Narrative lens

Hero section
How it works landing page
article space
real user test
test result success
compare perspectives news article test
more results from compare perspectives

Inspiration

News is written for a general audience that doesn't really exist, dense language, one-sided framing, and no accessibility for younger readers or anyone who just wants the facts without the noise. I wanted to build something that keeps the facts intact but lets the reader choose how the story reaches them.

What it does

Narrative Lens takes a news article and rewrites it through different lenses:

Teen-Friendly clear, simple language for younger or newer readers
Neutral stripped of emotional language, facts only
Story the same facts told as an engaging narrative
Comic Script a panel-by-panel comic-style rewrite

It also has a Compare Perspectives mode that generates the Teen, Neutral, and Story versions all at once, so you can see how the same event reads differently depending on the framing- side by side, in a clean collapsible layout. It's not a chatbot and it's not a summarizer, it's a narrative transformation engine.

How we built it

Built with Next.js (App Router) and TypeScript, styled with Tailwind CSS using a custom dark theme inspired by Linear's minimal UI language and Notion's block-based content layout. The rewriting itself is powered by Llama 3.3 70B through Groq's API , the Compare Perspectives mode fires all three requests in parallel so the wait time stays close to a single generation. The app is deployed on Render.

Challenges we ran into

Balancing scope against a tight timeline the original idea included URL scraping and real bias detection, both of which are much harder problems than they first appear (fragile scraping against paywalls, and genuine bias detection being closer to a research problem than a prompt). I cut both to focus on making the core narrative transformation feature and the Compare Perspectives view solid and polished instead of shipping several half-finished features.

Accomplishments that we're proud of

Getting a fully working, deployed AI-powered tool live end-to-end, real API integration, input validation, error handling, and a design system built from scratch rather than a template within a single hackathon weekend.

What we learned

How much of building a good AI product is actually about scope discipline cutting ambitious features early to protect the ones that matter most to the core idea, and how much a clear visual identity changes how "finished" a project feels even with a small feature set.