings > Access - Row label: SSO - Status text: SSO required - No additional helper text visible on that row Why reviewers are pausing - Screenshot 1 reads like a workspace-admin setting that can be configured here. - Screenshot 2 reads like bank policy is already enforced and no configuration question remains. - We want to describe the current product truth accurately and not overstate enforcement. A short current-state clarification is enough. Thanks, Evergreen admin testing team On Sun, May 5, 2024 at 6:41 PM PT, Evergreen admin testing team wrote: Sarah — Two late wording reactions from the admin/procurement side after the matrix below. 1) "View activity history" - Acceptable if it stays clearly present-state and does not imply a CSV, downloadable export, or report artifact exists somewhere behind the label. - The issue is not the word "activity" by itself; the issue is whether reviewers read the label as a path to a file/export surface that does not currently exist. 2) "Configured by workspace admin" - This tests better for the SSO surface than "required." - It reads as a workspace-level configuration state rather than bank-wide enforcement that is already in force. 3) "SSO required" - This still sounds like bank policy enforcement is already complete here. - Our reviewers do not read it as "a workspace admin can configure this"; they read it as "the bank already requires this." No timing ask in this note. We just wanted the late label reactions in before the next copy pass hardens. Thanks, Evergreen admin testing team Prepare customer-safe wording Sarah and Priya can use. It should clarify the current workspace-admin configuration state without implying bank-wide enforcement or a broader SSO commitment.

Clerk’s delayed archived-user replay batch came through. Jordan checked Honeycomb: archived workspace metadata stayed preserved in the normalized path, and the replay traces don’t show a JWT metadata mismatch.

Devon’s proof-versus-pressure grid should become a caveat page in Nadia’s taxonomy, not the center of the meeting. Merge it into the draft without letting Acme or Pinecone read as Mercury activation proof. Friday May 10 customer-growth read — proof-versus-pressure grid Date: Wed, May 8, 2024 Author: Devon Hayes For: Nadia Singh Working distinction - Proof = repeatable interpretation or adoption problem we can state in current-product truth. - Pressure = customer heat, procurement vocabulary, pricing/package concern, or roadmap gravity that feels important but is not yet reusable proof. - Use the grid as a caveat page, not as the meeting center. | Example | What it is evidence of | Useful internal use | What it is not evidence of | Current lane | Guardrail | | --- | --- | --- | --- | --- | --- | | Evergreen admin-control clarity | Buyer-readiness friction around wording, visible state, and role clarity inside the current Mercury flow | Valid repeated signal for the Friday read, especially alongside visible-state/admin-role examples | Enterprise-readiness proof; launch proof; SSO timing; audit-history capability; admin-model commitment | Sarah live-thread continuity + existing product-truth lane; Devon only for caveat | Keep present-state wording. Do not let a wording example become a capability claim. | | Acme visible-number trust | Finance-confidence context from a live NDA-constrained thread | Narrow contrast row only if we need a non-Evergreen example | Mercury activation proof; generalized expansion pattern; safe external example | Devon/Sarah caveat lane under the active NDA boundary | Do not externalize. Do not let it read like cross-customer proof. No new Acme materials. | | Pinecone retry semantics | Support/process clarity on the current API v2 path and retry explanation lane | Useful if the read needs one support/process comparison row | Enterprise-readiness proof; broader connector-market story; reason to reopen docs scope | Support + current product-facts lane | Keep it on the current path only. No broader connector narrative. | Pricing/package caveat - If a row starts sounding like pricing architecture, procurement sequencing, or package design, stop and annotate rather than upgrade it. - That caveat stays in Devon's lane even when the underlying customer signal is useful. Practical test for Friday - If the point can be stated in current-product truth and survives a caveat, it is closer to proof. - If the point becomes interesting only once we imply a date, promise, or package story, it is pressure and should stay marked that way.

Sink scare was not a new Blueline repair issue. I sent them a photo of the dampness under the repaired sink; they reviewed it by message and said it looks like condensation on the supply line.

