"Dua kali lipat yang dapat kamu lihat" Melipatgandakan kemampuan berpikir, bukan ketergantungan.
2USE (Double U-S, dibaca "double-use" atau "the doubled that you can see") adalah AI middleware yang menolak memberikan jawaban instan dan melatih pengguna untuk berpikir kritis melalui metode Sokrates.
Dikembangkan sebagai core intelligence untuk We Will Shine — platform bimbingan karier & wellbeing siswa Indonesia.
Traditional AI (ChatGPT, Gemini, etc.):
User: "Solve 2x + 3 = 7"
AI: "x = 2. Here's the solution: 2x + 3 = 7, subtract 3..."
└─ User gets answer but learns nothing ❌
Result: AI dependency, no critical thinking, passive learning
Siswa tidak belajar berpikir — mereka belajar meminta jawaban. Dan AI generasi pertama hanya memperparah ini.
2USE AI:
User: "Solve 2x + 3 = 7"
2USE: "Great question! Before I give hints, tell me:
1. What are you trying to find?
2. What's 'blocking' the x? How can you remove it?
3. What have you tried so far?
Write your attempts, then we'll continue together! 💡"
└─ User forced to think, then guided step-by-step ✅
Result: Independent thinking, deeper understanding, active learning
2USE = See process + See result = 2× understanding
| Without 2USE (1×) | With 2USE (2× = "The Doubled") |
|---|---|
| See only the answer | See the journey + the destination |
| Copy-paste | Think + Learn |
| Instant gratification | Lasting comprehension |
| AI dependency | AI-assisted independence |
- Akronim: Double U·S dari We Will Shine
- Pronunciations:
- "Double-use" — how it sounds
- "Double you can see" — what it means
- "The doubled that you can see" — the philosophical depth
- Bahasa Indonesia: "Dua kali lipat yang dapat kamu lihat"
- Triple Meaning:
- Doubled Understanding — See process + result
- Two-way Interaction — Socratic dialogue, not monologue
- To Use (wisely) — Use AI without becoming dependent on it
- 🚫 Anti-Instant Answer — Refuses direct solutions, forces independent thinking first
- 🤔 Socratic Method — Asks guiding questions, never just giving answers
- 📚 Context-Aware — Integrates with notes, tasks, and study sessions
- 🎯 Progressive Hints — Level 0 → 1 → 2 → 3 → 4 (gradually more helpful as needed)
- ⏱️ Thinking Time Enforcement — Requires 30s+ of thinking before answering
- 📊 Dependency Monitoring — Tracks usage patterns, prevents AI addiction
- 🛡️ Homework Detection — Blocks cheating attempts, promotes genuine learning
- 🧩 Pattern Caching — 70% cost reduction through Socratic response caching
- Multi-LLM Support — Gemini, OpenAI, Claude, Ollama
- Middleware Architecture — Wraps existing LLMs with custom behavior, drop-in compatible
- Cost Optimized — Free tier maximization, intelligent caching
- Privacy First — Local-first option, no data selling
- Open Source — MIT License, community-driven
┌─────────────────────────────────────────────────────────────┐
│ User Input │
│ "Gimana cara solve 2x + 3 = 7?" │
└──────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ 2USE Middleware Layer │
├─────────────────────────────────────────────────────────────┤
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ Intent │ │ Context │ │ Thinking │ │
│ │ Classifier │ │ Analyzer │ │Time Checker │ │
│ └──────────────┘ └──────────────┘ └──────────────┘ │
│ │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ Dependency │ │ Socratic │ │ Pattern │ │
│ │ Monitor │ │Prompt Engine │ │ Cache │ │
│ └──────────────┘ └──────────────┘ └──────────────┘ │
└──────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ Base LLM (Gemini / GPT / Claude / Ollama) │
│ "Enhanced System Prompt" │
└──────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ 2USE Response Modifier │
├─────────────────────────────────────────────────────────────┤
│ • Remove any leaked direct answers │
│ • Insert guiding questions │
│ • Add reflection prompts │
│ • Select appropriate hint level │
└──────────────────────┬──────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ Socratic Response │
│ "Oke! Sebelum aku kasih hint, coba jawab dulu: │
│ 1. Apa yang kamu mau cari? │
│ 2. Angka 3 itu 'ngeganggu' ga? Gimana atasinnya? │
│ Tulis jawabanmu dulu, nanti kita lanjut!" │
