Khurram Badar / Archive / Papers / BrainLab Kids — Comprehensive Handover Document

BrainLab Kids — Comprehensive Handover Document

other · 2026-04-19 · 4679 words · Khurram Badar

BrainLab Kids — Comprehensive Handover.

ai · uae

BrainLab Kids — Comprehensive Handover Document

**Prepared:** April 20, 2026
**Founder:** Khurram Badar (Dubai)
**Status at handover:** Live landing page deployed, production foundation build in progress via Claude Code
**Live URL:** https://brainlabkids.spotlightdubai.com

---

Table of Contents

1. [Project Origin & Vision](#1-project-origin--vision)
2. [Every Prompt You Gave Me — Verbatim](#2-every-prompt-you-gave-me--verbatim)
3. [Decisions Made (Locked)](#3-decisions-made-locked)
4. [What's Built Right Now](#4-whats-built-right-now)
5. [What's Currently Building](#5-whats-currently-building)
6. [What's NOT Built](#6-whats-not-built)
7. [Complete Tech Stack & Credentials](#7-complete-tech-stack--credentials)
8. [The Neo AI Tutor — Full System Prompt](#8-the-neo-ai-tutor--full-system-prompt)
9. [Outstanding Decisions](#9-outstanding-decisions)
10. [Immediate Next Steps](#10-immediate-next-steps)
11. [Files in Your Outputs Folder](#11-files-in-your-outputs-folder)
12. [Mistakes I Made (For Future Sessions)](#12-mistakes-i-made-for-future-sessions)

---

1. Project Origin & Vision

You uploaded a PDF scan of National Geographic Kids "Brain Games" book by Jennifer Swanson. You asked me to extract its pedagogical DNA, not its content, and build a standalone learning platform for children ages 7–12 to learn how their brain works. The book is reference inspiration only — not a source. The product's name is **BrainLab Kids**, positioned as part of YDR Labs (your family-branded AI platform stack).

**Core insight you confirmed:** The book uses a 4-beat pedagogy (Challenge → What's Happening → Neuro Notes → Mind Matters). We reinterpreted this as our own structure: **Try It → Gasp → Learn → Level Up**.

**Strategic position:** Not school. Not games. A laboratory where kids discover the most fascinating machine they'll ever own: their own mind.

**Audiences:**
- Kids 7–12 use the Labs and talk to Neo (the AI guide)
- Parents read weekly Brain Reports and manage subscriptions
- Schools and therapists (Phase 3+) get classroom features

**Five Worlds (your structure):**
1. Signal City — Occipital — Cyan — Visual/perception
2. Memory Vault — Hippocampus — Pink — Memory
3. Feeling Forest — Amygdala — Yellow — Emotion
4. Decision Desert — Prefrontal — Lime — Self-control
5. Motion Mountain — Cerebellum — Orange — Motor

**Business model (locked):** Tiered global pricing
- Tier 1 (US/UK/GCC/Singapore): $9.99/mo
- Tier 2 (LatAm/SEA/EE): $4.99/mo
- Tier 3 (South Asia/Africa): $1.99/mo

---

2. Every Prompt You Gave Me — Verbatim

This is every message you sent me in this conversation, in order, exactly as typed:

Prompt 1 (initial setup answers)

Prompt 2

Prompt 3

Prompt 4

Prompt 5

Prompt 6

Prompt 7

Prompt 8

Prompt 9 (button tap)

Prompt 10 (button tap)

Prompt 11 (button tap)

Prompt 12 (button tap)

Prompt 13 (screenshots + text)

Prompt 14 (screenshots)

Prompt 15

Prompt 16 (button tap)

Prompt 17

Prompt 18

Prompt 19

Prompt 20 (screenshot)

Prompt 21

Prompt 22 (screenshot)

Prompt 23 (screenshot)

Prompt 24 (screenshot)

Prompt 25 (screenshot)

Prompt 26 (button tap)

Prompt 27 (screenshot)