This final packet is useful but too wide to read straight through tomorrow. Friday May 10 customer-growth read — prep packet Date: Thu, May 9, 2024 Prepared by: Nadia Singh Contributors: Sarah Kim, Devon Hayes, Anna Martinez, Jake Audience: Morgan Chen, Sarah Kim, Devon Hayes, Jake, Anna Martinez Intent - Internal pattern capture only. - No external routing change in this packet: Sarah remains live Evergreen thread continuity owner; Devon remains commercial/procurement caveat owner. - Keep an owner route attached to every pattern so customer-growth does not become the place every customer-adjacent issue goes to die. - Exact customer quotes are intentionally omitted. Roadmap language is intentionally omitted. Suggested live-use order for Friday - 1) Anna metrics cut — 5 min - 2) Two anchor rows: Mercury activation friction; Evergreen admin handoff / role confusion — 12 min - 3) Procurement/security packaging row — 8 min - 4) Boundary-case contrast rows only if time: Acme visible-number trust; Pinecone retry semantics Section 1 — Nadia taxonomy v0, condensed for live use | Bucket | Working definition | Likely evidence sources | Candidate example rows | Owner route to confirm live | | --- | --- | --- | --- | --- | | Activation friction | Workspace stalls before activation or around activation. Activated means real_source_connected or first_live_sync_completed within 7 days. | Anna metrics cut; Jake support examples; setup notes | admin_first_action without first_live_sync_completed; source never connected inside day 7; setup looks active before real value lands | ____ | | Admin handoff and role confusion | Interpretation problems around roles, authority, resend/remove actions, source-workspace wording, and visible admin state | Sarah continuity notes; Jake support examples | role label over-read without visible powers; stale visible state after cancel; source-workspace copy read as permissions risk | ____ | | Procurement/security packaging | Current product truth gets translated into buyer/security/procurement requirements faster than we can safely generalize | Sarah continuity notes; Devon caveat page | audit wording implying export/reporting; SSO timing ask; admin vs billing language; procurement packet ask | ____ | | Expansion stalls after first live sync | first_live_sync_completed happens, but follow-through does not broaden into broader use | Anna metrics cut; support notes; customer thread notes | first live sync lands but no second admin motion; champion enthusiasm outruns implementation confidence | ____ | | True product gaps | repeated issue survives wording/support/packaging and needs a product-owner lane | repeated internal examples; product review input | durable capability gap after copy/support fixes; workflow gap with no stable workaround | ____ | Section 2 — Sarah live-thread continuity notes Keep explicit - Sarah remains the named continuity owner for live Evergreen follow-up. - Current external-safe lane remains the present Mercury test flow only: magic links, org invites, and the bounded preview experience. - Questions from Evergreen continue through Sarah. Useful internal patterns from the thread - Admin-role wording is not the issue by itself; reviewers want the visible authority on the same surface. - 'Admin' becomes more acceptable when resend/remove powers are visible enough next to the role state. - Source-workspace confusion is most reusable when framed as display clarity, not permissions risk, unless the backend grant is actually wrong. - Stale visible admin state after cancel reads as trust debt even when backend membership data is correct. - Audit wording triggers contractual/procurement inference faster than current capability supports. Keep out of the general read - Exact Evergreen reviewer routing and follow-up sequence - Exact wording already committed in the live thread - Anything that implies SSO timing, audit export capability, a standalone procurement/security packet, or a roadmap commitment - Any wording that makes Nadia the new external owner Section 3 — Devon proof-versus-pressure grid Working distinction - Proof = repeatable interpretation or adoption problem we can state in current-product truth. - Pressure = customer heat, procurement vocabulary, pricing/package concern, or roadmap gravity that is real but not yet reusable proof. | Example | What it is evidence of | Useful internal use | What it is not evidence of | Current lane | Guardrail | | --- | --- | --- | --- | --- | --- | | Evergreen admin-control clarity | Buyer-readiness friction around wording, visible state, and role clarity inside the current Mercury flow | Valid repeated signal for the Friday read | Enterprise-readiness proof; launch proof; SSO timing; audit-history capability; admin-model commitment | Sarah continuity + existing product-truth lane | Keep present-state wording. Do not let a wording example become a capability claim. | | Acme visible-number trust | Finance-confidence context from a live NDA-constrained thread | Narrow contrast row only if we need non-Evergreen context | Mercury activation proof; generalized expansion pattern; safe written example | Devon/Sarah caveat lane under the active NDA boundary | Do not externalize. No new Acme materials. | | Pinecone retry semantics | Support/process clarity on the current API v2 path and retry explanation lane | Useful if we need one support/process comparison row | Enterprise-readiness proof; broader connector-market story | Support + current product-facts lane | Keep it on the current path only. No broader connector narrative. | Section 4 — Anna metrics cut Directional internal cut only — not board language Definition reminders - Activated within 7 days = real_source_connected or first_live_sync_completed within 7 days of workspace_created. - sample_import_completed is excluded from activation. - invite_sent is supporting activity only. - retained_activity_week2 is directional and useful for operating questions. - repeat_admin_action_week3 stays visible, but it is still an