Launch Briefing Prototype

Source: starlight-md-prototype.netlify.app Auth (HTTP basic): starlight / CrushAtlas2026! Explored: May 7, 2026 Author: Deepak Surana (Builder/CPO) Repo: dsurana04/starlight-md (private)

What this prototype is

A four-artifact launch briefing positioning Starlight.MD as the explicit replacement for Atlas.MD, the incumbent DPC pediatric EMR. It is not the same as the earlier March prototype — it is a substantially larger, planning-grade artifact that ships:

1. A single-page launch briefing (/) — pitch, prototypes, 7-step tour, build plan, stack, doc library, asks of Erik.
2. A clinician web app (/visit/) with 12 navigable views, 11 chart tabs, two AI-scribe state machines, AI assistant panel, ⌘K command palette.
3. An iPhone-sized parent enrollment flow (/onboarding/) — 8 screens from email capture to first-visit booking.
4. A 10-dimension head-to-head vs Atlas.MD (/vs-atlas/) with quantified ROI for Yogini and the broader DPC TAM.

Every page carries a CONFIDENTIAL · FOR ERIK & VICTOR stamp.

Why it matters for engineering

The hardcoded data inside /visit/ defines Deepak's intended v1 schema:

• Family + Guardian + Patient as a three-table model (one Stripe customer per family, not per patient).
• 5 pre-seeded SOAP notes with section-tagged provenance (AI / auto / Yogini) — the contract for the note editor.
• Two parallel scribe state machines (office 1→5, newborn home 1→3) — the contract for AI ambient scribe.
• 6 canned AI_RESPONSES keyed by question — the quality bar for production Claude calls.
• 8 detail-drawer modals for ICDs, Rx, immunizations, SMS threads — the contract for chart drill-downs.
• 22-item ⌘K command palette — the v1 keyboard-shortcut surface.

These are not demo dressing. They are a planning surface to build against.

Section map

📄️ Clinician App

The desktop / iPad-class clinician surface. 12 navigable views, 11 chart tabs, two AI-scribe state machines, persistent AI assistant panel, universal ⌘K command palette.

📄️ Parent Onboarding

The 8-screen mobile enrollment flow that turns a marketing-site visitor into a paid Starlight family. Wrapped in a CSS iPhone bezel; only renders correctly at narrow viewport. State machine driven by goto(n).

📄️ vs Atlas Comparison

A single-page, marketing-grade head-to-head between Starlight.MD (2026) and Atlas.MD (2014). Hero stats, 30-second pitch, 10 numbered dimensions, a side-by-side chart screen, and bottom-line ROI for both Yogini specifically and the broader DPC + concierge category.

📄️ Hardcoded Data Models

The JS literals embedded in the prototype's /visit/ source define the v1 schema Deepak has in mind. Treat them as the planning surface for our DB models, not as demo dressing.

📄️ Design System

Extracted from the prototype's :root rules. Light-only — no dark-mode variant. A single typeface (Nunito) is used across all four artifacts.

Key links inside the briefing

• Build plan: Sprint 0 → Sprint 7+, ~7 months to Yogini cutover.
• 24 epics linked to GitHub issues (dsurana04/starlight-md/issues/{1..24}), each tagged with a sprint label.
• 9 doc-library cards linked to repo paths — MVP definition, Atlas teardown, v1 priorities, Record/docs audit, Visit flow spec, 18-patient roster, Chart design rationale, 4 QA agent reports, Epic backlog.
• 12 opinionated stack picks: Postgres+Drizzle (Neon BAA), Next.js 15 + React 19, Clerk + WebAuthn, Expo, DoseSpot, Stripe Billing, Twilio, LiveKit, Claude Sonnet + Whisper, Health Gorilla, Vanta, AWS + Vercel.

Asks of Erik (from the briefing's footer)

1. Approve the v1 scope — confirm we never build insurance / claims / RCM.
2. Pick the build pod size — 2 engineers + designer = 9 mo · 4 engineers + designer = 6 mo.
3. Confirm tech-stack picks — DoseSpot vs Rcopia · Clerk vs Auth.js · Anthropic vs OpenAI · LiveKit vs Twilio Video.
4. Pilot terms with Yogini — free during build → $0 first year, or paying customer from M7.
5. Add Victor to GitHub.
6. Brand polish — wordmark + 3 marketing screenshots.