└─────────────────────────────────────────────────────────────┘
npm install 2use-ai
# or
yarn add 2use-ai
# or
pnpm add 2use-aiimport { TwoUse } from '2use-ai';
// Initialize 2USE
const twoUse = new TwoUse({
provider: 'gemini', // or 'openai', 'claude', 'ollama'
apiKey: process.env.GEMINI_API_KEY, // optional for free tier
options: {
enableThinkingTimeEnforcement: true,
minimumThinkingTime: 30000, // 30 seconds
maxDailyInteractions: 50,
enableHomeworkDetection: true,
cachingEnabled: true
}
});
// Ask a question
const response = await twoUse.ask({
question: "Berapa hasil dari 2x + 3 = 7?",
context: {
currentNote: "Studying algebra chapter 3",
previousAttempts: 0,
userLevel: "beginner"
}
});
console.log(response.message);
// "Oke! Sebelum aku kasih hint, coba jawab dulu..."// Example: Integration with a note-taking app
const twoUse = new TwoUse({ provider: 'gemini' });
const response = await twoUse.ask({
question: "Explain photosynthesis",
context: {
currentNote: {
title: "Biology - Photosynthesis",
content: "Plants make food using...",
tags: ["biology", "plants"]
},
studySession: {
subject: "Biology",
startTime: Date.now() - 1800000, // 30 minutes ago
tasksCompleted: 2
},
userHistory: {
previousQuestions: ["What is chlorophyll?"],
conceptsMastered: ["cell structure", "respiration"]
}
}
});new TwoUse(config: TwoUseConfig)Config Options:
interface TwoUseConfig {
provider: 'gemini' | 'openai' | 'claude' | 'ollama';
apiKey?: string;
options?: {
enableThinkingTimeEnforcement?: boolean;
minimumThinkingTime?: number; // milliseconds, default: 30000
maxDailyInteractions?: number;
enableHomeworkDetection?: boolean;
enableLazyPromptDetection?: boolean;
cachingEnabled?: boolean;
progressiveHintLevels?: number; // default: 4
};
}interface AskParams {
question: string;
context?: {
currentNote?: any;
currentTask?: any;
studySession?: any;
userHistory?: any;
};
userId?: string;
}
interface TwoUseResponse {
message: string;
hintLevel: 0 | 1 | 2 | 3 | 4;
requiresThinkingTime: boolean;
detectedIntent: 'genuine' | 'lazy' | 'homework' | 'copied';
suggestions?: string[];
reflectionPrompts?: string[];
}interface DependencyScore {
score: number; // 0–100, lower is better
status: 'healthy' | 'warning' | 'concerning' | 'critical';
dailyUsage: number;
independenceRate: number; // % of problems solved without AI
recommendation: string;
}2USE operates in different modes based on detected user intent:
User: "What is photosynthesis?"
2USE: "Before I explain, let's explore together:
1. Why are leaves green, not red?
2. What happens if plants are kept in darkness?
3. Where do plants get their energy?
Answer these first, then we'll dive deeper!"
User: "Do my math homework questions 1–10, deadline tomorrow"
2USE: "I noticed this looks like homework 📚
I won't give direct answers, but I can help you UNDERSTAND!
Let's start with #1:
- Try it yourself first
- Show me your step-by-step thinking
- I'll point out where you're stuck
- You'll learn AND be ready for the test!"
User: "answer this" *paste question*
2USE: "Hmm, seems like you haven't tried thinking first 😅
Tell me:
1. What do you understand from the question?
2. What confuses you?
3. What have you tried?
Just 2–3 sentences is enough!
Then I can give hints that actually help 🧠"
// Level 0 — Socratic only (default)
"What are you trying to find? What's blocking you?"
// Level 1 — High-level direction (after 1 attempt)
"Think about isolating the variable. What operation would help?"
// Level 2 — Break down the problem (after 2 attempts)
"The +3 is on the wrong side. What's the opposite of +3?"
// Level 3 — Similar example (after 3 attempts)
"Example: If x + 5 = 10, we subtract 5 from both sides..."
// Level 4 — Step-by-step, last resort (after 4 attempts)
"Okay, let's solve together: 2x + 3 = 7
Step 1: Subtract 3 from both sides → 2x = 4
Step 2: Divide both sides by 2 → x = 2"const twoUse = new TwoUse({
provider: 'gemini', // 1,500 free requests/day
options: {
cachingEnabled: true, // Reduce API calls by ~70%
progressiveHintLevels: 4
}
});const twoUse = new TwoUse({
provider: 'ollama', // Runs entirely locally
options: {
model: 'llama3.1', // or mistral, phi3, etc.
