When we first sketched the idea for ADHDwear, the goal was simple: give people with ADHD a lightweight, offline companion that could remember reminders, respond to voice commands, and calm them in stressful moments. The product was a web app called Felix, a voice buddy that listens for a wake word and can pair users with a partner for a “Peace mode.” Behind Felix was a tiny on‑device language model, a minimal Express + SQLite backend, and a handful of pre‑orders for a necklace, ring, clip, bracelet and witty tees. It sounded like a niche project, but the journey taught us more about designing for ourselves than it did about building a consumer brand.
Design with Presence in Mind The first lesson came from the requirement that Felix work offline. We had to fit a 48 MB MiniLM model into the phone’s cache and run it with transformers.js and ONNX runtime. That constraint forced us to strip the app down to its essentials: no heavy frameworks, no unnecessary analytics, no pull‑to‑refresh that could distract the user. In our own workflows, we adopted a similar mindset. We began to question every feature we added to internal tools or client sites: does it serve the user’s immediate need, or just add noise? The result was a cleaner, faster experience for clients and a leaner, more focused codebase for us.
Give Your Product a Personality, Then Test It Felix’s default personality, “Felix,” was a playful, non‑judgmental voice that could be named by the user. We built a voice‑buddy that could respond in different tones, and we let users choose a voice or a personality. Before launching, we ran a small beta with a handful of friends and family, asking them to give feedback on the wake word, the clarity of the voice, and the usefulness of the reminders. That rapid, low‑cost testing cycle revealed that people wanted a clearer wake word and a more flexible reminder trigger. We pivoted quickly, adding location‑triggered reminders and partner pairing via push notifications. The same iterative approach works well when building internal systems: test with a small group, gather honest feedback, and iterate before scaling.
Embrace the Unexpected, Then Adapt When we launched the app and the store together on October 15, we were optimistic. Yet the first day we noticed that the on‑device model didn’t persist across app restarts on some phones. We had assumed the cache would survive, but the reality was that certain devices cleared the cache after a reboot. Rather than delay the launch, we added a quick fix that copied the model to the app’s internal storage during installation. The lesson here is that you can’t predict every edge case, but you can respond quickly and keep momentum. In our own projects, we now build in “fail‑fast” checks and design for graceful degradation, so a single bug doesn’t stall the entire release.
Build a Brand That Reflects Your Values After releasing the initial dark‑theme pitch deck, we received feedback that it felt too clinical for a consumer product. We rebuilt the deck in a warm, playful style using Fredoka and Nunito fonts, and a palette of cream, sun, coral, and teal. The redesign didn’t just look better; it aligned the visual identity with the product’s purpose: to feel like a friendly companion rather than a medical device. This experience taught us that branding is not a separate step; it’s intertwined with product design. When we create internal dashboards or client websites, we now ask: does the visual language reinforce the user’s emotional journey?
The Power of Offline Autonomy One of the most surprising aspects of Felix was its ability to understand commands offline. We had to train a small language model, compress it, and run it in the browser. This process taught us the value of autonomy: users don’t need a constant internet connection to get help, and neither do we need a constant server connection to deliver value. In our own work, we moved some of the heavy lifting to the client side, reducing server load and improving responsiveness. We also learned that giving users control—like choosing when to sync or when to go offline—creates a sense of ownership that translates into higher engagement.
A Mirror for Personal Growth Building ADHDwear forced us to confront our own distractions and habits. The wake‑word feature reminded us that we need to listen actively to our own internal signals before we can respond effectively. The partner‑pairing feature showed us the importance of community and support. When we launched the first pre‑orders, we had to manage expectations, communicate delays, and keep the conversation open. These were the same skills we use when we lead projects for clients: active listening, clear communication, and a willingness to adapt.
The Takeaway Creating a voice‑buddy for ADHD was more than a product launch; it was a practice in mindful design, rapid iteration, and empathetic communication. The constraints of offline operation, the need for a relatable personality, and the necessity of quick fixes all mirrored the challenges we face in building immersive 3D websites, CRMs, and festival apps. By treating our own lives as a living product, we can build tools that serve us better, stay present, and adapt gracefully to change.
How will you redesign one part of your daily routine to make it more mindful and responsive?