IMA
A live AI and technology news feed built to surface signal without forcing you to hunt across sources.
- Category
- AI / Product
- Year
- 2026
- Role
- Builder / Product Engineer
- Status
- SHIPPED
IMA is a small product experiment built around a simple idea: technology news is abundant, but useful signal is fragmented.
The product pulls live AI and technology stories into one feed instead of asking the user to repeatedly check separate publications and research sources.
Problem
Keeping up with AI usually means bouncing between news sites, Hacker News, research feeds and whatever social platform happens to be shouting that day.
The problem isn't a lack of information. It's fragmentation.
Approach
IMA uses public RSS feeds and normalises them into a single internal news model.
The current implementation pulls from:
- TechCrunch AI
- The Verge AI
- Hacker News
- ArXiv
Each item is normalised with a title, source, category, publication time, summary and article URL.
The feed uses short-lived in-memory caching, and failed sources are skipped rather than allowing one unavailable feed to break the entire experience.
Design
The interface is intentionally closer to a compact information console than a traditional news homepage.
Articles are presented in a responsive feed with source filters, category labels and relative timestamps. The design prioritises scanability over turning every article into a visual event.
Technical
Built with Next.js, React and TypeScript.
RSS feeds are parsed and normalised into a common news model. The application also includes testing infrastructure and handles individual feed failures gracefully.
Learning
The useful lesson was that a live product does not necessarily require a complicated data stack.
A small number of reliable public feeds, a normalised data model and sensible failure handling were enough to build a useful first version.