The systems that failed me: Why I build survivor-first

I don't talk about my personal story much. Not because I'm ashamed of it, but because I don't want NovaHEART to be about me. I want it to be about the work.

But people keep asking why I build this way. Why the obsession with safety. Why the insistence on audit trails and crisis detection and forensic accountability. Why I can't seem to build anything without asking "what happens when someone is at their worst?"

So here's the honest answer: I build this way because I know what it feels like when systems fail you.

What failure looks like

I've been on the receiving end of systems that were supposed to help. Waiting rooms that felt like judgment. Forms that asked the wrong questions. Crisis lines that put you on hold. Professionals who said the right words but clearly weren't listening.

I've experienced what it's like to finally work up the courage to ask for help and find the wrong thing on the other side.

That experience shapes everything I build.

When I design a crisis detection system, I'm not thinking about the abstract "user." I'm thinking about myself at my lowest, and what would have actually helped versus what would have made things worse.

When I insist on clear communication about AI limitations, it's because I know what it feels like to expect help and get something that can't actually provide it.

When I build grounding tools into every mental health product, it's because I know what it's like to need something that works right now, not after a referral process.

What I learned

The shame is the biggest barrier. Asking for help is hard. Anything that increases shame—judgment, cold bureaucracy, having to repeat your story multiple times—makes it harder. Anything that decreases shame—warmth, normalisation, privacy—makes it easier.

Small details matter enormously. The difference between "if you're in crisis, please call" and "it sounds like you're going through something really difficult—here are some people who can help" is enormous. One is a disclaimer. The other is care.

People in crisis can't navigate complexity. When you're at your worst, you have almost no capacity for decision-making. Systems designed for crisis need to be simple, clear, and directive. Every additional step is a place where people fall away.

Consistency builds trust. Vulnerable people have often been let down before. Every interaction that matches expectations—every promise kept, every boundary maintained—builds trust. One broken promise can destroy it.

Autonomy matters even when you're struggling. Crisis doesn't remove the need for agency. Systems that take over completely can feel as violating as systems that abandon you. The goal is support, not control.

How this shapes NovaHEART

Every design decision at NovaHEART goes through a filter: "Would this help someone at their worst?"

Phoenix Fail-Safe exists because I know that people in crisis interact with systems that weren't designed for them, and those interactions can make things worse. Phoenix tries to catch that before it happens.

Clear limitations are communicated everywhere because false hope is worse than no hope. If you're struggling and reach out to an AI expecting a therapist, the disappointment can be devastating.

Grounding tools are accessible at all times because sometimes you don't need conversation—you need to feel your feet on the floor.

Audit trails exist because accountability isn't optional when you're working with vulnerable people. If something goes wrong, we need to be able to explain what happened.

Session limits prevent dependency because a tool designed to support shouldn't become a crutch that prevents people from accessing real help.

The responsibility

Building for vulnerable populations is a responsibility I don't take lightly.

People will use these systems at their worst moments. They'll interact with our AI agents when they're scared, or overwhelmed, or considering whether to keep going.

That means every feature, every interaction, every design decision carries weight. Getting it wrong could genuinely harm someone. Getting it right could genuinely help.

This isn't abstract for me. It's personal. And that's why I build this way.

I share this not for sympathy but for context. NovaHEART exists because I know what bad systems feel like—and I'm trying to build better ones.