I wrote a diary for 20 years.
Then I couldn't anymore.
So I stopped writing — and started talking. That's how DailyVox was born.
The Problem
I kept a diary for over 20 years. Almost every day, I sat down and wrote about my life. It wasn't discipline — it was just something I did. Like brushing my teeth.
Then, in recent years, I started missing days. One day missed. Then a week. Then I'd look back and realize I hadn't written anything in a month. The habit that had been part of my life for two decades was disappearing, and I could feel it happening but couldn't stop it.
I tried different apps. Different formats. Different routines. Nothing stuck. Writing every day had become a chore. And when something becomes a chore, you stop doing it.
The Moment
Then I tried speaking instead of writing. And everything changed.
When you speak, you don't worry about grammar. You don't edit yourself. You don't stare at a blank page wondering what to say. You just talk. About your day. About what's on your mind. It flows — naturally, honestly, quickly.
Two minutes of talking captures more than ten minutes of typing. And it feels nothing like a chore.
Building DailyVox
The problem was: there was no good tool for voice journaling that I could trust with my most private thoughts. Every app I found either sent my audio to cloud servers for processing, or charged a subscription, or both.
I'm a developer. So I built my own.
DailyVox is simple at its core: you open the app, press record, talk about your day, and the AI transcribes everything on your device. No cloud. No servers. No account. No subscription. Your voice literally never leaves your phone.
AI tools made this possible as a solo developer. What would have taken a team and months a few years ago, I shipped from idea to the App Store in a fraction of the time. That's the power of this moment in technology — one person with the right tools can build something real.
The Digital Twin Vision
After I started building DailyVox, I realized it could go far beyond just recording voice entries.
If you talk to this app every day — about your thoughts, your decisions, how you react to things, what makes you happy, what frustrates you — it builds a picture. A detailed, honest picture of who you are.
That's when I built the Digital Twin.
The Digital Twin learns your personality from your entries. It understands how you think, predicts your moods, identifies your triggers. And here's where my real ambition comes in:
We already leave digital traces everywhere — social media, messages, emails. But all of that is scattered across platforms owned by other companies. None of it is organized. None of it is truly yours.
What if you had one place where you deliberately stored the most honest version of yourself? Not the curated version you put on social media. The real version.
I believe keeping a digital version of yourself will be an important part of the near future. Not science fiction — practical. And DailyVox is built to make that happen.
But not by sending your data to a server. Everything works on-device. Your Digital Twin lives on your phone. Nobody else has access to it. I won't compromise on that.
Why iOS Only
DailyVox is on iPhone and iPad only. There's no Android version — yet.
That's a deliberate choice. Apple gives me strong security guarantees — the Secure Enclave, the Neural Engine, the privacy frameworks. The whole promise of DailyVox is that your data never leaves your device. I need to be absolutely sure I can keep that promise before putting it on another platform.
I'd rather be on one platform and keep your data safe than be on two platforms and compromise.
Android will come — after iOS is solid, after the user base grows, and after I'm confident the same privacy guarantees can hold.
What I Believe
Privacy Is the Foundation
Not a feature. Not a toggle. The entire architecture is built so your data physically cannot leave your device.
Mental Health Tools Should Be Free
A diary shouldn't have a paywall. DailyVox is free because no one should have to choose between their wallet and their mental health.
AI Should Work for You, Not on You
AI that processes your thoughts should run on your device, under your control. Not on a server training someone else's model.
Speaking Is More Honest Than Writing
When you write, you edit. When you speak, you're real. The best insights come from unfiltered expression.
The Roadmap
Voice Journaling + On-Device AI
Tech: Apple Speech framework (SFSpeechRecognizer, requiresOnDeviceRecognition=true) for zero-network transcription. NaturalLanguage framework NLTagger for sentiment scoring and named entity recognition. Core Data + NSPersistentCloudKitContainer for local-first storage with optional encrypted iCloud sync. AVFoundation for AAC 44.1kHz audio capture. CryptoKit AES-256-GCM for encrypted backups. LocalAuthentication for Face ID/Touch ID. WidgetKit for Home Screen and Lock Screen widgets. AppIntents for Siri Shortcuts. All AI computation on Apple's Neural Engine.
Digital Twin + Predictions
Tech: Custom DigitalTwinEngine building a multi-dimensional personality model. CommunicationStyle tracking (Type-Token Ratio for vocabulary richness, expressiveness/directness/formality scoring, signature word frequency analysis). EmotionalSignature with valence/arousal/dominance baseline, time-based mood patterns (morning vs evening, weekday vs weekend), trigger topic correlation. PersonalKnowledgeGraph via NLTagger NER mapping people, places, and topics with emotional weight scores. TwinPredictions using temporal pattern analysis and day-of-week mood forecasting. All models serialized as Codable JSON in Core Data's AIState entity.
