AI-Powered Mental Health Infrastructure — built in one afternoon.

15.5 million U.S. adults live with diagnosed ADHD. Between 18.6% and 53.3% of them also experience depression. Meanwhile, productivity apps see a 30-day retention rate of 4.1%. The supposed solution is failing the people who need it most.

Quinn isn't a productivity app. She's a personal system layer between my brain and my commitments — captured by voice or text, classified silently in the background, and protected against my own worst instincts during depressive episodes.

I'm a product leader who builds workflow tools for a living. I still couldn't maintain a productivity system for myself for more than three weeks. The reason wasn't discipline — it was design. Every tool I tried was built for a brain that works differently than mine.

Before I touched a single tool, I spent time on an uncomfortable question: what are my actual failure points? These four observations became the design requirements that every feature of Quinn maps back to.

The first architectural decision: Quinn classifies incoming information into five distinct entity types rather than dumping everything into a single task list. Tasks are one of five things — not the default container for everything.

The headpat category deserves its own paragraph. Most productivity systems accidentally reinforce people-pleasing — the dopamine hit ties to completing tasks, which are almost always things done for someone else. The headpat inverts that. The wins captured are yours. Self-generated, self-directed, independent of external validation. For anyone who's experienced the ADHD-adjacent loop of approval-seeking as a learned response to chronic underperformance, this isn't a small design choice. It's a therapeutic one.

Each protocol fundamentally changes what Quinn surfaces and how she behaves. They exist because without them, the system would be useless on the days I needed it most.

I'm not a developer. I'm a UX practitioner and systems designer. What I contributed wasn't engineering — it was design thinking: understanding the actual user (me), mapping the actual failure points, and writing prompts specific enough that the output was usable.

Sprint 2 — Memory. Everything saves immediately. Zero friction. No questions. Classification happens later, in the background, via LLM. The user experience is: type, submit, done.

Sprint 3 & 4 — Review and adaptive cards. Quinn surfaces unreviewed items one at a time. Cards scale by confidence: high-confidence headpats route silently to the carousel; simple tasks get one sentence and two buttons; commitments get triage; uncertain items get an honest "help me understand this one."

Sprint 5 — Real infrastructure. Google Calendar, Gmail, task bundling. On hard days, the social labor of rescheduling is often what breaks you. Quinn drafts the email — warm, professional, no reason given — and sends it with one tap. Nobody gets an explanation. Nobody gets an apology.

Sprint 6 — The analysis lockout. The single most important design decision of the entire build.

Given access to data about my own performance during a depressive episode, I will use it to hurt myself. Not dramatically — just: the completion rate becomes a verdict. So the analysis section locks automatically during Sad Panda mode, for 7 days after it deactivates, during Red mode, and within 48 hours of last access. When locked, the screen shows exactly one message, in plain sans-serif, no Quinn warmth, no personality.

Sprint 7 — A phone number. Quinn gets a Twilio number. Family can text it. Business leads from marketing materials can text it. Someone picks up a flyer, texts Quinn — she captures them, answers FAQ, and surfaces the lead in Review. I never have to be available. Quinn holds the whole thing.

During Sad Panda mode, after the three questions, a quiet link sits at the bottom of the screen: "Show me the receipts."

Quinn compiles everything — every headpat captured, every commitment kept, everything she's noticed about what I've been carrying — and displays it on one warm, plain, scrollable screen.

Context comes before the wins. Then the headpats — real ones, captured by me, in the words I actually used. A sizzle reel I worked hard on. A meal I was proud of. My son looking at me and saying Mom. Then the closing line, in italic serif at the bottom:

"The fact that you're still trying is the proof."

This isn't a productivity feature. It's evidence-based self-compassion delivered at the moment of maximum need, in the words you chose when you had enough capacity to be kind to yourself. Depression lies about what happened. Quinn has the receipts.

Quinn isn't therapy. She isn't medication. She isn't a replacement for professional support. She's a system layer — between my brain and my commitments, between my hardest days and the world's expectations — completely customized to how I actually work.

Most people don't know they can build this. They think you need to be a developer. You don't. You need to be honest about your failure points and willing to design something that serves your real life.