KO | EN

GitLite

Minimalist Open Source Repository Search Engine

gitlite — search

#dsh #dsh-plugin #developer-tools #ai-agent #claude-code #codex #typescript #ai-agents #deepseek-harness #mcp #cli #rust

biological-twin

★ 12 Open GitHub ↗

Meet your biological twin — a digital mirror born from your DNA that explains you to yourself. Built with Opus 4.6 for the Claude Code Hackathon.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Biological Twin

**Meet your biological twin — a digital mirror born from your DNA that explains you to yourself.**

Your genome is the longest text ever written about you. 3.2 billion characters, edited continuously for 3.8 billion years. Until now, you couldn't read it. We gave it to the first mind capable of actually reading it — Opus 4.6.

Upload your raw DNA file. Your twin reads your genome thematically — not individual SNP lookups, but deep reasoning across hundreds of variants to find convergent patterns. It discovers themes in your biology and connects them to your lived experience. Then you go deeper: ask anything, explore your possibility space, and understand why you are the way you are.

**This is not health analysis. It's self-knowledge through biology.**

https://github.com/user-attachments/assets/placeholder

## How It Works

1. **Upload** — Drop your 23andMe or AncestryDNA raw data file
2. **Parse** — Your genome is parsed entirely client-side. It never leaves your browser.
3. **Read** — Opus 4.6 reads your variant profile with extended thinking, finding themes across multiple pathways
4. **Discover** — Themes bloom onto a 3D DNA helix as they're discovered, in real time
5. **Explore** — Three depth levels per theme: your twin's personal insight, accessible science, and full evidence essays
6. **Converse** — Ask your twin anything about your biology. It holds your full genetic context.

## The Reading

Every genetics tool treats your DNA as a medical record. We treat it as a text — the most personal text in existence — and interpret it.

Opus finds **themes** in your genome: not "you have variant X," but "five variants across three pathways converge on the same pattern — here's what that means for how you experience the world." Every theme is backed by 3+ convergent variants with evidence tiers and citations.

Three voices for each theme:
- **Twin voice** — "This is why you've always felt things intensely." The recognition moment.
- **Street level** — Accessible science. Gene names explained, pathways made intuitive.
- **Deep dive** — Full essay with specific variants, effect sizes, evidence tiers, and citations you can bring to your doctor.

## Why Opus 4.6

This product could not exist before Opus 4.6:

- **Extended thinking** — Cross-variant reasoning requires genuine deliberation. Finding that five methylation variants interact with your COMT status isn't a lookup — it's interpretation.
- **1M context window** — Holds your full significant variant profile + research literature + conversation history simultaneously.
- **Opus-level intelligence** — Weaving genetics, neuroscience, pharmacology, and nutrition into a coherent personal narrative requires the deepest reasoning available.

## Tech Stack

- **Next.js 16** + React 19 + TypeScript
- **React Three Fiber** + drei + postprocessing — 3D DNA helix visualization
- **Zustand** — State management
- **PapaParse** — Client-side genome parsing
- **Anthropic SDK** — Opus 4.6 streaming with extended thinking, Haiku for post-processing
- **Tailwind v4** + shadcn/ui — UI components
- **Motion** (Framer Motion) — Cinematic transitions

## Architecture

```
Browser (client-side)          Server (API routes)
┌─────────────────────┐        ┌──────────────────────────┐
│ DNA file upload     │        │ /api/reading             │
│ PapaParse parsing   │───────▶│   Opus 4.6 + 24K thinking│
│ SNP reference match │        │   SSE streaming          │
│ (genome never sent) │        │                          │
│                     │◀───────│ /api/reading/process     │
│ 3D helix (R3F)      │        │   Haiku post-processing  │
│ Theme cards + pins   │        │   Theme extraction       │
│ Journey navigation  │        │                          │
│ Conversation UI     │───────▶│ /api/conversation        │
│                     │        │   Opus + genome context   │
└─────────────────────┘        └──────────────────────────┘
```

**Privacy model:** Your full genome is parsed in the browser. Only the relevant SNP subset (~300 matched variants out of 600K+) is sent to the API for Opus to reason about. Nothing is stored server-side.

## Setup

### Prerequisites
- Node.js 18+
- An [Anthropic API key](https://console.anthropic.com/) with access to Opus 4.6

### Install & Run

```bash
cd app
cp .env.example .env.local
# Add your ANTHROPIC_API_KEY to .env.local

npm install
npm run dev
```

Open [http://localhost:3000](http://localhost:3000).

### Sample Data

Sample genome files from the [Harvard Personal Genome Project](https://pgp.med.harvard.edu/) are included in `data/sample-profiles/` for testing. Drop any of these into the upload zone to generate a full reading (requires an Anthropic API key and takes 2-3 minutes as Opus reasons through the genome).

### Demo Mode

To see the full experience without an API key or the wait for Opus:

- Click **"Try a demo reading"** on the landing page, or
- Visit `http://localhost:3000?demo=true` to auto-start

Demo mode replays a pre-generated reading through the same visual pipeline — helix assembly, theme blooming, crystal drops, and journey navigation all work identically to a live reading. The full post-reading experience (street view, deep dives, essay mode) is available after the replay. Note: interactive features like asking your twin questions and discovering more themes do require an API key.

The demo session data lives in `app/public/demo/session.json`. To generate a new one: run a real reading with any genome file, then export from the browser console:

```js
copy(localStorage.getItem("biological-twin-session"))
```

Paste the result into `app/public/demo/session.json`.

## Project Structure

```
app/
├── src/app/page.tsx              # Main page (landing → parsing → reading)
├── src/app/api/
│   ├── reading/route.ts          # Opus SSE streaming reading
│   ├── reading/process/route.ts  # Haiku theme extraction
│   ├── reading/expand/route.ts   # Discover more themes
│   └── conversation/route.ts     # Twin conversation
├── src/components/
│   ├── scene/                    # 3D: ComposableHelix, CrystalPins, VaporText
│   └── ui/                       # ReadingView, ConversationView, JourneyNav
├── src/hooks/                    # useGenomeReading (SSE), useConversation
├── src/lib/                      # store, types, dna-parser, parseReading
└── public/snp-reference/         # 301 variants across 10 themed JSON files
```

## How We Built It

Built in 6 days by b (design, product, creative direction) and Claude (architecture, implementation, reading engine) using Claude Code. The process is documented in:

- `HOW-WE-GOT-HERE.md` — Origin story
- `meta/PROCESS-LOG.md` — How we used Claude Code and Opus
- `PROGRESS.md` — Daily build log

## Credits

- **Fonts** — [Inter](https://rsms.me/inter/), [Space Grotesk](https://fonts.floriankarsten.com/space-grotesk), [JetBrains Mono](https://www.jetbrains.com/lp/mono/) — OFL 1.1
- **Sound effects** — Generated with [ElevenLabs](https://elevenlabs.io/)
- **Sample genomes** — [Harvard Personal Genome Project](https://pgp.med.harvard.edu/) — CC0

## Built For

**[Built with Opus 4.6: A Claude Code Hackathon](https://anthropic.com)** — February 10–16, 2026

## License

MIT

biological-twin

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README