M MemberIntel KB

reference

Meeting Log — Blair × Seth, 2026-05-11

Full decision log from the Blair × Seth working session on 2026-05-11: friction-points resolutions, team structure changes (Product Lead retired → Santiago expanded; Russ added as Lead Designer), free-tier LLM strategy (local Ollama-class), GCP hosting confirmed, 14-day Pro trial for existing MP members, domain registrations, action items.

Date: Tuesday, May 11, 2026, 15

MDT
Attendees: Blair Williams, Seth Shoultes
Source notes: Gemini meeting notes (~/Downloads/intel - 2026_05_11 15_14 MDT - Notes by Gemini.md) + the friction-points prep doc Blair had read in advance (/blair-friction-points-prep)
Recording: see Google Calendar event metadata for the link (internal)


TL;DR

Seven friction-points decisions resolved. Team structure restructured: Cindy’s vacated Product Lead role not backfilled — Santiago picks up Product+Project Lead duties (80% MemberIntel / 20% cross-Caseproof). Russ joins as Lead Designer. Free tier runs cheaper local-LLM (Ollama-class) to manage cost; Pro stays on paid Anthropic models. GCP confirmed for hosting. Existing MP customers get a 14-day Pro trial via frictionless toggle, then downgrade to free tier. Action items below.


Strategic decisions

Team structure

  • Cindy’s Product Lead role: NOT backfilled. Blair’s reasoning: too much predicted Cindy-Santiago coordination friction. Santiago absorbs Product Lead duties on top of his existing Project Manager mandate. Capacity allocation: 80% MemberIntel / 20% remaining cross-Caseproof work (shopping-cart issues + other initiatives). See /santiago-jd for the expanded role definition. The historical Product Lead JD is preserved as a retired-reference for the duties Santiago now owns.

  • AI Engineer hire — Blair owns the JD. Blair plans to hire a Senior AI Engineer under Seth. A LinkedIn candidate was surfaced during the meeting with “AI-driven financial transaction intelligence” experience + LoRA/QLoRA fine-tuning + synthetic data work — flagged as a strong fit. JD already drafted at /ai-engineer-jd; Blair to take it from there and run the search.

  • Code review hire — meticulous reviewer. Blair explicitly wants a careful, testing-oriented code reviewer for the volume of code Seth will produce. Ronald will migrate from current work to MemberIntel once defined work exists for him; potentially Kalpesh after.

  • Russell (Russ) joins as Lead Designer. New seat on the team. Initially part-time / contract through Phase 1; potential full-time conversion in Phase 2–3 based on design surface area. JD: /russ-jd. Reports to Blair.

  • Brain content lead — Sam from support (not external hire). Friction-points item 1 resolution: instead of hiring an external content lead, repurpose Sam (and possibly one other technically-inclined support agent) once the MP support-assist light reduces ticket load. Cost saving + faster start than external sourcing.

  • Consultants attached:

    • Kurt — growth consultant (marketer / small-business owner perspective)
    • Thomas — technical consultant (continues prior involvement, e.g., GCP recommendation)

Hosting & infrastructure

  • GCP confirmed as the hosting platform. Aligns with prior Thomas conversations. Consolidates billing with other Caseproof GCP usage. ADR-0004 in the V1 product repo already locked this; the meeting ratified.
  • A/B staging strategy: Seth to set up two staging sites for cost/quality comparison — one running Ollama-class local LLM (target free tier), one running paid Anthropic / Google Cloud models (target Pro tier). Compare before GA.

Pricing & tier strategy

  • Free-tier LLM = cheaper local model (Ollama-class). Driving force: paid LLM cost on a free tier at 50,000 users projects to ~$50,000/month, which doesn’t pencil. Blair accepts slower replies + higher latency on free tier as the upsell lever to Pro.
  • Open-source model choice considerations (Seth flagged): Nvidia-hosted Ollama models avoid the “Chinese bleed-through” issue some open-source models have (e.g., Qwen). Google Cloud–hosted models also under consideration if hosting consolidation pays off.
  • 14-day Pro trial for existing MP customers. Existing MemberPress members are auto-enrolled in a 14-day Pro trial via a frictionless in-MP toggle. Data ingestion starts immediately. Downgrades to free tier on day 15 if no upgrade. (Resolves v1-pro-30day-launch-trial; duration tuned from 30 to 14.)
  • Frictionless data ingestion is the primary onboarding priority. Toggle-on → analyze → personalize. A smarter paid model may be needed in the background to distill raw customer data into the brain before serving it through the free-tier RAG.

Brand & domain

  • Platform-agnostic positioning is locked. MemberIntel must read as a standalone product, not a MemberPress add-on. Required for V2 BuddyBoss launch.
  • Domain registrations approved:
    • memberpress.cloud (Blair to register)
    • membersintel.com (Blair to register)
  • Final naming TBD; lean toward standalone “MemberIntel”.

Privacy & data anonymization

  • Strong anonymization required. Customer financial / membership data must not be traceable back to specific users (Seth’s example: an oncologist’s MemberPress site). Important for legal exposure.
  • Privacy doc for Allen (privacy council) — Seth to prepare a presentation outlining the security & anonymization strategy. Goal: feedback from Allen before June (in time for the late-May counsel architecture review per friction-points item 4).

