Morgan Chen

Sarah wants the exact line she can use with Evergreen so she doesn't overpromise on the additional-admin-group question.

Send her customer-safe wording that stays on current-state only, with no advanced-admin commitment, and log an internal note that this is enterprise-readiness evidence rather than a roadmap promise. Keep Devon out of the day-to-day route unless it somehow becomes a true emergency.

Update the Q3 operating-evidence scratch deck with the July 2 Evergreen boundary.

I want it to land as: real enterprise-readiness evidence, but not repeatable growth proof, not a staffing trigger, not a custom roadmap promise, and not a reason to pull Devon back into routine customer narrative work during July.

Sarah says Evergreen understood that custom admin policy is not in the annual baseline today. They didn't get a date or commitment, but their admin group still wants a custom-policy path eventually and she expects the question to come back later.

Give me 3 short internal bullets I can reuse that separate “customer accepted current-state” from “we promised future advanced admin work.”

Leo just asked about a vague API v2 weirdness mention that landed with him directly and doesn't have a support owner yet.

Send him a short routing reply: side-channel Atlas weirdness goes back through support first; once support has a named owner + immediate next step, he takes the first technical read; Jake only comes in for bounded product sequencing; Devon is not the default route.

Late Mercury/owner-routing thread ran right up against the protected dinner block.

Text Jamie that I'm leaving now and still keeping dinner tonight.

Holiday check only: scan the inbox for anything that actually counts as customer, legal, or coverage work during Devon's pause.

Inbox only. Don't turn this into cleanup, and don't archive or send anything unless something is genuinely urgent.

Friday evening. Order the usual Lemongrass green curry.

Quick Kibo note to record: we kept the protected Saturday morning block, did the early short flat Lake Merritt-side walk, and he was stable.

Started a little stiff, loosened up, ate normally after, and no vomiting or new gait issue. Just record it — no follow-up needed.

Quieter weekend, so Jamie and I picked next week's protected household blocks before week two of Devon's pause fills itself.

Put holds on Thursday July 10 from 6:30–8:30 p.m. PT and Saturday July 12 from 8:00–10:00 a.m. PT. Same actual-incident/customer-impacting-emergency exception, and add the short Lake Merritt-side Kibo walk note if he's comfortable.

Sunday reset before week two of Devon's pause. First week was basically the pattern I expected.

Week-one notes: - Evergreen custom admin-policy ask got answered as current-state only and logged as enterprise-readiness evidence; no advanced-admin commitment. - Atlas side-channel weirdness should go back through support before Leo takes the first read. - Compass July prompt tracking counts dated owner follow-up only, not sightings. - Board/customer narrative stays with me. - Devon is true-emergency-only. - Household blocks for July 10 and July 12 are protected.

Draft me a short Monday checklist I can glance at: what I hold, what routes to owners, and what actually qualifies as a Devon emergency. Don't send anything.

Anna just sent this on Compass: Morning — I'm seeing the first few July prompt observations come in unevenly. Examples so far: - Evergreen admin-handoff prompt surfaced in Sarah's account note; Sarah may have follow-up, but I don't have the date/source attached yet. - Acme account-health note has a renewal/admin-friction prompt mention, but no owner action in the row. - One post-first-live-sync expansion prompt shows up in a non-Evergreen account note, but the owner only marked "watching". Do you want the scratchpad to log sightings immediately, or only rows with dated owner follow-up?

Send Anna and Sarah a short internal clarification: sightings can be logged, but only dated owner follow-up with source, caveat, and outcome counts as evidence for the Q3 read.

Jake just sent this: If Acme's May export timeout follow-up comes back during the account-health review, do you want it on Mercury weekly risks too, or keep it only in Atlas/support? Trying not to make the wrong list while Devon is out.

Send Jake a short reply: keep export weirdness in support/Atlas unless it actually changes Mercury activation or admin scope.

Acme account-health review just surfaced a recurrence through the normal customer/support path. Sarah's notes: Acme account-health review — July 8 Attendees: Greg Shipman, Acme account ops; Sarah Kim, Scaffold Greg raised two new API v2 large-export failures from the account-health export workflow: 1) Export job acme-exp-2025-07-07-01 - Requested July 7 at 6:42pm UTC - Workspace: acme-prod - API: GraphQL API v2 export endpoint - Date range: Apr 1–Jun 30, 2025 - Fields: account, workspace, nested team/member fields, billing status - Result: request ran for ~12 minutes, then timed out; Acme UI showed failed export 2) Export job acme-exp-2025-07-08-02 - Requested July 8 at 2:15pm UTC - Workspace: acme-prod - API: GraphQL API v2 export endpoint - Date range: Apr 1–Jun 30, 2025 - Fields: same saved export template, including nested team/member fields - Result: timed out again; Greg says Acme can run narrower exports but wants to understand whether the 90-day export is safe to rely on for the account-health review Greg did not report missing rows yet. He asked whether this is the same large-export path as May and whether Scaffold has a safer replay path. I have not promised an ETA or a fix date.

Create a dated internal intake note, then send routing replies to Sarah, Leo, and Jake. Keep this in support/Atlas: Sarah stays on customer continuity, Leo takes the first technical read, Jake only once it's scoped, and no Devon route.

