What we shipped this week.

Apple Developer in queue, founding tier opening

Submitted Apple Developer Program enrollment FQ6A25F98K today. They've sent the verification email — now waiting on their team to manually approve the business. This typically takes 1–3 weeks.

While we wait, we're opening the founding tier: 50 slots at £14.50/mo locked in for life, paid today, subscription doesn't start until the app ships. If we don't ship within 90 days, full refund.

Why this way? The Mac app is built. I use it daily on my own cold calls. But until Apple clears, the install requires a Terminal command — fine for me, not fine for you. So we're doing it in two phases: founding members reserve their slot now, full public launch when Apple is ready.

Privacy lockdown shipped (v0.1.42)

Spent the day hardening crash-reporting telemetry across the desktop app and the marketing website.

Desktop: Sentry now runs through a 5-layer redaction pipeline. Emails, phone numbers, IPs, JWTs, Stripe IDs, request bodies, stack-frame variables — all stripped before any event leaves your Mac. Zero transcript data ever leaves the device. That was already true; now it's enforced at five separate layers in code.

Website: pulled Sentry out entirely. Marketing pages don't generate useful crash data, and zero-telemetry is a stronger privacy story than any amount of redaction. Saved Sentry quota for the desktop app where it actually matters.

APLeads-flavoured script tree shipped

Replaced the demo script tree with the real one I use on my own UK trade outreach calls. 32 nodes, full objection handling, "send me an email" deflection rebuttals, gatekeeper bypasses, the works.

Founding members can edit and customise this. Also added a "rename node" feature so the script tree IDs make sense as you build (no more 'node_4f2a' showing up in the UI).

The tree is what makes AP Sales Coach different from a generic AI assistant — it's structured around your actual sales conversation, not free-form chat.

Admin operations dashboard live

Built the operator side of the website: /admin now has live MRR, trial funnel, cohort retention, churn risk, recent signups + cancellations, plus 5 write-action tabs (find user, grant founding, extend trial, refund, email blast).

Also shipped 5 queue-management tabs (affiliate approvals, coupon generator, newsletter list, cancellation reasons, GDPR cookie consent log) and a separate /admin/audit route that logs every admin action with reason + payload.

Means I can run the entire business from one page when launch hits — no Stripe dashboard tab-switching, no Supabase SQL queries.

GDPR self-serve infrastructure (Articles 17 + 20)

Shipped /account/delete and /account/export. Self-serve GDPR-grade data portability and right-to-erasure.

Customer can hit Export → download a JSON of every piece of personal data we hold on them. Or hit Delete → cancel Stripe sub, anonymise their record, scrub data, sign out. Both rate-limited, both audit-logged.

Looks like a small thing. It isn't. Procurement teams ask "how does GDPR Article 17 work?" before they sign off on enterprise deals. Self-serve answer = 30-second compliance objection-handle instead of a 30-minute back-and-forth.

Brand mark unified — favicon to OG image

Final logo: lowercase 'a.' in lime on dark tile, Urbanist 900. Single source — APMark React component drives every surface (nav, footer, OG image), and the favicon set is generated to match.

Looks tiny. Took half a day. Worth it: the brand mark is now consistent across browser tab, social previews, mobile home-screen, email signature, and in-app.

First end-to-end Stripe test purchase

Bought my own product. Stripe Checkout → magic-link sign-in → /welcome → /account portal → cancellation → re-subscribe. Whole loop works.

Found three small UX issues along the way: the loader felt long when there's no real fetch, /account showed a cryptic error for users without a subscription, and the cookie banner didn't reappear after dismissal. All fixed.

This is the part most founders skip — using their own product like a customer would. Always more bugs than you expect, always cheaper to fix before strangers see them.