Website Build

Building a wellness platform for vulnerable audiences presented unique technical and ethical challenges that required careful consideration beyond typical web development.

Safety and Responsibility: The primary challenge was creating an AI-powered affirmation generator that would never produce harmful content for users in crisis. This required architecting multiple safety layers—AI prompt engineering, server-side content filtering, and fallback mechanisms—while maintaining the authentic, non-toxic-positivity voice that differentiates the platform. Balancing genuine support with safety guardrails without creating sterile, corporate-feeling responses demanded extensive prompt iteration and testing.

Rate Limiting Architecture: Implementing a rate limiting system that encourages signup without manipulating vulnerable users posed both technical and ethical challenges. The localStorage-based MVP solution needed to balance operational costs (API expenses) with user experience, requiring careful UX writing for limit warnings and conversion modals that feel helpful rather than punishing.

Mobile-First Design for Vulnerable Moments: The target audience often accesses wellness tools during difficult moments—in bed during pain flares, in waiting rooms, during insomnia. This demanded mobile-first responsive design with high contrast for readability, minimal cognitive load, and fast performance, all while maintaining brand authenticity and visual appeal.

**Balancing Features with MVP Scope:** The vision includes blog, forum, e-commerce, and social sharing features, but launching required ruthless prioritization. Determining which features were essential for Phase 1 versus future phases required strategic product thinking to deliver value quickly while architecting for future expansion.

Brand Voice Consistency: Maintaining the authentic, permission-giving voice throughout all copy—from affirmations to error messages to rate limit warnings—required careful attention to ensure the platform never slipped into toxic positivity or corporate wellness language.

Our team addressed these challenges through strategic technical architecture and thoughtful product design decisions.

Multi-Layered Safety System: We implemented three concurrent safety mechanisms for the affirmation generator. First, comprehensive safety guidelines embedded directly in the Claude API prompt instruct the AI to avoid harmful content patterns. Second, a server-side regex filter scans all generated affirmations for concerning keywords before delivery. Third, a fallback system with five pre-vetted safe affirmations deploys if harmful content is detected. This layered approach ensures user safety while maintaining authentic, helpful messaging.

Intelligent Two-Tier Rate Limiting: We designed a progressive rate limit system with distinct user tiers: 5 affirmations daily for anonymous users and 20 for signed-up users. The system provides a gentle warning at 15 affirmations and presents a value-focused signup modal at the limit, framing email signup as unlocking more support rather than punishing usage. Copy emphasizes permission and help rather than restriction.

Mobile-First Responsive Architecture: Using Tailwind CSS v4 with fluid typography (clamp() functions), we created a design system optimized for mobile consumption with touch-friendly button sizing, high-contrast color pairing (dark blue #2C3845 + yellow #F6DD1C), and generous spacing. The component library includes mobile-optimized cards and modals with simplified layouts for small screens.

Phased Roadmap with Scalable Foundation: We built Phase 1 MVP with Next.js 16 and TypeScript, creating reusable component architecture and centralized design tokens that support future expansion. Documentation includes detailed roadmaps for Phases 2-4, cost analysis, and migration paths (localStorage to server-side rate limiting, ConvertKit integration, CMS addition), ensuring the codebase scales without technical debt.

Brand Voice Documentation: We created comprehensive copy guidelines and component variants that encode the authentic voice into the design system itself, ensuring consistency as the platform grows.