Little by Little

Inspiration

This project was born from a deeply personal place. As someone who lives with anxiety, depression, and executive dysfunction, I know firsthand how a simple to-do list can become a source of stress. A task that might take a few minutes gets put off for weeks, not out of laziness, but from a feeling of being completely paralyzed.

On the surface, it looks like procrastination. But underneath, it’s the exhaustion of seeing a simple task, like "do the dishes," as ten overwhelming steps: get up, open the dishwasher, unload, load, hand wash, dry, wipe counters, clean the sink... and so on.

The inspiration for Little by Little came from realizing that the key wasn't to force productivity, but to work with my brain. It's about breaking an overwhelming mountain into a single, manageable first step to make starting, which is always the hardest part, significantly easier.

What it does

Little by Little is a task manager that puts mental well-being first. It's built on principles from Cognitive Behavioral Therapy (CBT) to help users make gentle progress.

Its core feature is a mood-adaptive AI that adjusts the entire user experience based on how the user is feeling. The AI uses Cognitive Restructuring to break down daunting tasks into small, achievable steps, directly combating the "All-or-Nothing" thinking that leads to task paralysis. It also includes a non-negotiable Crisis Support feature that detects harmful language and provides immediate access to help resources like the 988 hotline, ensuring user safety is always prioritized.

How we built it

The project was kickstarted using the Bolt.new Vite + React + TypeScript template. The application is built with a modern tech stack, featuring a React frontend styled with Tailwind CSS. State management is handled through React's Context API. The backend is powered entirely by Supabase for secure user authentication, a PostgreSQL database with Row Level Security, and file storage for assets. The intelligent, mood-based task breakdown is driven by the Google Gemini API. The app is deployed via Netlify.

Challenges we ran into

The most significant challenge was in the AI implementation. My initial "Planner/Writer" agent architecture for the "Motivated" mood was sophisticated, but sometimes produced steps out of logical order. I solved this by refining the instructions given to the "Planner" agent, adding a strict rule that forced it to think sequentially from start to finish. This fixed the ordering issue while preserving the advanced two-step agent logic. Another real-world challenge was debugging the custom SMTP setup, which involved methodically testing credentials, ports, and finally correcting the Site URL in the Supabase settings to fix the email confirmation flow.

Accomplishments that we're proud of

We are incredibly proud of successfully translating complex, evidence-based psychological principles from CBT into simple, intuitive, and genuinely helpful app features. Building the sophisticated, multi-step "Planner/Writer" AI agent is a technical accomplishment that allows the app to provide truly nuanced support. Most of all, we are proud of building the app with an ethics-first mindset, making the Crisis Support modal a core, non-negotiable feature from day one.

What we learned

Throughout this project, we learned the immense importance of precise prompt engineering. Simply telling an AI what to do is not enough; you must rigorously define how it should think, the format of its output, and the rules it must follow to get a reliable and safe result. We also learned the value of systematic debugging when integrating multiple third-party services (Supabase, DreamHost, Netlify), and how a single configuration setting can be the key to solving a complex problem.

What's next for Little by Little

This hackathon MVP is the foundation for a much larger vision where Little by Little becomes an invaluable tool in the field of psychology. Our roadmap includes evolving the app to support professional therapy:

Clinical Integration: We plan to introduce a secure, two-way login system. This would allow a patient's therapist to gain real-world insights into their progress between sessions.
Data-Driven Therapy: A therapist could see trends, such as, "My patient used the 'Overwhelmed' mood adjustment 90% of the time this week," allowing them to custom-tailor future sessions to address what's happening in their patient's daily life. This data helps track which treatment strategies are making a real difference.
Deeper AI & CBT Tools: We will continue to enhance the AI, implementing a "What's Next?" loop for a guided, one-step-at-a-time workflow and a "Values Clarification" module to better personalize task suggestions. The ultimate goal is to evolve the AI using Retrieval-Augmented Generation (RAG) to pull from our extensive CBT research in real-time, making it an even more powerful supportive tool.