Persona 1: The Skeptic

People don't use the term "cognitive decay," but they say "I forgot where I put that," "Who was that person?", or "I wish I could remember what my dad said." We are solving the universal pain of forgetting, which everyone experiences daily but has accepted as inevitable. We are making the inevitable optional.

Google Photos is a library for files; Dzikra is a search engine for memories. Google can find "a dog," but it can't find "the dog Sarah mentioned." We index the context and relationships (Social & Verbal memory), not just the pixels.

Exactly. That is our core advantage. Dzikra requires zero manual organization. Our AI passively analyzes visuals, voice, and location to build the index. The user does nothing but live their life and ask questions later.

The average person takes 20 photos a day and feels anxiety about losing them. It's not hoarding; it's fear of loss. Our beachhead (Hajj pilgrims) proves this: they capture once-in-a-lifetime moments and are terrified of forgetting the spiritual experience.

It works for nouns ("Car", "Beach"). It fails for episodic memory ("The car I rented in Italy"). Apple searches labels; we search episodes. Try asking Apple Photos "Review the books I saw last year"—it fails. Dzikra succeeds.

Because Google's business model is advertising; ours is privacy. We use on-device processing and zero-knowledge encryption for cloud sync. We literally cannot see your memories. Google analyzes them to sell ads; we analyze them to serve you.

Apple focuses on the "highlight reel"—making you feel good about the past. We focus on "cognitive utility"—making you smarter in the present. Furthermore, we are cross-platform (Android/iOS/Web/Wearables), whereas Apple is locked to their ecosystem. Your memory shouldn't depend on your phone brand.

They pay for iCloud to avoid the "Storage Full" popup. They will pay for Dzikra to access "Superpowers." We are positioning this as a productivity and wellness tool (like Notion or Headspace), not a utility bill.

Never. We have a "Local-First" guarantee. Your data is always yours, stored locally on your device. If you stop paying, you lose the Cloud Sync and proprietary AI processing for new memories, but you never lose access to your past data.

Those products failed because they required new hardware (Humane) or were Mac-only screen recorders (Rewind). Dzikra is software-first, meeting users where they are (on their current phone), focusing on the high-emotion mobile capture use case, not just desktop productivity.

It is the perfect "high emotion, high volume" wedge. 20M people/year, distinct start/end date, willingness to pay is high (trip costs $10k+), and the fear of forgetting the spiritual experience is acute. It allows us to dominate a niche before attacking the general massive market.

Saudi Arabia has 98% smartphone penetration. Pilgrims use apps for visas, navigation, and rituals (Nusuk). They are mobile-first users who already document everything on WhatsApp and Instagram.

The "Core Loop" is identical: High-emotion event -> Fear of forgetting -> Capture -> Recall. A pilgrim uses it for the Kaaba; a parent uses it for a graduation. We strip the religious branding for the Global launch, keeping the underlying cognitive engine.

Through "Shared Memory Context." When I recall a shared event with you, I send you a "Memory Link" that unlocks that specific episode. To view the full AI context, you onboard. It's the same loop Dropbox used for shared folders.

For a 20-year-old, perhaps. For a parent, an executive, or anyone over 40, memory loss is a source of genuine anxiety. We are selling "Memory Security," which is a painkiller for the fear of aging and loss.

We don't market "Cognitive Augmentation." We market "Never forget your mother's voice" or "Recall any document instantly." We focus on tangible use cases. In the Hajj beachhead, our CAC is near zero due to B2B2C partnerships with travel agencies.

They are text-first and manual. You have to *type* to remember. Dzikra is capture-first and passive. We are for the 99% of life that you don't have time to write down.

Moves was a location tracker with no semantic understanding. It could say "You were at Starbucks," but not "You discussed a merger at Starbucks." The AI technology to unlock the *meaning* of the memory didn't exist until the LLM era.

The network effect is "Data Gravity." The more you store, the smarter the AI gets about *you*, and the harder it is to leave. Additionally, our "Family Memory" tier creates a local network effect within households.

