YOUR AI - learns you, acts for you
Not everything on your phone needs your eyes, mind or ears
Design philosophy : We are not reinventing wheels. We leverage best-in-class third-party components and focus on the unique value : behavioral memory that learns YOUR patterns and acts on YOUR behalf.
This shall be the only app on your phone that you ever have to open. It evolves from observer to helper to doer in background, without ever asking for your personal credentials like phone number or email ID. In true sense, it does not know you - it's only about your behavior, context and preferences.
This shall be the only app on your phone that you ever have to open. It evolves from observer to helper to doer in background, without ever asking for your personal credentials like phone number or email ID. In true sense, it does not know you - it's only about your behavior, context and preferences.
User journey with un-app (example view)
Watch how behavioral memory evolves over 4 weeks
Same pattern engine, infinite use cases
Cricket Scores
Live match updates delivered before you check
Cab Bookings
Your 6pm Uber booked automatically
Food Orders
Friday pizza ordered without lifting a finger
Groceries
Tuesday milk restocked on schedule
Stock Updates
Market movements before trading hours
News
Your preferred sources, curated daily
One behavioral engine. Infinite possibilities.
Early validation
958
People surveyed
26
Active testers
30
Past testers
4
Patterns detected
1
Patent published
₹1/day
North star pricing
IIT-Madras
Incubated
Not everything on your phone needs your eyes, mind or ears
👆 Click any step to see details
1
User input
You type or say a natural query in plain English or any Indian language.
- Example : "Latest from Lex Fridman"
- No commands, no syntax and just natural language.
- Supports voice input and multiple Indian languages.
- Interface : Simple text input on web/mobile.
- Tracked : Every query is logged for pattern detection.
↓
2
Intent extraction
AI understands what you want and who you are asking about.
- Extracts entity : "Lex Fridman"
- Understands context : "latest content"
- Uses : Third-party AI for natural language understanding.
- Output : Structured intent with entity and action.
↓
3
Memory lookup
System checks : Have we seen this entity before?
- Checks behavioral memory (on-device).
- Retrieves : Past queries, timestamps and frequency.
- Storage : Browser localStorage for sovereignty.
- Privacy : Your data, your device and your control.
↓
4
Pattern analysis
Behavioral graph detects YOUR unique patterns.
- Temporal patterns : Do you check this person at specific times?
- Frequency tracking : How often do you check?
- Sequence learning : Do you check multiple people in order?
- Confidence scoring : How reliable is this pattern?
↓
5
Content fetching
Multi-source intelligent retrieval.
- YouTube : Latest videos via third-party API.
- Web search : Recent posts/articles via search API.
- Social : Public posts from various platforms.
- Caching : Saves results to avoid redundant fetches.
↓
6
Display results
Clean and organized content presentation.
- Source-tagged results (YouTube, Web and more).
- One-click access to original content.
- INSTANT badge if from cache (already fetched).
- Mobile-optimized interface.
↓
7
Memory update
Learning happens with every interaction.
- Increments check count for entity.
- Records timestamp (hour and day of week).
- Updates behavioral graph.
- Builds knowledge graph for future decision flows.
↓
8
🤖 Autonomous agent (background)
Every 60 seconds, predicts and fetches based on your patterns.
- Runs continuously in background.
- Checks : Is it Monday 9am and user checks Lex Fridman at this time?
- Fetches : Content ready BEFORE you ask.
- Cache hit : Shows INSTANT badge when you query.
- Result : Time saved and no manual searching.
🎚️ Three modes of autonomy
Observer mode
System watches and learns. You do everything manually. Behavioral graph builds silently. No autonomous actions yet. You are in full control.
Helper mode
System suggests and assists. Fetches content based on patterns. Shows predictions. You still initiate but system makes it instant. Hybrid control.
Doer mode
System acts autonomously. Orders milk Tuesday 8am. Books Uber at 6pm. Fetches NASDAQ Monday evening. All based on learned patterns. You approve autonomy level.