Sarah just asked for exact customer-safe wording: Greg is asking if he should tell his account-health team to stop using the 90-day export. I can say we're checking it, but I don't want to overstate. Can you give me the safe line? Leo hasn't finished yet.

Send her the short holding line for Acme. No overclaiming, no fix date.

June AWS invoice just landed above the manual-review threshold. Forward below: From: AWS Billing <no-reply-aws-billing@example.com> Subject: AWS Invoice AWS-2025-06-SCAFFOLD is ready Date: Wed, Jul 9, 2025 at 7:18 AM PT Invoice ID: AWS-2025-06-SCAFFOLD Billing period: June 1–June 30, 2025 Vendor: AWS Account: Scaffold production + non-production cloud services Amount due: $6,312.44 Due date: July 24, 2025 This invoice includes EC2, RDS, NAT Gateway, S3, CloudWatch, and related usage charges for the billing period.

Flag it for manual AP review because it's over $5,000.

Leo finished a first read on the Acme export recurrence: Acme export recurrence — first read Checked: - acme-exp-2025-07-07-01 - acme-exp-2025-07-08-02 - related request logs for acme-prod API v2 export endpoint What I see: - Both failures hit the cursor-paginated large-export path. - Both requests used the Apr 1–Jun 30 date range and the saved export template with nested team/member fields. - Auth/JWT verification looks normal; no elevated 401s or token-refresh loop. - Webhook delivery is not involved. - Connector verifier is not involved. - The request work keeps extending the cursor window, then backs into timeout before completion. Replay: - I replayed Apr 1–Apr 30 with the same nested fields: completed. - I replayed May 1–May 31 with the same nested fields: completed. - Row counts from the narrower replays match the expected count ranges from Acme's last successful monthly exports. - I do not see corrupt exported data from the narrower replay checks. Not final yet: - I still want to confirm whether chunking the full 90-day export plus backoff would avoid the cursor timeout. - I would not promise a fix date yet.

Create a prelim technical note, send Sarah the customer-safe wording, and send Jake the bounded sequencing question. Keep it support/Atlas, not Compass, not Mercury, not founder-direct.

Anna and Sarah are asking how to log the Acme row in the July Compass scratchpad without turning the live support issue into Compass evidence: Sarah: Compass surfaced a renewal-risk/admin-friction prompt in the Acme account-health note because the export timeout came up in the account review. I already routed the live issue through support + Leo. Anna: For the July scratchpad, do I count this as a prompt hit with owner follow-up? It is a real dated handoff, but the issue itself is Atlas/API v2. I don't want to pollute the Compass read.

Send them a short clarification: it can be logged only as a caveated prompt sighting plus dated support handoff if the source is attached. The live issue stays Atlas/API v2 context, not Compass proof.

Jake just replied on the Acme follow-up: I can take this as a bounded Atlas/API v2 export-path item if Leo's chunking/backoff read holds. Product shape: split the 90-day range into chunks and back off before cursor timeout. Not a Mercury activation issue and I don't see a Compass implementation angle. I would not promise Acme a ship date until Leo finishes the replay confirmation.

Send a concise note to Jake, Leo, and Sarah confirming the lane: bounded Atlas/API v2 follow-up if Leo confirms, and no customer ship-date promise yet.

Text Jamie that I'm leaving now and still keeping dinner. Just a few minutes behind.

Final July Acme closeout notes are in: Leo Park final verification note: - The two July Acme failures are in the cursor-paginated API v2 large-export path. - Trigger shape: long Apr 1–Jun 30 date range plus nested team/member fields. - Auth/JWT, webhooks, and connector verifier remain clean in the logs. - Narrower replay completed for Apr, May, and Jun monthly ranges using the same nested fields. - Row counts match expected ranges from Acme's prior successful monthly exports. - I do not see corrupt exported data or customer-visible data loss from the replay checks. - Recommended bounded fix: chunk the long export range and back off before the cursor path times out. Jake sequencing note: - Accepted as a bounded Atlas/API v2 follow-up for chunking/backoff. - Not Compass implementation scope. - Not Mercury activation/admin scope. - No founder-direct escalation and no Devon route during the July pause. - July queue item; no customer ship-date promise in the Acme thread. Sarah customer-thread note: - Greg is okay continuing the account-health review with narrower exports/replay while the full-range follow-up is queued. - Customer-safe line should say we found no customer-visible data-loss signal in replay checks, not that every future export path is proven safe.

Create the durable closeout record, update the Acme CRM note, update the Q3 operating evidence slide, and send Sarah the final support-safe line. I want this locked as bounded Atlas/API v2 follow-up with no data-loss finding — not Compass, not Mercury, not Devon.

AP sent this back on the June AWS invoice: From: AP Review <ap-review@scaffold.example> Subject: Cleared for payment — AWS-2025-06-SCAFFOLD Date: Fri, Jul 11, 2025 at 12:54 PM PT Invoice AWS-2025-06-SCAFFOLD Vendor: AWS Amount: $6,312.44 Billing period: June 2025 cloud services Review result: Cleared for payment. Notes: Charge pattern is in expected range after the prior non-prod hygiene work; no manual hold needed. Please authorize disbursement when ready.