Ask Your Twin + Social Sharing
Tech: TwinChatView with pattern-matched query system — 10 question types mapped to DigitalTwinEngine data queries. TwinResponseGenerator constructs natural-language responses from emotionalSignature, knowledgeGraph, communicationStyle, and behavioralPatterns. ShareablePersonalityCardView renders cards to UIImage via SwiftUI ImageRenderer at 3x scale — Instagram Stories (1080×1920) and Twitter/X (1200×675). UIActivityViewController for native share sheet. ReviewManager with SKStoreReviewController triggered at milestone entries (5, 15, 40) with 90-day cooldown via UserDefaults.
Entry Embeddings + Local Vector Index
Tech: NLEmbedding for 512-dimensional sentence embeddings per journal entry, persisted in Core Data. Custom cosine-similarity vector search index for semantic queries ("entries where I felt conflicted about work"). Proactive insight engine detecting statistical anomalies — z-score deviations from emotional baseline trigger local notifications. K-means clustering on embedding space to surface hidden thematic groupings across months of entries. Preparing the retrieval layer for v2.0's RAG architecture. All computation on Neural Engine — zero network calls.
Multi-Language + Wrist Capture
Tech: Xcode 15+ String Catalogs (Localizable.xcstrings) for UI localization — Hindi, Spanish, Japanese, German as first wave. Speech framework already supports 60+ transcription languages. WatchKit companion app with WatchConnectivity for iPhone sync. Quick voice entry from wrist. Watch face complications showing mood and streak data.
Apple Foundation Models + Tool Calling + SpeechAnalyzer
Tech: Apple's on-device 3B-parameter Foundation Model (WWDC 2025) accessed via LanguageModelSession. Twin becomes a real chatbot with multi-turn conversation and session transcript memory. Tool calling architecture: Twin autonomously queries Core Data via custom Tool protocol implementations — FetchEntriesTools retrieves journal entries by topic/date/mood, GetPersonalityTool returns CommunicationStyle and EmotionalSignature data, GetMoodPatternsTool surfaces temporal trends. Model generates personalized responses grounded in real data — no hallucination. Tone matching via instructions: DigitalTwinEngine feeds the user's expressiveness, formality, directness scores, signature words, and emotional baseline into session instructions — model adapts its voice to match the user's natural style. Guided generation via @Generable macro for type-safe structured outputs (mood reports, weekly summaries as Swift structs, not raw text). Streaming responses via streamResponse() for real-time chat UI. SpeechAnalyzer replaces SFSpeechRecognizer — faster, more accurate, supports long-form audio, no user setup required, volatile results for instant feedback. Requires: iPhone 15 Pro or later, iOS 26. Entire pipeline on-device — zero network calls, zero API costs.
LoRA Fine-Tuning — Twin Learns to Sound Like You
Tech: Apple's Foundation Models Adapter Training toolkit for Low-Rank Adaptation (LoRA). Export 100-1,000 journal entries from Core Data as JSONL prompt/response pairs. Train a personal adapter on Mac (32GB+ Apple Silicon) or Linux GPU — original model weights stay frozen, only small adapter matrices are trained. Resulting adapter is ~160MB, delivered to device via Background Assets framework. Loaded via SystemLanguageModel(adapter: personalAdapter). What changes: The Twin doesn't just know your data — it sounds like you. It learns your sentence structure, your emotional vocabulary ("rough" vs "terrible" vs "challenging"), your punctuation habits, your hesitation patterns, your level of self-reflection. ~95% tone accuracy vs ~60% with instructions alone. Privacy: Training data never leaves the user's Mac. Adapter stored on-device only. No cloud training, no federated learning, no data sharing.
A True Digital Version of You
The end state: An on-device AI that acts like you, speaks like you, responds like you — built from years of your own journal entries. Full architecture: SpeechAnalyzer for voice input → Foundation Model with personal LoRA adapter for generation → Tool calling for real-time data retrieval from Core Data → NLEmbedding vector index for semantic memory search → Guided generation for structured personality outputs → Session transcript management for long-term conversation memory with context condensation when window is exceeded. Context-aware emotional intelligence: The Twin understands not just what you said, but when you said it, how you felt, who you were talking about, and what patterns repeat. It can answer "Why have I been stressed this month?" by retrieving relevant entries, analyzing sentiment trends, correlating with entity mentions, and generating a response in your own voice. Digital self-preservation: Your Twin is the most honest, complete digital record of who you are — your thoughts, your patterns, your growth over years. Encrypted. On your device. Inaccessible to any external process. Exportable only by you, for device migration. Your data never touches a server. Your digital self is yours alone.
Get in Touch
I'm Karthikeyan. I built DailyVox because I needed it. Turns out other people need it too.
If you have questions, feedback, or just want to say hi:
Email: intrepidkarthi@gmail.com
GitHub: github.com/intrepidkarthi