Inspiration
The inspiration for POCO WINS came from a gap that clinical psychology has known about for decades but technology has largely ignored. Adapted Cognitive Behavioral Therapy works for individuals with intellectual disabilities — the research is clear on that. But therapy is one hour a week. Life is the other one hundred and sixty seven. For adolescents and adults with mild to moderate ID who are also managing depression or anxiety, that gap between sessions is where skills go to disappear. They learn a calm breathing technique on Thursday. By Monday it is gone — not because they didn't try, but because working memory limitations, difficulty with abstraction, and the absence of structured repetition make independent skill retrieval genuinely hard in ways that most digital tools have never been designed to address.We were also moved by something quieter than the clinical data. Individuals with ID experience significantly higher rates of depression and anxiety than the general population — and significantly fewer tools designed specifically, thoughtfully, and respectfully for them. Most wellness apps assume a neurotypical user and add an accessibility toggle as an afterthought. We wanted to build something where the entire architecture — every animation, every word, every mechanic — started from this population's actual cognitive profile and worked outward from there. Not an app adapted for them. An app built for them from the ground up.The name came from a Spanish word. Poco means small. And in our work understanding this population, we kept coming back to the same clinical truth — for someone managing intellectual disability alongside depression or anxiety, showing up on a hard day and doing one small thing is not a minor achievement. It is a victory worth celebrating loudly and building upon consistently. POCO WINS is built around that belief entirely.
What it does
POCO WINS is a therapy reinforcement platform that bridges the gap between weekly therapy sessions and the real lives of adolescents and adults with mild to moderate intellectual disabilities. It operates across three core challenges that adapted CBT faces in this population — encoding skills clearly, practicing them independently without a therapist present, and generalizing them spontaneously across different real-world situations.At its heart, the app gives each user a living digital world — Maya's Island — that grows visibly and meaningfully every time they engage with a therapeutic skill. The world is not decoration. It is a clinical tool. Each domain of the island maps to a domain of the user's life — home for personal skills, garden for mindfulness and emotional regulation, community center for social skills, cove for calm and safety. When a skill is practiced, the world grows in the domain that skill belongs to. When a skill is used in a new context — at work, at home, on the bus — a new part of the world unlocks. Generalization is not an abstract clinical concept in POCO WINS. It is a flower that blooms in a new part of the garden. It is a light that turns on in the community center. The user can see their own growth in a way that working memory limitations would otherwise make impossible to hold.The client-facing experience includes an AI-powered Practice Buddy that provides daily micro-practice sessions calibrated to the user's functional tier, emotional state, and time of day. The Buddy initiates — the user never has to self-motivate from zero. Environmental anchoring ties practices to existing daily routines — toothbrushing, morning phone pickup, mealtimes — so independent practice happens inside moments that already exist rather than requiring new slots in an already demanding day. A Real-Time Skill Matcher surfaces the right skill at the moment of distress — the user taps one button and the AI finds the skill they need right now from everything they've already learned. A rich gamification ecosystem including a personal flame streak, quest system, community collective glow, and behavioral activation features like the Mood Garden, Kindness Ripple, and Before I Sleep ritual keep engagement consistent and clinically meaningful rather than cosmetically rewarding.For therapists, a dedicated dashboard called Maya's Clinical Map provides a between-session window that clinical practice has never had before. The Emotional Weather Strip, Skill Mastery Map, Learning Fingerprint, Between-Session Gap Analyzer, and AI Recommendation Engine give therapists a data-informed session agenda in under two minutes of review — without requiring patients to remember, report, or explain anything. In-session assignment tools and a personal voice note feature allow therapists to set Maya's week directly from the dashboard and send their voice into the app through the Buddy, maintaining the therapeutic relationship across all seven days rather than just one.Every word in the app is written at or below a sixth-grade reading level. Every instruction is paired with audio. Every piece of feedback is reinforcing, never punishing. The flame never dies. The world never resets. The Buddy never says wrong.
