Mobile MoleCheck: Swipe to Flag Lesion Concern

This is a mobile-first tool for screening perceptions of skin lesions. Users swipe or tap to indicate whether an image looks concerning, not concerning, or if they’re unsure. A desktop fallback is available, but the experience is optimized for phones.

Key Points

• The app is optimized for mobile use and can be accessed via a QR code.
• A desktop continuation option is provided as an alternative.
• Users assess whether a skin lesion appears concerning.
• Interaction is via swipe gestures (left for concerned, right for not concerned) or buttons.
• An 'I'm not sure' choice and navigation to the next image are available.

Sentiment

Overwhelmingly positive. The Hacker News community embraced this as a heartwarming example of AI coding democratizing application development. The combination of genuine medical expertise with a simple, useful tool resonated strongly. Criticism was constructive and minor, focused on improving the app rather than questioning its premise.

In Agreement

• The app is an excellent example of AI coding tools empowering domain experts to build practical applications they couldn't have built otherwise
• Patient education about what skin cancer looks like is genuinely valuable and could lead to earlier detection and better outcomes
• The gamified approach mirrors how dermatologists actually learn — through repeated pattern recognition with feedback
• Vibe coding lowers barriers to entry, allowing creative professionals to act on ideas without needing development teams or large budgets
• The simplicity of the implementation (single file, no backend, no dependencies) is a strength, not a weakness

Opposed

• The dataset is heavily skewed toward cancerous images (~75%), making the task feel unrealistically alarming and potentially less educational
• Images are too zoomed in — wider-angle views would better represent how people actually check their own skin
• The binary concerned/not-concerned classification is oversimplified compared to real dermatology, which involves history-taking and distinguishing among thousands of conditions
• A programmer using AI to understand skin cancer could theoretically build the same thing, and the 'vibe coding' label adds little beyond buzz
• Without more diverse lesion types (infectious disease, benign conditions that mimic cancer), the learning experience may give false confidence