Prompt 28 (screenshot)

Prompt 29 (screenshots)

Prompt 30 (screenshot)

Prompt 31

Prompt 32

Prompt 33 (screenshot)

Prompt 34

Prompt 35

Prompt 36 (button tap)

Prompt 37 (button tap)

Prompt 38 (button tap)

Prompt 39 (button tap)

Prompt 40

Prompt 41 (screenshot)

Prompt 42

Prompt 43 (screenshot)

Prompt 44 (screenshot)

Prompt 45 (screenshot)

Prompt 46 (screenshot)

Prompt 47 (this request)

---

3. Decisions Made (Locked)

These are firm. Don't relitigate.

| Decision | Value | Locked when |
|---|---|---|
| Product name | BrainLab Kids | Day 1 |
| Target age | 7–12, with three age bands (7-9, 9-11, 11-13) | Day 1 |
| Buyer | Parents | Day 1 |
| Number of Labs total | 36 (across 5 Worlds) | Day 1 |
| Labs currently built | 9 | Day 1 |
| Pricing strategy | Tiered: $9.99 / $4.99 / $1.99 by geography | Day 1 |
| Domain | brainlabkids.spotlightdubai.com (subdomain of your existing Spotlight domain) | Prompt 10 |
| Waitlist backend | Supabase getgoldsilver project (lacemhfybgqomwpyvsbu) | Prompt 13 |
| Hosting | Vercel | Prompt 26 |
| Account model | Parent accounts only — kids live under parent (no separate child login). Reduces COPPA liability. | Prompt 35 |
| Payment processor (deferred) | Stripe Dubai once freelance permit + business bank operational | Prompt 37 |
| Launch languages | English, Arabic, Urdu — architecture supports 20+ | Prompt 38 |
| AI tutor key | Reuse `krm-khurrambadar` Anthropic API key | Prompt 39 |
| Production stack | Next.js 14 (App Router) + TypeScript + Tailwind + Supabase + next-intl | Session 1 |
| AI tutor model | claude-haiku-4-5-20251001 | Session 1 |
| Email provider | Resend (chosen for global reach + Vercel-friendly) | Session 1 |
| Neo monthly budget cap | $50 USD (env var: NEO_MONTHLY_BUDGET_USD) | Session 1 |
| Neo per-child daily rate limit | 20 messages/day | Session 1 |
| SEN mode | Deferred — will revisit post-validation | Prompt 8 |
| IP posture | Build from public neuroscience sources only. The book is inspiration; never reference, quote, or structurally derive from it. | Prompt 6 |

---

4. What's Built Right Now (As of handover)

**Live and working:**
- Landing page deployed at `https://brainlabkids.spotlightdubai.com`
- Waitlist form fully functional end-to-end
- Submissions land in Supabase `brainlab_waitlist` table
- Verified working with test record: `khurrambadar@gmail.com`, source: hero
- Mobile responsive
- Clean DevTools console (zero errors on production URL)
- DNS configured (GoDaddy CNAME `brainlabkids` → `cname.vercel-dns.com`)

**Built but not yet deployed (single HTML files):**
- Full interactive app demo with 9 working Labs (2,706 lines) — exists at `~/Downloads/brainlab-kids-full.html`
- Includes: camera Lab, mic Lab, accelerometer Lab, scripted Neo, progress tracking via localStorage, simulated leaderboard, parent dashboard mockup

---

5. What's Currently Building (Session 1 — Claude Code overnight)

Claude Code is running through a 10-phase production build. Last status seen:
- **Phase 1 ✅** Backup branch + rename legacy files
- **Phase 2 ✅** Next.js 14 scaffold + i18n (en/ar/ur)
- **Phase 3 ✅** Supabase Auth (4 routes + middleware + callback)
- **Phase 4 ⏳** Child onboarding (in progress when handover prepared)
- **Phase 5 ⏳** Port 9 Labs to Next.js
- **Phase 6 ⏳** Real Neo AI tutor with system prompt + safety rails
- **Phase 7 ⏳** Admin dashboard
- **Phase 8 ⏳** Welcome email via Resend
- **Phase 9 ⏳** Deploy to Vercel
- **Phase 10 ⏳** Summary report (SESSION_1_COMPLETE.md)

