Inspiration

I grew up seeing my father read the morning paper front to back. Today news is scattered across feeds, half of it unverified, and the classic front page is basically dead. I already run a Turkish AI news site, and I kept wondering: what if AI could bring back the newspaper itself, not just another feed, but a real front page you trust?

What it does

Manşet AI is an AI newsroom for Turkish news:

  • Scan: agents watch sources and pick up stories as they break.
  • Verify: each story is cross-checked against multiple sources before it's accepted.
  • Write: verified stories are written in clean, neutral Turkish.
  • Print: everything is laid out as a classic newspaper front page, with a headline, columns, and a masthead.

How I built it

I built it from scratch in Kiro. I wrote specs for each agent (scanner, verifier, writer, layout) and let Kiro generate and iterate on the code. Agent hooks run the pipeline step by step, and the front page is rendered with HTML/CSS using a vintage print style.

Challenges

  • Turkish news sources are messy, so there are duplicates, clickbait, and missing dates.
  • Deciding when a story counts as "verified" without being too strict or too loose.
  • Fitting dynamic content into a fixed newspaper grid without breaking the layout.

What I learned

Spec-driven development changed how I work. Writing clear specs first made the agents far more reliable than prompting on the fly. I also learned that verification is the hardest and most important part of AI news.

What's next

Daily printable PDF editions, regional editions, and a public "how this story was verified" trail for every article.

Built With

Share this project:

Updates

Submission history