What's shipping next.

Pure-TypeScript rules engine encoding 14 CFR §61.23 / §61.56 / §61.57(a/b/c) / §91.409(a/b) / §61.31. 160 unit tests. AI brief layer via Claude (rule-based fallback when key isn't set).

Add, edit, remove. IFR currency editor. Day/night currency tracking. Per-aircraft squawk management with severity tiers (advisory / MEL / grounding). Persists in browser localStorage.

NO-GO can be overridden with typed chief-pilot signature + reason. Every override logged forever. CSV-exportable audit history (16 columns).

Print-friendly /briefing dashboard. /demo route ships 3 hard-coded scenarios (GO / HOLD / NO-GO) so visitors can see the engine work without setting up data.

Live at auto-brief-web-pied.vercel.app since 2026-05-28. Auto-deploys on every push to master. README badges, sitemap, og:image, full SEO meta.

Flip ANTHROPIC_API_KEY env var in Vercel → AI-generated briefs replace the rule-based fallback. Prompt caching already configured. Roughly 1 minute of work, blocked on the credentials being committed to the environment.

Vercel-assigned URL works fine but autobrief.app is the long-term home. ~$9 / year + a DNS hop. Replaces the per-Vercel-deployment alias across the site and journal references.

Migrations 0001-0005 are written (schools + dispatchers + pilots + aircraft + squawks + checks + override_log with RLS isolation and SECURITY DEFINER RPCs). Blocked on Supabase Pro upgrade to clear the free-tier 2-project cap. Drops in cleanly behind the localStorage adapter — no UI changes required.

First-customer onboarding accelerator. Drop a roster CSV in, get a populated school. Bridge between v1's manual entry and v2's FSP API.

Daily digest to chief pilot when something in the school's roster crosses the 30-day-out window. Resend integration, same pattern as Trim's cron.

Pull roster + bookings directly from the school's existing scheduler. Currently manual entry per the v1 honesty rule. Defer until a paying customer confirms which scheduler they use.

First 3 schools to sign up get founder pricing locked for life. Plumbing exists in TarmacLabs already (Trim pattern). Defer until cold email reply #1 says 'let's do it.'

v1 is responsive web on phones, which works but isn't optimal at the FBO desk. Capacitor wrapper of the existing Next.js app is the cheapest path. Defer until first paying school has used the web version for 30 days and asks.

Hobbs/Tach predictions, cost-per-hour by tail, predictive 100-hr scheduling. Was in the original 3-module charter, cut on 2026-05-02 to keep MVP focused on GO/NO-GO. Sold as an expansion to the first paying school, not built into MVP.

Hours flown vs. available, no-show rates, instructor revenue-per-hour. Original 3-module charter scope. Defer or drop based on what the first paying school actually asks for.