instrumentation-caveated measure and not a headline claim. Completed cohort cut available as of Thu, May 9 | Cohort week | New Mercury workspaces | Activated within 7d | sample_import_completed only within 7d | admin_first_action but no live sync by day 7 | retained_activity_week2 among activated | repeat_admin_action_week3 | | --- | --- | --- | --- | --- | --- | --- | | Apr 1 | 18 | 10 (56%) | 4 | 3 | 6/10 (60%) | 4/10 (40%) | | Apr 8 | 21 | 12 (57%) | 5 | 4 | 7/12 (58%) | 5/12 (42%) | | Apr 15 | 24 | 13 (54%) | 6 | 5 | 7/13 (54%) | 4/13 (31%) | | Apr 22 | 19 | 11 (58%) | 4 | 3 | 6/11 (55%) | pending | Admin-friction split, directional | Split | Workspaces | Activated within 7d | retained_activity_week2 among activated | Note | | --- | --- | --- | --- | --- | | Admin-setup friction tagged | 32 | 15 (47%) | 5/15 (33%) | Weakest durability cluster; most examples involve invite/role/source-workspace confusion or visible-state trust | | No admin-setup friction tag | 50 | 31 (62%) | 21/31 (68%) | Better week-2 durability once the setup path is clear | Read notes from Anna - The biggest drop is not initial interest; it is the cluster where admin motion appears, then stalls before real source connection or first live sync. - retained_activity_week2 is worth keeping visible because it tracks the accounts that looked alive early but did not stabilize. - repeat_admin_action_week3 should stay caveated and should not carry a customer-growth conclusion by itself. - The Apr 29 cohort is omitted from week-2 and week-3 views because the observation window is not complete as of this packet. Section 5 — Jake support examples Classification note: putting a support example into a bucket does not change the live reply owner. | Support example | What kept happening | Candidate bucket | Live owner note | | --- | --- | --- | --- | | Pending invite resend/remove unclear | Reviewer sees role state but not obvious authority for resend/remove and asks who can actually clear the issue | Admin handoff and role confusion | Weekly support response stays in support; useful as recurring interpretation pattern | | Source-workspace/header confusion after invite acceptance | Header can briefly reflect last active workspace and gets read as permissions risk before acceptance flow is understood | Admin handoff and role confusion | Use current-product wording; only escalate as permissions problem if membership record is wrong | | Stale member-count display after cancel until refresh | Visible count lags the real action and gets interpreted as trust trouble | Admin handoff and role confusion | Bounded visible-state example; not a wrong-membership claim unless backend record is wrong | | First live sync completed, then no second admin motion | Team thinks setup is done for everyone after first live sync, but broader rollout never starts | Expansion stalls after first live sync | Useful only as an internal follow-through example; not a forecast row | | Pinecone retry semantics on current API v2 path | Teams read retry behavior as product reliability unless current path and retry expectations are explained clearly | Boundary-case support/process clarity | Support lane only; not enterprise-readiness proof | Section 6 — Candidate rows for Friday live discussion Use paraphrased examples only. Exact customer quotes intentionally omitted. Live rows | Evidence source | Repeated pattern | Concrete example | Caveat | Likely owner route | Parking lot | | --- | --- | --- | --- | --- | --- | | Anna metrics cut + Jake support examples | Admin motion can create false comfort before real activation | Team invites members and touches admin surfaces, but no real source connection or first live sync lands inside day 7 | Directional internal signal only; not customer-facing proof | Jake/Priya product-truth lane + Anna metric caveat + Nadia synthesis | Whether activation instrumentation should split admin_first_action by workspace state | | Sarah continuity notes + Jake support examples | Role labels need visible authority on the same surface or buyers over-infer | 'Admin' lands better once resend/remove powers are visible enough next to the pending state | Wording/visibility pattern only; not capability expansion claim | Sarah continuity + support/product-truth lane | Whether admin examples need a support subroute | | Sarah continuity notes + Devon caveat page | Audit wording creates procurement inference faster than product truth supports | Safer wording reduces over-read, but even lighter audit labels can still imply export/reporting expectations | Buyer-language sensitivity only; not evidence of current audit-history/export capability | Sarah continuity + Devon caveat + product-truth lane | What is the minimum product-owner check before a product-gap row appears | Contrast rows only if time | Evidence source | Repeated pattern | Concrete example | Caveat | Likely owner route | Parking lot | | --- | --- | --- | --- | --- | --- | | Existing Acme notes | Visible numbers lose trust when explanation burden stays high | Finance reviewer confidence drops when numbers require repeated interpretation even if source data is correct | Narrow NDA-constrained context only; not Mercury activation proof and not safe for new written materials | Devon/Sarah under current NDA boundary | Keep internal only | | Pinecone support notes | Retry behavior gets read as product reliability unless the current path is made explicit | Retry semantics question is really a support/process clarity problem on the current API v2 path | Support/process example only; not enterprise-readiness proof | Support + current product-facts lane | First row to cut if the packet is too broad | End note - The packet is intentionally wider than the live discussion set. - Goal for Friday is not to read every section aloud; goal is to use the packet to keep the conversation evidence-bounded and owner-routed. Pick the final order and cuts for the live read. I want it to fit the slot and stay on taxonomy, evidence, caveats, and owner routes—not a packet walkthrough.