ChatGPT is a text chatbot. It doesn't have access to your camera roll, your location history, or your offline conversations in real-time. We own the "Capture Layer" on the device; ChatGPT is stuck in the cloud waiting for text input.

We use "Opportunistic Processing." We only index when the phone is plugged in and on Wi-Fi (like iCloud). During the day, we only capture lightweight metadata. This ensures zero impact on daily battery life.

We don't store 4K video in the cloud. We store "Embeddings" (vectors) and heavily compressed thumbnails for indexing. The original high-res files stay on the user's device or their own personal cloud (Google Drive/Dropbox integration). We store the *index*, not the *archive*.

Yes. The cost of storing text vectors for a lifetime of memories is under $1/year per user. The heavy lift is the initial processing (inference), which we run on-device (Edge AI) via CoreML/TFLite whenever possible to offload cloud costs.

We use Whisper-large models fine-tuned on conversational fragments. We also don't need 100% transcript accuracy; we need "Semantic Gist" accuracy. If we capture "book... recommend... Sci-Fi", that's enough for the index to work.

Crucial point. We implement "Source Anchoring." Every AI answer cites the specific photo, audio log, or location data point it came from. Generally, we don't synthesize new memories; we retrieve existing records.

We use a Hybrid approach. Basic indexing (OCR, Face ID) is on-device. Complex query reasoning ("What was the vibe?") sends anonymized vectors to the cloud within a secure enclave. The raw media never leaves the user's control without explicit consent.

We focus on "Tier 1" Android devices (Samsung S-series, Pixel) for the initial launch to manage hardware capability. The basic capture features work on any phone, and cloud AI processing works universally across all devices.

Our architecture is "Privacy by Design." Deleting a memory deletes the vector key, making the data cryptographically irretrievable instantly. We are fully GDPR and CCPA compliant because we treat data as user-owned property.

Vector search is O(log n). Searching 1 million vectors takes milliseconds. The bottleneck is LLM generation, which is optimizing rapidly. We also pre-fetch frequent questions ("Where is my passport?") for instant recall.

We are model-agnostic. We can swap between GPT-4, Claude, or Llama 3 instantly. We are currently fine-tuning open-source models (Mistral/Llama) to run on our own infrastructure to reduce dependency and cost.

We don't record *everything* 24/7 like a bodycam. We offer "Active Capture" (user takes a photo/note) and "Passive Context" (location, calendar). The user sets the dial. We find most users start manual and enable more passive features as they build trust.

Dzikra allows you to live in the moment *because* you don't have to worry about capturing it perfectly. You can be present, knowing your "Second Brain" has your back. It reduces the anxiety of "Did I take a photo of that?"

Did calculators make us worse at math? No, they allowed us to do *advanced* math. Dzikra offloads the low-level task of "storage" so your brain can focus on "creativity" and "synthesis." It evolves the brain, it doesn't atrophy it.

Our Verbal Memory feature has a "Consent Indicator" (visual ripple on screen) and we encourage transparency. But largely, this is for *personal* utility—remembering your *own* experience of the conversation, not spying on others.

Early adopters are productivity enthusiasts, yes. But the "Retained Users" are parents and people with busy social lives. The "Killer Feature" is usually finding a lost item or recalling a detailed instruction from a spouse.

We feature "Digital Muting." You can mute a person, a date range, or a location. The data remains (if you want), but it stops resurfacing in "Flashbacks" or search. You control the narrative of your memory.

We drive the UI primarily through Voice. My 70-year-old mother can't search through folders, but she can ask "Show me the photos of the grandchildren." Voice is the universal interface for the elderly.

We engineer serendipity. The "Morning Review" shows you memories relevant to *today* (e.g., "On this day 5 years ago"). We don't remove accidental discovery; we curate it so it's delightful, not overwhelming.

Because Dzikra gets more valuable the longer you use it (compounding value). The first month is useful; the 12th month is indispensable because it holds a year of context. Our churn drops significantly after the "Magic Number" of 100 memories.

We support 50+ languages natively via our Whisper + LLM pipeline. You can search in English for a conversation you had in Indonesian. This "Cross-Language Recall" is a major feature for international users.

