WholeTech — Session Notes

🧲 The PolyMagnet meeting → transcript → follow-through

• Meeting run-of-show + numbered talking points built pre-call: wholetech.com/script/ and /script2/ (nginx if(-d) self-loop fixed with an explicit location ^~ /script/ block).
• Meeting ran ~90 min (2:01–3:29 CT), Tim only (Melissa had a conflict). Pulled the full lifelog transcript via the Limitless API (date=2026-07-07) — the B2→droplet sync lags a day, so query the API directly for same-day. Key discovery: Tim = CEO of Correlated Magnetics / PolyMagnet; polymagnet.com is on Shopify; invented by Larry Fullerton, licensed to Apple.
• Two emails drafted to Tim + Melissa (Gmail drafts): a progress recap and a "big update + what to test" note. The final one is addressed to both, ready for Paul to send.
• LESSON: Limitless lifelog same-day data is only on the API, not the B2 mirror — see [[project_melissa_morman_bdx]] (updated with the full meeting outcome).

✅ Fixed Melissa's objection (the "money shot")

• Her experience was "quick but the dashboard wasn't remotely correct — instantly you don't trust it." Root cause: onboarding dumped users on a pre-baked demo dashboard (data.js) instead of their real audit.
• Fix: the 5 onboard forms (deptzy/deptless/automarketingdept/automarketing.wholetech/amdept) now POST the URL to /api/analyze, then /api/activate (generates real deliverables), then land on the accurate report. Verified end-to-end: onboard "multipolemag.com" → deptmatic.com/api/report/multipolemag.com shows a real content brief, title/meta rewrites, schema — held for approval.
• Also fixed onboarding field wording: "Business name" now clarifies brand-vs-corporate; "Location/service area" → "Where you sell" (5 forms + a brand hint on 110 marketingapp instances via /root/marketingapp-shell/app.js).

🔌 Answered Tim's connection question — /connect/ + hosting

• New /connect/ page on all 10 product sites: 4 honest connection methods (copy-paste + we-host = "works today"; platform connector + one-line tag = "rolling out"/"coming"), an approval/rollback promise band, Shopify/WordPress section, FAQ schema. Framed as a real "Connect & Ship" step in the loop.
• The connect step is IN the flow: every engine report now ends with a "Connect your site to ship →" CTA (/opt/autoengine/server.py).
• "We host it" is actually implemented: a working Request hosting form on /connect/ posts to the launch-intake API → lands in the admin panel's leads. Backed by real capability (~226 hosted sites, admin panel edit/deploy, the Path B plan). All 10 sites have a host-request path.
• Path B plan documented at deptmatic.com/backend/polymagnet-plan/ (hybrid: keep Shopify checkout, we host the marketing surface, staging subdomain, rollback).

🔬 Competitor research → enrichment applied cluster-wide

• 8 deep teardowns (Jasper, HubSpot, Copy.ai, Writesonic, Hootsuite, Surfer, BrandWell, AirOps+Relevance) + a ranked plan, saved at deptmatic.com/backend/competitors/. Convergent finding: they all abandoned SMB and hedge on control — deptmatic owns the vacated bottom and the "you approve, then it ships" wedge.
• Applied across all 10 sites (each in its own design): a named department org-chart (Auditor, SEO Manager, Content Desk, Social Scheduler, Email Writer, Approval Queue), the Audit→Draft→Approve→Connect & Ship→Measure loop, a hero "nothing ships without your approval" promise, and an honest /compare/ page (real competitor pricing + FAQ schema).
• Flagship role pages (Jasper's /agents/* lesson): deptmatic.com/department/ + /audit, /seo, /content, /social, /email.
• Free public audit tool built on the flagship engine: automarketingengine.com/audit/ (0–100 gauge, "do these 3 first", reuses the existing engine).
• All 12 lessons encoded as a copy-paste prompt card on every product /act/ worksheet (/root/gen-act.py).

💲 $499/mo founding rate — reconciled across all 10

• Was inconsistent ($2,400 / $2,900); Paul chose a $499/mo founding rate anchored against ~$2,900 market. Repriced all 10 carefully — changed the product's OWN price, KEPT market anchors, competitor prices (Content Cucumber/Verblio), word-counts ("2,400-word") and blueprint drawing numbers. Calculators, invoices, JSON-LD, payroll counters all recomputed.
• LESSON: never blind-sweep a price across content sites — "$2,900" is variously the product price, a market anchor, a competitor's price, and a word count. Per-page judgment required.

🧲 Magnetics sites + worksheet fixes

• Registered maxelmag.com + codedmag.com (GoDaddy, ~$10.69 each); built both as distinct sites (maxel/pixel field-instrument vs coded-magnets signal spec-sheet); cross-linked the 4 magnetics sites (multipolemagnets, multipolemag, maxelmag, codedmag) into a network.
• deptmatic.com/all/ + /marketingsites/ worksheet: refined "Topic" from generic "page" to real per-page topics; fixed the Act column sort (numeric rank, was a string/number mix — fix is in shared /root/gen-alldomains.py, benefits wholetech.com/alldomains too); made /marketingsites/ domains-only; per Paul, reverted the # column to a clean sequential counter.
• Removed the developer-capital ($250K/$70K "skin in the game") paragraphs + card from deptmatic.com/backend/briefing/.

🔐 Secure admin control panel

• New panel to track users + full edit/deploy of the 10 product sites: deptmatic.com/admin/ (basic-auth). Backend /opt/admin-panel/admin.py (systemd admin-panel.service, 127.0.0.1 only).
• Layered security: basic-auth on UI + API, loopback-only backend, shared-secret proxy header, realpath-confined to each webroot (path-traversal rejected), 10-site whitelist, backup-before-write, never deletes. Independently verified: 401 without creds, traversal blocked. (Credentials in the private session record, not here.)

📋 What changed — complete list

🔎 Verified

🧭 Decisions & why

• $499 founding rate (Paul's call): strongest per the research — undercuts a DIY tool stack, owns the "small business everyone priced out" wedge, reads as an intentional discount vs ~$2,900 market.
• Central admin panel (not one login per site): fewer doors to secure; still targets all 10 webroot-guarded.
• /connect/ "we host" links to the central form vs per-site forms: economical, one working intake, consistent with the report CTA.
• Kept a few /transcript/ build-log diaries at $2,900: they narrate the research where $2,900 was the working number — repricing would falsify the historical account.
• Did NOT set any API key / API billing anywhere — the free audit + activate reuse the existing engine (subscription-only rule intact). Tim's $500 token offer = reimbursement, not API billing.