🧠 Two types of AI at work
Agentic AI
Behavioral memory engine
Pattern detection algorithm. Temporal analysis and confidence scoring.
Autonomous agent logic
60-second prediction loop. Fetch triggers and cache management.
Tool-based decision flows
Cab booking, food ordering and commerce - all based on learned patterns.
Generative AI
Knowledge graph builder
RAG-based system that understands context across interactions.
Calling use case
YOUR AI receives incoming calls and quantifies which ones deserve your attention.
Context understanding
Builds semantic understanding of your preferences over time.
🔧 What we built vs what we leveraged
What we built
Behavioral memory engine
Pattern detection, temporal analysis and confidence scoring. This is our core IP.
Autonomous agent logic
60-second prediction loop, fetch triggers and pattern-based automation.
Behavioral graph
Your unique pattern web. Entity relationships and sequence learning.
Knowledge graph
On-device semantic understanding using RAG for context-aware decisions.
User interface
Single-file HTML app. Black + lime theme. Mobile-first design.
Event tracking system
Permanent behavior analytics. Privacy-respecting insights.
What we leveraged
AI models
Third-party AI for intent extraction and natural language understanding.
Content APIs
YouTube API and web search API. We integrate best-in-class sources.
Deployment infrastructure
Serverless compute and global CDN for zero-maintenance scalability.
Persistent storage
Third-party database for optional sync when user changes device.
🎯 Wedge strategy : Lazy prompting, knowledge graph and behavior-as-a-memory
💾 Data sovereignty and privacy
On-device sovereignty
Behavioral memory lives on YOUR device. Pattern graph, knowledge graph and user preferences - all stored locally. You control everything.
Optional sync (user choice)
Want to change devices? Sync with another AI app? You can choose to back up your data. Otherwise, it stays on your device. Your choice, always.
Privacy first : We see aggregate patterns, not individual data. No phone number required. No email ID required. No personal credentials. Just your behavior, context and preferences.
🚫 What we are NOT
NOT another chatbot : We do not answer questions. We act on patterns.
NOT a search engine : We do not help you find things. We fetch what you need before you ask.
NOT competing with big AI : Claude, ChatGPT and Gemini are incredible. We are complementary.
NOT a super app : We do not want you to stay in our app. We want to reduce your screen time.
NOT cloud-dependent : Your AI lives on YOUR device. Cloud is optional for sync only.
⏰ Why now?
AI agents are here : OpenAI Operator, Google Mariner and Anthropic's use show the market is ready for autonomous AI.
Device OS evolution : On-device AI chips in phones make behavioral memory feasible without cloud dependence.
Small AI models on phones : Open source models can now run directly on your device and learn from your behavior patterns without sending data anywhere.
Timing for India : 4+ hours daily phone usage. ₹1/day pricing hits the sweet spot for mass adoption.
Clear incentive alignment : We succeed when your phone usage becomes more natural and less effortful. You are not locked into any specific AI tool and we have no cost pressure from usage limits.
🚀 Go-to-market : Collaboration over competition
We are building an open ecosystem. We look forward to collaborating and partnering with players who benefit from user-specific AI decision flows.
Device manufacturers (ODMs)
Pre-installed YOUR AI on new phones. On-device sovereignty from day one.
Telecom operators
Bundle with data plans. Behavioral AI as a service for subscribers.
Digital public infrastructure
Integrate with India Stack. Leverage existing identity and payment rails.
Non-first players
Partner with Namma Yatri (not Uber) and Swish (not Zomato). They have incentive to build user-specific AI flows.
Possible tailwind
Content platforms, browser extensions, agent stores, GPT store, ADKs, ACPs and more.
Why partners win : First-party apps (Uber and Zomato) optimize for their platform. Non-first players (Namma Yatri and Swish) can differentiate by optimizing for USER patterns instead.