Whisp
Designing an AI product that doesn’t pretend to know
Whisp is a journaling app for signs, dreams, repeating numbers and meaningful moments. I designed, built and shipped it myself — with AI at the center of the product, and all through the process of making it.
The two-minute version
- Lots of people notice repeating numbers, vivid dreams and animals that keep showing up — and most apps for that either lecture them or promise things no one can know.
- I designed Whisp to remember everything someone logs, find patterns across it over time, and reflect it back with AI — under strict rules about what that AI is allowed to claim.
- I built and shipped it solo, using AI heavily for design exploration, code and content. It’s my proof that I can still take something from zero to a live product — and that I take AI seriously as both material and tool.
One sign is a moment. A pattern is a message.
A lot of people keep noticing things — 11:11 on the clock, the same bird on the fence, a dream that won’t let go. The apps that serve them tend to fall into two camps: static dictionaries (“a hawk means X”), or fortune-telling that promises certainty and leans on fear and urgency to sell subscriptions.
I wanted something calmer and more honest. Whisp doesn’t tell you what a sign means. It remembers what you notice, connects it to your dreams, journal and life events over time, and reflects patterns back so you can decide what they mean to you.
Notice
Dreams
The weekly Weave
The hard part wasn’t the AI. It was what the AI shouldn’t do.
Trust with personal material
People log dreams, grief, hopes. The product has to feel safe and never manipulative.
AI that stays humble
Language models are very happy to sound certain. Whisp must never predict outcomes or claim to know what something “really” means.
Belief-aware, not belief-prescriptive
Users range from devout to curious-skeptic. Reflections have to respect their framework without pushing one.
Everything, alone
Product, design, engineering, subscriptions, App Store, marketing — one person, so every decision is also a scope decision.
Notice → remember → connect → reflect.
Whisp’s value is memory. Every sign, dream and journal entry is kept with its context, so patterns can emerge across weeks and months. AI reflections follow one rule of thumb: meaning first, evidence second, action third — what this might mean to you, what in your own history points that way, and a gentle next step.
Designing what the AI is allowed to say.
Write the guardrails as product requirements.
I wrote down, as product requirements, what Whisp would never do: claim supernatural certainty, predict outcomes, use fear or urgency, or invent testimonials and metrics. Those became the acceptance criteria for every AI response and every piece of marketing.
Whisp explores meaning. It does not manufacture certainty.
Name the sighting. Never assign the meaning.
Animal signs were a special case: a dictionary definition feels authoritative and can land badly on something personal. Whisp names what you saw, shows how it has shown up in your history — instead of handing down a meaning.
A sign, with the reflection split into general meaning and personal context.
The timeline, where threads start to form.
A whisper, not a scream.
The visual and verbal identity carries the same restraint: deep navy and indigo, warm gold, an editorial serif for emotional moments and a clean sans for utility, painterly celestial imagery. No flashing badges, no fake scarcity, no streak guilt.
The weekly Weave reading.
Energy centers and celestial context.
A business model that doesn’t fight the product.
A free tier with a limited number of full-quality AI reflections and 90 days of history, and a Premium subscription for the rest. Early supporters get a locked-in founder rate. Checkout runs on the web (Stripe → Adapty → Firebase entitlements), which keeps pricing flexible and honest.
AI as a collaborator, not a vending machine.
I used AI throughout the build: exploring UI directions, writing and refactoring code, drafting the reflection prompts that sit behind the guardrails, and producing marketing imagery and copy that I then edited hard. It didn’t replace design judgment — it compressed the distance between an idea and something I could put in front of people.
I’m now building a content pipeline on the same principles: the app’s daily content is adapted per platform by a language model, voiced, rendered to short video and scheduled — with the same “no certainty, no fear” rules applied automatically.
Live, early, and honest about it.
Whisp is live on iOS and Android. The product works; the next problem is distribution, which is where my attention is now. Building it solo has made me a sharper product leader: when you’re the whole team, every feature has to earn its place.
What this project says about how I work.
AI product design
Designing guardrails, voice and trust — not just a chat box.
0 → 1
Strategy, UX, visual design, engineering and launch, end to end.
AI-native workflow
Using AI seriously to design, build and market — with judgment in charge.
Restraint
A business model and brand that refuse the manipulative patterns common in the category.