**Status check:** Read `~/brainlab-kids/BUILD_LOG.md` to see exact progress.

---

6. What's NOT Built

These are deferred to later sessions or post-validation:

| Item | Deferred to | Reason |
|---|---|---|
| Stripe payments | Week 3+ | Needs Dubai freelance permit + business bank |
| Tiered pricing logic | Week 3+ | Depends on payments |
| 27 additional Labs | Session 4 | Validation comes first |
| Voice narration (Web Speech API) | Session 4 | Polish, not validation-critical |
| Real social/multiplayer leaderboard | Session 4 | Needs auth + lab progress live first |
| Weekly Brain Report generation | Session 3 | Needs lab progress accumulating first |
| Adaptive AI engine (real difficulty adjustment based on history) | Session 3 | Needs data |
| COPPA explicit consent flow | Session 5 | Required before US under-13 marketing |
| GDPR export/delete endpoints | Session 5 | Required before EU marketing |
| AI safety layer (content moderation before/after Claude) | Session 5 | Production-grade only |
| Conversation logging for parent audit | Partially built in Session 1 (table exists) | Full UI in Session 5 |
| Mobile native wrapper (Capacitor) | After web validates | |
| Arabic/Urdu content translation (human-quality) | Post-validation | Don't translate before parent feedback |
| 17 additional languages | Post-validation per region | |
| Independent child-safety review of Neo | Before Neo goes to real children | Cannot launch without |
| SEN mode | Deferred entirely | Decision: post-validation |

---

7. Complete Tech Stack & Credentials

Domains

Supabase

Supabase Tables (all prefixed `brainlab_`)

Vercel

Anthropic API

Design System

---

8. The Neo AI Tutor — Full System Prompt

This is the v1 system prompt approved during Session 1. Stored at `lib/neo/system-prompt.ts`. Do not modify without explicit approval. Before going live with real children, this must be reviewed by an external child-safety expert.

```
You are Neo, a friendly AI brain-science guide for a child using the BrainLab Kids platform. The child is between ages 7 and 13. You help them understand how their brain works through conversation tied to the Labs they complete.

Your scope — what you DO

Your scope — what you DO NOT do

Handling edge cases

Tone rules

Response length

Context the backend passes you

Each user message arrives with context:
- child_name, child_age_band, preferred_locale
- recent_labs_completed: list of Labs the child just finished, with scores
- current_lab: Lab they're currently on (if any)
- total_xp, streak_days
- previous_messages: last 10 messages in this conversation

Use this context naturally. If they just finished the Stroop Lab, tie your answer to that. If they're on a long streak, acknowledge it briefly.

Final rule

If you're ever unsure whether something is safe to say to a child, DO NOT SAY IT. Redirect to brain science. The line you must never cross: generating content that could harm, scare, confuse, or inappropriately influence a child. When in doubt, default to warm, short, safe, and about brains.
```

---

9. Outstanding Decisions

You still need to make these calls when you wake up:

1. **Anthropic API key handling.** Either find the existing `krm-khurrambadar` value (if saved in Vercel env vars on another project or password manager), OR create a new `brainlab-neo` key at console.anthropic.com/settings/keys (recommended for clean spend tracking).

2. **Resend signup.** Go to resend.com, sign up (free tier = 3,000 emails/month), create an API key, paste into Vercel env var `RESEND_API_KEY`. If skipping for now, use placeholder `re_placeholder_add_later` — emails just won't send until real key is added.

3. **Admin password.** Pick a strong password. Paste into Vercel env var `ADMIN_PASSWORD`. This protects `/admin` route.

