Behind the briefing

From scattered news
to a daily signal.

AI news moves quickly. The Gradient turns a collection of feeds into one readable edition, then keeps that edition available long after the inbox moves on.

An end-to-end TypeScript project connecting data ingestion, rule-based ranking, AI summarization, and a responsive publication.

Explore the source on GitHub (opens in a new tab)
The product

88 editions · 436 archived stories

The interface

Next.js · React · Tailwind CSS

The pipeline

Node.js · TypeScript · GitHub Actions

01 / The architecture

One pipeline. Six steps.

From original source to published edition.

  1. 01

    Gather

    RSS feeds from research, company blogs, and industry publications are fetched concurrently. A failed feed does not stop the others.

  2. 02

    Deduplicate

    Normalized URLs and similar titles remove repeated stories before they reach the reading list.

  3. 03

    Curate

    A weighted score combines source priority, recency, relevance keywords, and engagement metadata when available. The highest-scoring stories make the edition.

  4. 04

    Summarize

    Groq or OpenAI produces structured summaries. Without an API key, or after a provider failure, an extractive fallback uses the source text.

  5. 05

    Publish

    The pipeline writes Markdown and HTML. GitHub Actions schedules the daily run and commits the generated editions.

  6. 06

    Read & deliver

    Next.js serves the Markdown archive with search and story links. Resend optionally manages contacts and delivers newsletter emails.

02 / Engineering decisions

Built with clear tradeoffs.

Files before a database

Markdown makes editions portable, inspectable, and versioned in Git. This suits a small publication; a larger archive would benefit from an indexed data store.

Explicit ranking

Source, recency, and relevance weights make selection straightforward to inspect. They are editorial heuristics, rather than a personalized recommendation model.

AI with a fallback

Provider summaries give the briefing structure, while extractive summaries keep publishing possible without credentials. Generated claims still require source verification.

03 / The reading experience

Follow the story.
Keep the context.

Search across headlines, summaries, topics, and sources. Jump to individual stories, follow original articles, and move between editions without losing your archive filters.

Explore the archive