Friction-points resolutions (full)

All 7 decisions from /phase-1-2-friction-points and the prep doc at /blair-friction-points-prep resolved:

#DecisionResolution
1Brain content lead in Phase 1Approved — staffed internally with Sam (not external hire)
2Phase 2 milestone language calibratedApproved
3Differentiation eval as Phase 3 milestone + monthly reviewApproved — monthly 30 min with Blair starting Phase 4
4Privacy counsel architecture review late MayApprovedSeth preps privacy doc for Allen
5Infra hire on V1.5 roadmapApproved as budget signal — Ronald migrates first, Kalpesh later
6Model abstraction layer Phase 1Approved — already shipped in V1 product kickoff (ADR-0005 + llm.call)
7Brain versioning in Phase 1 schemaApprovedversion + superseded_at columns

V1-spec open-questions resolved or moved forward

Logged inline on /v1-spec; summary:

Q IDStatus
v1-naming-trademark (Q1)Partial — domains approved, name still TBD, trademark check owed
v1-content-lead-sourcing (Q2)Resolved — Sam (internal)
v1-hosting-choice (Q8)Resolved — GCP
v1-free-tier-opt-in (Q12)Direction set — frictionless toggle; consent UX owed
v1-pro-30day-launch-trial (Q13)Resolved — Yes, 14 days
v1-brand-presentation-free-tier (Q15)Direction set — standalone “MemberIntel” with Pro upsell

Questions remaining open: Q3 (token budgets), Q4 (Stripe Connect vs OAuth), Q5 (customer support model), Q6 (eval methodology), Q7 (MP API surface), Q9 (vector store), Q10 (cross-pollination cadence), Q11 (observability), Q14 (quarterly free-tier review).


KB / process decisions

  • Blair commits to a daily KB review. He’ll start each morning on memberintel.caseproofagent.com, answering open questions and reviewing standup updates. Specifically asked for the standup to be emailed daily as a reminder.
  • Email notifications via ResendSeth to design and ship daily standup emails + new-question / new-comment notifications. Spec drafted at docs/superpowers/specs/2026-05-11-resend-email-notifications-design.md.
  • Design mockup hand-off to RussSeth to share the current Claude design artifact with Russ for review. Index lives at /design-references. Russ’s Phase 1 deliverable is migrating it to a Figma source-of-truth file.
  • Knowledge base demoed and praised. Seth walked Blair through memberintel.caseproofagent.com, including the chat (RAG), open-questions side panel, friction-points doc, and the personalized Blair view. Blair committed to daily use.

Action items

From the Gemini notes plus what surfaced during the working session:

  • Blair Williams — Prepare and publish the AI Engineer JD (post the role for sourcing). Note: JD already drafted internally at /ai-engineer-jd and at /ai-engineer-job-posting; Blair to finalize and post externally.
  • Blair Williams — Review the friction-points doc immediately and confirm the 7 resolutions logged above. (Resolved in-session; this is the durable check.)
  • Blair Williams — Register memberpress.cloud and membersintel.com.
  • Seth Shoultes — Prepare privacy strategy presentation for Allen (Blair’s privacy counsel) ahead of the late-May architecture review. (Drafted 2026-05-12 at /privacy-strategy-for-counsel — Brandeis-counsel reasoning pass + McPhee-prose rewrite + panel review (Lessig, Sunstein, Rawls) + Lessig/Rawls debate on permanent exclusion. Ready for Allen.)
  • Seth Shoultes — Send the current design mockup to Russ for review (Claude design link in /design-references).
  • Seth Shoultes — Update job-roles documentation, delegating Product Lead responsibility to Santiago. (Done 2026-05-11 — see /santiago-jd and /product-lead-jd retirement banner.)
  • Seth Shoultes — Implement email alerts via Resend (daily standup + new-question / new-comment). Spec drafted; plan + implementation pending.
  • Seth Shoultes, Santiago — Schedule customer interviews to inform onboarding flow + early product decisions.
  • Seth Shoultes — Stand up the A/B staging sites (Ollama vs. paid model) for free-tier comparison.

Open questions raised in this meeting (carry-forward)

  • Smart data-distillation model for the free tier. A paid model may be needed to process raw customer data into brain entries even on the free tier, since the local LLM may not be smart enough to distill it well. Cost/quality trade-off pending the A/B staging comparison.
  • Sam’s exact workload as brain content lead. Sam’s current support workload must drop before he can absorb the content authoring role. Depends on MP support-assist light shipping and reducing ticket load. Date for Sam’s transition: not committed.
  • Russ — full-time conversion date. Tied to Phase 1–2 design load. Decision point: end of Phase 2 (June 30, 2026), revisited end of Phase 3.
  • AI Engineer sourcing speed. Target close: mid-June 2026 per /ai-engineer-role. Depends on Blair’s external posting + interviewing.
  • Final product name. Domains secured for two candidate names; final decision still owed.

Document version: v1 — drafted 2026-05-11 directly after the meeting. Source: Gemini auto-generated notes plus the friction-points prep doc Blair had read in advance. Decisions logged here are authoritative; downstream pages (decision-rights matrix, role JDs, friction-points page, V1 spec open-questions) updated to reflect resolutions in the same commit.

For: B Blair Williams S Seth Shoultes S Santiago Perez Asis P Product Lead A AI Engineer R Russell