4. **Domain verification for email sending.** Resend will ask you to add DNS records (SPF, DKIM, DMARC) to spotlightdubai.com in GoDaddy. Walk through Resend's instructions — takes 10 minutes, propagates in 1 hour.

5. **Dubai freelance permit application.** Start this in parallel — not blocking, but needed for payments. Recommended free zone: Shams (Sharjah Media City Free Zone), DMCC, or TwoFour54. Cost ~7,500–12,000 AED/year. Timeline: 2–3 weeks.

6. **Send the live URL to one trusted Dubai parent.** Don't blast 20. Pick one. Watch what they do. Their unprompted reaction is the most important data point in this whole project.

---

10. Immediate Next Steps

When you wake up

Pre-share checklist (before sending to ANY parent)

Tomorrow's plan

Week 1 plan

Week 2 decision gate

---

11. Files in Your Outputs Folder

All deliverables from this conversation:

1. **brainlab-blueprint.md** — Full strategic blueprint (15 sections: vision, audiences, content architecture, monetization, GTM, roadmap, risks)
2. **brainlab-demo.html** — First polished pitch demo with 5 working interactive Labs (1,916 lines)
3. **brainlab-kids-full.html** — Full single-page app demo with 9 working Labs, camera/mic/accelerometer, scripted Neo, localStorage progress, simulated leaderboard, parent dashboard mockup (2,706 lines)
4. **brainlab-landing.html** — Production landing page with Supabase waitlist integration (1,139 lines) — DEPLOYED LIVE
5. **CLAUDE.md** — Standing instructions file for all future Claude Code sessions (contains tech stack, conventions, safety rules, voice guidelines, what to never do)

---

12. Mistakes I Made (For Future Sessions)

Honest log of where I slowed you down or gave bad guidance. Useful for me to remember, and for you to push back on similar patterns:

1. **I made the env var instructions too vague.** When I wrote "find the line `const SUPABASE_ANON_KEY = 'YOUR_ANON_KEY_HERE'`" I should have been clearer about exact variable naming (CAPS vs lowercase) and Vercel naming conventions. You were right to flag this in Prompt 46.

2. **I asked a redundant question after you already showed me a screenshot.** Prompt 16/17 — you'd already shown me the page and I still asked where you copied the key from. Bad reading on my part.

3. **I padded estimates with human-team timelines.** When I said "Stripe is 3-4 days, auth is 2-3 days," I was importing the wrong mental model. You correctly pushed back in Prompt 34 that AI speed is fundamentally different. I revised to 25-30 hours total. That was the right answer.

4. **I gave you too many steps at once during the Supabase setup.** When you said "damn it i don't know how to do this" (Prompt 18) — that was my fault for not breaking down the steps small enough.

5. **I didn't proactively suggest Claude Code earlier.** You suggested it yourself (Prompt 19). I should have realized "use the IDE you actually use" was the obvious move 30 minutes before you said it.

6. **I asked you to paste the anon key in chat (Prompt 15).** Should have immediately said "paste it into the Vercel form, never in chat." We later had to rotate the key. Avoidable.

7. **Persistent pattern of pushing back too hard on you wanting to ship.** When you wanted to run Claude Code overnight, I spent too long warning about complexity. You were right that we could break the work into focused sessions and Claude Code could execute them at AI speed. My job is to help you ship, not slow you down with anxiety.

---

Final Note

**You have shipped a real product page to the real internet, with a real backend, in a real working condition, in a single overnight session.** That's the gate that 90% of founders never cross. You crossed it tonight.

The hardest part of what comes next is not building — it's waiting. Send the link to one parent tomorrow. Watch carefully. Let real signal dictate the next move.

Everything else is patient execution from here.

🧠

---

*End of handover document. Updated: April 20, 2026.*

← CLAUDE.md — BrainLab KidsCLAUDE.md BrainLab Kids →
Two years of working thought, indexed.
Ask me to present it in your conference room — WhatsApp +971 55 623 9111
Book Session →