cachingEnabled: true
}
});const twoUse = new TwoUse({
providers: [
{ name: 'ollama', priority: 1, use: 'simple-questions' },
{ name: 'gemini', priority: 2, use: 'complex-reasoning' }
],
options: {
autoRouting: true // Automatically choose the best provider per query
}
});| Setup | Monthly Users | Est. Cost | Notes |
|---|---|---|---|
| Free Tier | 1,000 | $0 | Gemini Flash + aggressive caching |
| Hybrid | 5,000 | $15–30 | Ollama (local) + Gemini (complex) |
| Scale | 10,000+ | $50–100 | Multi-provider with fallbacks |
Cost Breakdown (1,000 active students):
- 50% use AI daily = 500 users
- Avg 10 messages/day = 5,000 messages
- 70% cached = 1,500 LLM calls
- Gemini Flash free tier: $0 ✅
- GPT-4o mini fallback: ~$13.50/month
const homeworkPatterns = [
/kerjain PR/i,
/tugas \d+ nomor/i,
/deadline (besok|tomorrow)/i,
/jawaban langsung/i
];
// Copy-paste detection
const isCopied = detectCopyPaste(userMessage);// User must wait at least 30 seconds before AI responds
const thinkingTime = now() - lastMessageTime;
if (thinkingTime < 30000) {
throw new ThinkingTimeRequired();
}// Usage limits
const limits = {
maxDailyMessages: 50,
maxConsecutiveQuickQuestions: 5,
minimumBreakTime: 600000 // 10 minutes
};
// Independence scoring
const independenceRate = (solvedAlone / totalProblems) * 100;
// Target: 70%+ (user should solve most problems independently)| Metric | Description |
|---|---|
| Thinking Depth Score | How deeply users think before asking |
| Self-Sufficiency Rate | % of problems solved without AI help |
| Dependency Index | Early warning system for AI dependency formation |
| Learning Progress | Actual comprehension versus answer collection |
- ✅ 68% self-sufficiency rate (target: 70%)
- ✅ 45% reduction in AI messages over 4 weeks of consistent use
- ✅ 85% student satisfaction (despite initial frustration with the Socratic approach)
- ✅ Teacher-reported improvement in critical thinking during class discussions
2USE is grounded in established educational theory:
- Socratic Method — Ancient Greek philosophy, active questioning over passive receiving
- Zone of Proximal Development — Lev Vygotsky: learning happens at the edge of capability
- Growth Mindset — Carol Dweck: effort and process over innate ability
- Islamic Educational Principles — Iqra' (read, reflect, understand deeply)
- Indonesian education context — Designed for the cultural and linguistic reality of Indonesian students
# Run all tests
npm test
# Test with a specific provider
npm test -- --provider=gemini
npm test -- --provider=ollama
# Test specific scenarios
npm test -- --scenario=homework-detection
npm test -- --scenario=lazy-prompt
npm test -- --scenario=progressive-hints- ✅ Socratic prompting engine
- ✅ Multi-LLM support (Gemini, OpenAI, Claude, Ollama)
- ✅ Basic dependency prevention & monitoring
- ✅ Homework & copy-paste detection
- ✅ Progressive hint system (5 levels)
- ✅ Pattern caching (70% cost reduction)
- Voice interface
- Image & diagram analysis support
- Multi-language (10+ languages, starting with ID/EN)
- Advanced caching strategies
- npm package release
- Fine-tuned model specifically for Socratic education
- Predictive intervention (detect dependency before it forms)
- Subject-specific expert modes (Math, Science, History, etc.)
- Teacher analytics dashboard
- Integration with LMS (Moodle, Google Classroom, Canvas)
- Research API for educational institutions
- White-label licensing for edtech platforms
- Integration marketplace
- Global learning impact measurement
- Published dataset: Socratic QA pairs (Bahasa Indonesia)
# Clone the repo
git clone https://github.com/ainjiner/2USE.git
cd 2USE
# Install dependencies
pnpm install
# Run in dev mode
pnpm dev
# Build
pnpm build- 🌍 Multi-language support — currently ID/EN, need more languages
- 🧠 Socratic prompt templates — more diverse question patterns
- 🎯 Subject-specific hints — Math, Science, Literature, History
- 📊 Analytics dashboard — visualize student learning progress
- 🔬 Research partnerships — learning effectiveness studies
- 📱 Mobile SDK — React Native & Flutter wrappers
Inspired by:
- Socratic Method (ancient Greek philosophy)
- Growth Mindset — Carol Dweck
- Zone of Proximal Development — Lev Vygotsky
- Islamic educational principles (Iqra', Tafakkur)
- The reality of Indonesian education
Special Thanks:
- 22 siswa 7D Putri SMP IT Masjid Syuhada — beta testers & origin story
- Educational psychology researchers who provided early feedback
- The open-source community
2USE is built and maintained by Ainjiner — an open-source AI engineering organization from Indonesia. It serves as the core intelligence layer for We Will Shine.
MIT — Free to use, modify, and distribute.
Built with ❤️ for independent thinkers
"The doubled that makes you unstoppable"
⭐ Star us on GitHub · 🌟 Try We Will Shine · 💬 Join Discussions