How we built it
We built POCO WINS as a functional prototype across a focused hackathon sprint, making deliberate technology choices that prioritized clinical integrity alongside technical execution.The frontend was built using v0 for rapid UI scaffolding and component generation, with the design system built around a single non-negotiable constraint from the start — every component had to pass a six-point accessibility check before it was considered complete. High contrast, large touch targets, single idea per screen, audio paired with every text element, reinforcement-only feedback language, and immediate world response to every user action. These were not features added at the end. They were the specification from which every component was designed.The AI layer was built using the Anthropic Claude API as the core intelligence powering three distinct functions — the adaptive calibration engine that assesses functional tier through disguised gameplay during onboarding, the Practice Buddy's conversational and responsive micro-practice guidance, and the Situation Remix Engine that generates endless contextually varied scenarios for a single skill to drive generalization without requiring new content creation. We used Claude's ability to simplify complex therapeutic content into concrete block-style instructions as the primary mechanism for translating CBT skills into accessible micro-practice modules — ingesting adapted CBT frameworks and outputting step-by-step illustrated practice sequences calibrated to the user's tier.The gamification and world-building layer was architected as a state machine tracking skill mastery across three levels — encoding, repetition, and generalization — with world growth events firing based on specific behavioral thresholds rather than simple completion counts. This ensured that the world's development was a genuine clinical signal rather than a participation trophy.The therapist dashboard was built as a separate interface pulling from the same behavioral data layer, with AI-generated recommendation cards processed through a clinical framing layer that translates raw behavioral patterns into actionable session guidance without crossing into autonomous clinical decision-making — a boundary we were deliberate and careful about throughout.The visual language — chibi illustrated characters, warm island world, Animal Crossing-adjacent aesthetic — was a deliberate clinical choice as much as a design one. Familiar, warm, non-threatening visual environments reduce cognitive load and anxiety in this population. Every color, every animation speed, every character proportion was chosen with the user's sensory and cognitive profile in mind.
Challenges we ran into
The hardest challenge we faced was not technical. It was the tension between clinical precision and cognitive accessibility — and it surfaced in almost every design decision we made. The things that make an app engaging for a general audience — variety, novelty, complex reward structures, rich navigation — are frequently the exact things that increase cognitive load and reduce usability for individuals with ID. Every time we built something that felt exciting from a product perspective, we had to run it through the question: does this serve Maya or does this serve our instinct to build something impressive? That question was uncomfortable and productive in equal measure throughout the entire sprint.The abstraction problem was a constant challenge at the content level. CBT as a therapeutic modality is built on cognitive constructs — thoughts, beliefs, patterns, schemas — that are inherently abstract. Translating these into concrete, visual, immediately actionable micro-practices without losing their therapeutic integrity required significant iteration. It is easy to make something simple. It is genuinely hard to make something simple that still does real clinical work underneath. We rebuilt the micro-practice module architecture three times before we arrived at a structure that satisfied both requirements.Designing the gamification system without introducing punishing mechanics was harder than we anticipated. Gamification's default grammar is punishment — lost streaks, empty bars, missed achievements, comparative leaderboards. Removing all of those default mechanics while maintaining engagement and motivation required building an entirely new reward language from scratch. The flame that never dies, the world that never resets, the community glow that adds without comparing — each of these required genuine invention rather than adaptation of existing patterns.On the technical side, calibrating the AI's adaptive difficulty engine to work meaningfully within the wide functional spectrum of mild to moderate ID — where two users with the same diagnosis might have dramatically different language comprehension, processing speed, and visual literacy — was a significant challenge that we addressed with a tiered functional assessment built into onboarding and a continuous passive calibration loop that adjusts throughout the user's experience rather than assessing once and locking in a level.Finally, building the therapist dashboard in a way that was genuinely useful rather than just data-rich required us to think deeply about the clinical workflow and the realistic constraints of a therapy practice. A dashboard that shows everything is a dashboard that helps no one. Deciding what to surface, what to suppress, and how to frame AI-generated observations as clinical suggestions rather than clinical decisions required careful thinking about the boundary between AI support and clinical judgment — a boundary we took seriously and encoded deliberately into every recommendation card the system generates.
Accomplishments that we're proud of
We are most proud of building something where the clinical purpose and the user experience are genuinely inseparable. In POCO WINS, the flower that blooms in Maya's garden is not a reward for completing a task. It is a visible representation of a regulated nervous system practicing a skill consistently enough for it to begin transferring into memory. The world growth is the therapy. The therapy is the world growth. Achieving that unity — where the engagement layer and the clinical layer are the same layer — is the accomplishment we feel most satisfied by because it is also the hardest thing to do and the thing most wellness apps never achieve.We are proud of the environmental anchoring system — specifically the toothbrushing anchor and its post-brush breathing extension — because it represents a genuinely novel approach to the independent practice problem. We did not build a reminder system. We built a behavior attachment system that makes therapeutic practice happen inside moments that already exist in the user's day, requiring zero new time, zero new motivation, and zero additional cognitive planning. That is a meaningful clinical innovation expressed as a product feature.We are proud of the Learning Fingerprint — the personalized learning profile built passively from behavioral data — because it addresses one of the most persistent failures of digital health tools for this population: the assumption that all users with the same diagnosis learn the same way. POCO WINS knows that Maya learns better with audio in the morning and with visual prompts in the evening, and it adjusts. No other tool we found does this for this population at this level of individual specificity.We are proud of the dignity embedded in every decision. The flame never dies. The buddy never says wrong. The therapist sees the map but not the private journal. The community has no leaderboard. These are not accessibility features. They are expressions of a fundamental belief that the people POCO WINS is built for deserve technology that respects their autonomy, protects their privacy, and never — not once — makes them feel like they have failed.And we are proud that we built something that a therapist, a caregiver, and a person with ID could each look at and immediately understand what it is for and why it matters. That clarity, achieved across three very different audiences, is something we worked hard for and feel genuinely good about.
