Dreamalizing: privacy-first AI for sensitive personal input
How Elk Solutions professionally guides sensitive dream input with bounded AI — privacy-first, verifiable and publicly available as an iPhone app.
Not every proof of what Elk Solutions builds sits behind an NDA. Dreamalizing is an Elk Solutions product that is fully visible: an AI that helps people explore their own dreams — privacy-first and bounded, with a live marketing site and an iPhone app in the Dutch App Store.
It is not a client case but a studio case: developed, deployed and operated by Elk Solutions. We carry its design choices and its boundaries ourselves. That makes Dreamalizing a visible example of how controllable AI can remain professional even with highly personal and sensitive input.
Privacy is not a promise but a boundary by design.
Where the data stays, what the model is allowed to see, and what is removable — that is established in the system, not in a statement.
This is a claim, not a measurement. The strike-through and underline are the same marking the renvooi service uses to show a change in law.
The challenge
Dream content is an extremely privacy-sensitive category. People share things in it they tell no one else. That imposes two hard requirements that seem to pull in opposite directions:
- AI has to be useful — otherwise it's just an empty text field
- AI must never pretend to know the truth — no therapy, no diagnosis, no "here's what your dream means"
A naive AI product picks one side: either an assertive dream-interpreter that fakes certainty, or a harmless journal with no value. Both are wrong. The real challenge was to use AI as a guide within strict safety and privacy boundaries — useful, yet honest about what it doesn't know.
The solution
I built Dreamalizing around a few deliberate choices, each one mirroring an Elk Solutions design principle:
- Guided exploration instead of black-box output. The AI doesn't interpret; it asks targeted questions. The user stays the authority over their own dream; the AI only helps explore.
- Bounded AI with clear limits. The model is explicitly barred from therapy, diagnosis or medical claims. Those limits live in the system, not in a disclaimer at the bottom.
- Privacy-first storage. Dreams are stored encrypted (AES-256-GCM, per-user key) on our own infrastructure — not at a large cloud vendor.
- Local inference. The language models run locally, enforced through configuration, so sensitive input doesn't leak to external AI services by default.
- Opt-in pattern recognition. Connections across multiple dreams are only sought when the user deliberately chooses it.
- Deletable and traceable. A deletion request actually wipes the data completely (GDPR), and the chain is auditable.
What this proves
Dreamalizing shows concretely how Elk Solutions takes a sensitive AI product from concept to controlled operation. The evidence is not a general claim about the difficulty of app stores, but a set of verifiable parts of the production chain:
- Public distribution and payment. The iPhone app is available in the Dutch App Store with a payment model; the marketing site supports 14 languages.
- End-to-end operation. Deployment, payments, maintenance and ongoing development are managed as one production chain.
- Technical privacy controls. Sensitive data is protected with encryption, own infrastructure and local inference.
- Enforceable AI boundaries. Guardrails are part of the system rather than a textual disclaimer.
The public distribution can be verified directly: Dreamalizing is listed in the Dutch App Store. The boundary is equally explicit: Android is not yet available and is therefore not presented as delivered.
The same principles — context as truth, measurability, governance and clear boundaries — are what I apply in client work around AI in production.
See it for yourself
Dreamalizing's marketing site and iPhone app are live; Android is still to follow. You can explore the product, and if you want to use AI with sensitive or high-risk data, Elk Solutions can help define enforceable boundaries and controls.