Founders Fund working session is done. One attendee used the Zoom backup, the Embarcadero Tower room setup held, Tava Kitchen headcount was right, and Sarah Kim sent a short logistics closeout afterward.

Sarah sent the narrowed audit/SSO wording reply to Evergreen. They accepted that CSV/export and “compliance-ready” language stay out of current copy; what’s left is just Priya’s next wording preference pass.

Leo reran the Atlas shadow scenario with Rishi’s archived-label example. He stopped exposing the raw migration label, but still hesitated when checkpoint evidence was partial and the hypothetical customer asked for manual replay. Jake confirmed there’s no product sequencing decision for him here.

HR asked whether Friday’s customer-growth read should go into the quarterly all-hands archive. I kept it as an internal operating read, not company-wide material.

Nadia’s first customer-growth read happened with me, Sarah, Devon, Jake, and Anna. Good shape: she did not turn it into a sales forecast. She classified the Mercury/Evergreen learning into activation friction, admin handoff and role confusion, procurement/security packaging, expansion stalls after first live sync, and true product gaps. The important rule we kept was an owner route attached to each bucket, so customer-growth doesn’t become the place every uncomfortable customer issue goes to die. We left with v1 of the signal taxonomy for turning messy design-partner learning into operating input.

Anna caught one spreadsheet row after the read: teams with unresolved admin setup had been labeled as expansion stalls. She corrected it, and retained_activity_week2 stays framed as activation/admin-setup friction, not expansion evidence.