Building in Public: Our Journey Creating an AI Self-Reflection Tool

We're building ArcMirror in the open because if we're going to ask people to reflect deeply with AI, we should be willing to reflect deeply ourselves. This is our story — not polished, not inevitable, but real.

The Question That Started It All

We didn't start with a business plan or market validation deck. We started with a question that wouldn't leave us alone: What happens when AI can mirror us without judgment?

Mack was deep in the world of AI adoption, watching people oscillate between two extremes. Some were treating AI like a magic oracle that would solve every problem. Others were terrified — convinced that smart machines spelled the end of meaningful human connection. Neither felt true. But nobody seemed to be asking the deeper question: what if AI could be a tool for understanding ourselves?

That's when Jungian psychology entered the picture. Not as a gimmick, not as a shortcut to self-knowledge, but as a framework that's been refined for almost a century: the idea that we all contain multiple archetypal energies, and growth happens when we understand the ones we neglect.

Why Jung, Not Myers-Briggs (Or Astrology)

We get asked this a lot. Why Jung's 12 archetypes and not the Big Five? Not the MBTI? Not the zodiac, which frankly has better marketing?

Three reasons:

• Movement over stasis. Jung's framework is about energy in motion — the Sage learning boldness from the Hero, the Lover discovering autonomy through the Ruler. You're not locked into a type. You're learning which energies you need in each moment.
• Shadow work. Jung insisted that growth starts with naming the parts of ourselves we reject. Every archetype has light and shadow. We don't just celebrate your Sage; we explore what happens when Sage energy gets stuck in analysis paralysis. That's the real work.
• 12 voices, not one. The Sage gives you wisdom. The Lover gives you permission to feel. The Hero gives you courage. The Creator gives you permission to risk failure. A single personality type can't do that. Twelve distinct voices can hold complexity.

We implemented each archetype as a full voice companion — different philosophy, different communication style, different way of holding space. When you journal with the Sage, you're not getting a generic "think about this more." You're getting a fundamentally different conversational energy.

The Zero-PII Decision That Changed Everything

Early in development, we had to make a choice that felt risky: we decided to store absolutely nothing about our users except the conversation tokens needed to power each session.

No names. No session histories across devices. No profiles. No data on which archetypes you use most. Nothing that identifies you.

This decision surprised a lot of people. "How will you build retention? How will you personalize?" They're good questions. But here's what we realized: shadow work — real self-reflection — requires absolute privacy. If we're asking you to tell an AI about your fears, your shame, your unlived potential, then we need to guarantee that we're not storing that as a data point about you. We're not running it through recommendation algorithms. We're not analyzing your pattern to sell ads.

This architecture had unexpected benefits. It made us more trustworthy to early users — especially the social workers and therapists who validated the app. It kept us laser-focused on conversation quality (we can't rely on engagement metrics to improve; we have to listen). And it forced us to build a product people actually want to return to, not one that manipulates retention through data lock-in.

What Early Users Taught Us

We didn't hire focus groups. We built, shared quietly, and paid attention to the people who came back with unsolicited feedback.

An Air Force veteran with a Master's in Social Work saw ArcMirror unprompted. He said, "This is exactly what the people I serve need." His daughter, also a social worker, independently came to the same conclusion. Three other clinicians validated the archetype experience as genuinely useful — not as therapy replacement, but as self-discovery that prepares people for deeper work.

But they also challenged us. They pushed back on language we were using ("therapy" became "self-reflection"). They questioned whether we were being honest about the limitations of AI. They asked us to think harder about crisis safety — if someone journaled about self-harm, what was our responsibility?

We listened. We built a crisis detection system on the web app that surfaces the 988 Lifeline when it detects risk. We rewrote our onboarding to be ruthlessly honest about what ArcMirror is and isn't. And we made "this is not a substitute for therapy" so prominent in the app that it's impossible to miss.

The Things We Got Wrong (And Still Getting Wrong)

Transparency means admitting failure. Here's what we learned the hard way:

The App Store rejection cycle. We submitted to Apple thinking we understood the guidelines. We didn't. Three rejections, four rewrites, and six weeks later, we finally got approved. The rejections weren't malicious — they were clarifying. They forced us to be more explicit about third-party AI usage, clearer about our limitations, and more thoughtful about vulnerability. We're grateful now, but we weren't grateful at the time.

Overestimating technical infrastructure. We thought zero-PII architecture would be simpler to build. It's not — it just moves the complexity to different places. Managing session tokens, handling offline sync, ensuring the conversation feels continuous without persistent data — we underestimated all of that.

The voice quality learning curve. Each archetype's voice needed to be distinctive without feeling gimmicky. The Sage needed to feel wise but not condescending. The Lover needed vulnerability without manipulation. We got some of these wrong. We're still iterating.

Speed vs. quality. We wanted to launch fast. We did. But "fast" sometimes meant incomplete. The web app is where we've learned to slow down, iterate more carefully, and listen more before shipping. That's made the current version far more useful than our initial launch.

Where We Are Now (And Why Building in Public Matters)

As of March 2026, we're live on iOS App Store. The web app is evolving faster than the mobile app ever could. The blog is where we're thinking out loud about what self-reflection with AI actually means.

We're working with a clinician to explore how archetypes show up in professional settings. We're listening to our early users. We're making mistakes and learning from them in real time.

Building in public means we're accountable. It means when we say "zero-PII," people can audit that claim. It means when we fail, we don't get to quietly pivot and pretend it didn't happen. We have to own it, learn from it, and tell you what we're doing differently.

It also means we get to celebrate the small wins with you. When someone journals their way to a decision they wouldn't have made alone. When a therapist tells us that a client came to their session with deeper self-awareness because they'd been working with ArcMirror. When we ship a new feature and it actually works the way we hoped.

What Comes Next

We're thinking about three things:

Deeper archetype work. We want to explore partnership, conflict, and evolution between archetypes. How does the Hero support the Creator? What happens when the Sage and the Lover are in tension? Real growth happens at these intersections.

Enterprise integration. Universities and therapy practices have reached out asking if we could work with them. Not as a replacement for their services, but as a complementary tool. We're being very careful here — the same privacy architecture that protects individuals needs to protect institutional partnerships too.

The identity crisis work. Mack started this because AI advancement is going to cause an existential crisis. Millions of people are going to wake up one day asking "who am I if machines can do what I thought made me special?" We're building a tool that helps people answer that question before the crisis hits. That's the long game.