What we learned
We learned that designing for cognitive accessibility is not a constraint on good design — it is a clarifying force that makes design better. Every time we removed a step, simplified a word, replaced an instruction with an image, or eliminated a decision the user didn't need to make, the product became cleaner, warmer, and more focused for every user — not just users with ID. The principles we applied here — concrete over abstract, repetition over novelty, reinforcement over punishment — are good design principles for any population. This population just makes the consequences of violating them more visible and more immediate.We learned that the between-session gap is not a content problem. Every solution we initially considered was a content solution — more skills, more modules, more exercises. The real problem is structural. Skills don't transfer because the conditions for transfer — repeated practice in varied contexts, with immediate feedback, in the moments when transfer is actually needed — don't exist in daily life without deliberate architectural support. POCO WINS is that architecture. Building it taught us that the most important thing technology can do for therapy is not to teach more — it is to be present more.We learned that AI is most clinically valuable in this context when it is invisible. The most powerful AI features in POCO WINS — the adaptive calibration, the situation remix engine, the passive learning fingerprint — are ones the user never directly interacts with or even knows are running. The AI's job is to make everything around it work better for this specific person. The moment AI becomes visible as AI in this context, it risks introducing abstraction, unpredictability, and cognitive load that undermines the experience it is supposed to support.We learned that gamification built for this population requires inventing a new reward grammar from the ground up — not adapting existing patterns. That was harder, took longer, and produced something we are significantly more proud of than anything we could have built by modifying what already exists.And we learned something quieter that will stay with us beyond this hackathon. The individuals POCO WINS is built for are not waiting for technology to notice them. They are managing real depression, real anxiety, and real daily challenges with fewer tools, less support between appointments, and less representation in the design rooms where those tools are built than almost any other population. Building something genuinely for them — not for an average user with an accessibility mode turned on — felt important in a way that went beyond the competition. We hope the work shows that.
What's next for PocoWins
The immediate next step is clinical validation — partnering with adapted CBT practitioners and ID-specialized psychologists to run a structured pilot with real users across a range of functional levels. The Learning Fingerprint and between-session data the app generates creates a genuine research opportunity: for the first time, there is a continuous behavioral record of how individuals with ID practice, generalize, and retrieve therapeutic skills in real life rather than in controlled settings. That data, handled ethically and with full participant consent and agency, could meaningfully advance the clinical understanding of skill transfer in this population. On the product side, the most important next development is the voice interface layer — replacing text interaction entirely for users who need it with a fully audio-navigated version of the app where the buddy speaks every prompt and the user responds by tapping large illustrated options or speaking simple words. For users at the lower end of the functional spectrum, a text-reduced experience is not an accessibility option. It is the only viable interface. Building it properly will extend POCO WINS's reach to a significantly larger and more underserved portion of the ID population. The Spanish language version is a near-term priority — specifically for the Miami community the brief identifies, where a significant portion of users and families are Spanish-speaking and where the name POCO WINS already carries bilingual resonance. This is not a translation project. It is a cultural adaptation — ensuring that the scenarios, the buddy's language, the community content, and the caregiver communications feel native to Spanish-speaking families rather than translated from an English original. The caregiver platform needs significant development beyond its current form. The caregiver is the most present support figure in most users' daily lives and the most underutilized resource in most digital mental health tools. A dedicated caregiver interface — with simple daily context about what the user practiced, what their mood has been, and one concrete suggestion for how to reinforce today's skill in natural conversation — would dramatically amplify the app's real-world impact without requiring caregivers to become clinical practitioners. Longer term, POCO WINS has the potential to become an infrastructure layer for adapted CBT delivery more broadly — a platform that therapists can use to build and assign custom skill modules, that researchers can use to study generalization in naturalistic settings, and that advocacy organizations can use to extend therapeutic support to individuals who lack consistent access to specialized clinical care. The between-session gap is universal in this population. The solution we've built is scalable. The next chapter of POCO WINS is making sure it reaches every Maya who needs it — not just the ones whose therapists happened to be in this room today.
Log in or sign up for Devpost to join the conversation.