18 months. This allows us to reach 50k active users and prove the unit economics of the Hajj beachhead.

We use Purchasing Power Parity (PPP) pricing. In Indonesia, it's roughly 30k-50k IDR/month. In the US, it's $8. We optimize for adoption in developing markets and revenue in Tier 1 markets.

Ads destroy trust in a memory product. If we show you an ad for shoes when you recall a memory of a funeral, we die. Subscription aligns our incentives with the user: Privacy and Utility.

Yes. The marginal cost of adding a user is storage + minimal compute. With our "Index Only" strategy, gross margins are above 70%.

We are building to be a standalone Memory Platform. However, strategic acquisition by Apple, Microsoft (for Copilot consumer integration), or even a health-tech company (Alzheimer's care) are viable scenarios.

Standard Seed dynamics; we are offering [X]% for [Y] investment, ensuring the founders remain motivated and in control to drive the long-term vision.

Because this is a Product & GTM challenge, not just a research challenge. The models exist; the difficulty is wrapping them in a UX that humans love. Our team has deep experience in Consumer Apps and the specific cultural nuance of our initial market.

We have provisional patents filed on our "Zero-Knowledge Semantic Indexing" workflow. The real moat, however, is the data compounding effect and user trust.

Our "Edge-First" strategy is our hedge. We push as much compute to the user's phone as possible. If the cloud gets expensive, we simply throttle the "Deep Analysis" speed without breaking the core functionality.

Convenience beats free. Privacy beats free. Better recall beats free. People pay for Spotify when radio is free. People pay for Dropbox when email attachments are free. We sell a premium experience that "Free" products can't match without violating privacy.

Critical. The app is "Offline-First". You can capture and search your local index without any signal. Sync happens when connection is restored.

That is our "Phase 2". We are building the backend software now so that when Ray-Ban Meta or Apple Vision become mainstream, we are the default "Memory OS" for them.

Since we use end-to-end encryption, we technically cannot comply with a request to "turn over user data" in a readable format. We can only provide encrypted blobs. This is a feature, not a bug.

Yes. We have a "Memory Migration" tool. It takes time to process, but it allows you to start Day 1 with a fully populated brain.

No. We continuously export a standard JSON/multimedia archive to a local folder or user-owned cloud bucket. We strictly avoid vendor lock-in of the raw assets.

It combines "Return" and "Memory". It's intuitive, globally pronounceable, and shorter than "Cognitive Augmentation Platform."

Product Engineers. Engineers who have messy empathy for the user, not just academic knowledge of Transformers.

Metric: "Successful Recall Rate." How often does a user search effectively vs. abandon the search? We aim for >90%.

Initially, yes (early adopters). But as compute costs drop (Moore's Law for AI), we will lower the price to make it a universal utility.

Because "Memory" is the fundamental unit of human identity. Owning the platform that preserves identity is the most valuable real estate in the digital world.

Founding team includes ML engineers with experience at [Tech Co], product designers who shipped consumer apps with 1M+ MAU, and a regional GTM expert for the Hajj market. We are technical and scrappy.

We have a functional MVP already. 6 months of nights/weekends. We're raising to go full-time and scale from MVP to market-ready product.

Standard 4-year vesting with 1-year cliff. The IP and codebase remain with the company. We've worked together for 3+ years already; the risk is minimal.

Because we live the problem daily. I've personally lost irreplaceable recordings of my late father's voice. This isn't an MBA case study; it's personal grief channeled into solution-building.

The moat isn't code; it's user trust and data gravity. A fork with zero users has zero value. We build relationships, not just algorithms.

The multi-billion dollar scrapbooking, photobook, and memorial industries say otherwise. People spend thousands on weddings to capture memories. The willingness-to-pay is validated.

Gen Z is the most anxious generation about identity and authenticity. They obsess over "core memories." We position Dzikra as the anti-performative platform—private, real, yours.

Data compounding and switching costs. After 2 years of memories, migration pain is too high. Also, our "Local-First" architecture is technically complex to replicate properly.

