DEV Community

Cover image for Show Dev: I built an AI research tool that tracks top labs daily
Flynn L
Flynn L

Posted on

Show Dev: I built an AI research tool that tracks top labs daily

The problem

Keeping up with AI research across labs is impossible manually. arXiv is a firehose. Twitter is noise. Newsletters land 3 days late.

I wanted one place to see what matters. It didn't exist.

What I built

Lattice — an AI research intelligence dashboard that tracks what 20 top labs publish daily.

🔗 layerthelatestinalattice.com

Features

  • Daily paper feed from 20 tracked labs with AI summaries
  • 24 research topics with trend velocity charts
  • Lab profiles with output sparklines and topic focus
  • Weekly digest with most-read papers and breakout topics
  • Collections with BibTeX/Markdown/JSON export
  • No account needed — anonymous, localStorage only

One interesting technical challenge

Matching paper authors to research labs. Papers list affiliations as free text ("Dept. of Computer Science, University of X") — no standard format.

The pipeline: OpenAlex API (by DOI) → arXiv HTML scraping → Semantic Scholar data → LLM extraction as a last resort. All matched against ~500 organizations. Hit rate varies wildly by source —
OpenAlex is great for established venues, arXiv scraping catches the rest.

Stack

  • Next.js 16 (Turbopack, standalone output)
  • Supabase (Postgres + RLS)
  • Vercel (hobby plan)
  • GitHub Actions (daily 6am UTC cron)
  • Semantic Scholar + arXiv + HuggingFace APIs
  • OpenRouter (Gemini Flash) for summaries

Feedback welcome

What features would be most useful? What labs or topics am I missing?

We're also live on Product Hunt today (https://www.producthunt.com/products/lattice-6).

Top comments (0)