Notion is manual text input. Dzikra is passive multimodal capture. Different behavioral loop. Notion is for what you decide to write; Dzikra is for what you live.

It's the wedge, not the market. We prove the model here, then expand. It's the same playbook PayPal used (eBay power sellers) and Airbnb used (SXSW attendees).

We comply with COPPA. Users under 13 require parental consent. The parent controls the child's memory vault until they reach the age of digital consent.

We're a neutral tool (like a camera). We respond to lawful subpoenas but cannot decrypt user data. Our liability exposure is minimal.

We diversify regionally. Indonesia, Malaysia, UAE, Turkey—all are backup markets. If one door closes, we have four others.

Face embeddings are generated on-device and stored locally. We never upload raw facial biometric data to the cloud, keeping us compliant with Illinois BIPA and EU regulations.

Voice-first design is inherently accessible. We support VoiceOver, TalkBack, and high-contrast modes. Memory should be accessible to everyone, including the visually impaired.

It's the default OS for wearable AI (glasses, pins). Every human interaction is passively indexed. Forgetting becomes a conscious choice, not a default state.

No. We partner with hardware OEMs. We're the "software brain" for their cameras. Think: Android for Memory Devices.

Yes. It's not our primary GTM, but caregivers are an incredible use case. Dzikra can help dementia patients "remember" their family through photos and voice. We'll explore healthcare partnerships in Year 3.

Product first, platform later. We nail the consumer app, then open APIs for developers to build "Memory Apps" on top (e.g., a memoir-writing tool that pulls from your Dzikra index).

Access to a category-defining company at the intersection of AI and human experience. You get bragging rights when this becomes the "Spotify of Memory."

If you can use WhatsApp, you can use Dzikra. We hide all complexity. You just take photos and ask questions. That's it.

Basic features work on iPhone X. Advanced AI features require iPhone 12+. We gracefully degrade functionality on older devices.

If you're on Premium (Cloud Sync), your memories are safe and restore to a new device. If you're Free (Local Only), you lose the data—like losing a physical photo album.

Yes. We provide standard JSON export. We don't believe in lock-in. If a competitor serves you better, we respect that.

The app + local index is ~2GB. Your photos stay in your camera roll. We don't duplicate storage; we reference existing files.

Someone who understands consumer behavior, has deep empathy for human memory, and ideally has portfolio companies in the "Personal OS" space (productivity, wellness, longevity).

Yes. We're in conversations with several funds. We're looking for the right partner, not just capital.

We aim to close within 60 days. We have strong inbound interest and want to move quickly to capitalize on the AI memory window.

Absolutely. We'll give you TestFlight access (iOS) or APK (Android). Use it for a week. If you don't see the magic, we don't deserve your capital.

Speed. Apple and Google move slowly, but they move. We need to capture users and data gravity before they launch competing features. That's why we're raising now.

People didn't know they "needed" Uber until they used it. The problem exists (frustration when searching photos); we're just naming it and solving it elegantly.

2015 didn't have on-device LLMs, Whisper, or CLIP. The tech stack matured in 2023-2024. The "Why Now" is that the AI foundation finally exists.

Cultural fluency in the Hajj market (native Arabic speakers on team) + deep technical chops (ex-FAANG ML engineers). Few teams have both.

Failure to achieve product-market fit in the Hajj beachhead before capital runs out. That's why we're laser-focused on this niche first.

If after 12 months, CAC > LTV and churn is above 10% monthly, we'd reassess the beachhead strategy—but the core mission (Memory OS) remains.

Yes. We're builders, not ego-driven founders. If the data says we're wrong, we pivot. That said, we won't compromise on privacy or user trust.

We've known each other for years and have weathered tough projects together. Clear division of labor: tech, product, GTM. Zero drama.

We have a small angel round committed ($200k from operators in AI/consumer space). This Seed round is the institutional layer.

We're model-agnostic. We can swap to Anthropic, Mistral, or self-hosted Llama within days. We're not strategically dependent on any one vendor.

Because memory is the most valuable asset a human has, and we're the only team building the infrastructure to preserve it with absolute privacy and intelligence. This is a once-in-a-decade opportunity.