‹ All Tutorials

Tutorial, 14 Nov 2026, 14 min read

How I read 100 sources a day without reading 100 sources

The full workflow behind this site: where the feeds come from, how the AI scores them, and the one step I refuse to automate.

This post contains a Hostinger affiliate link. If you sign up through it I earn a commission and you pay the same price.

Every morning this site shows a handful of links. Behind them is a workflow that reads far more than I could.

The three stages

  1. Collect. An n8n workflow pulls new posts from RSS feeds, newsletters and a list of accounts I follow.
  2. Filter. Each post gets a summary and a score from an AI model. Anything below the cut-off is dropped.
  3. Pick. I read the shortlist myself and choose what goes on the site.

The scoring prompt

The prompt that works best for me does not ask whether a post is important. It asks:

Would I forward this to a friend who builds things with AI?

That one change removed most of the announcement-style posts from my shortlist.

The step I don’t automate

The verdict. Every pick on this site has a note saying whether I would try it, watch it or skip it. A model can summarise a post, but it has not run the tool on my laptop.

Where it runs

The workflow runs on the same Hostinger VPS as this site. Starting it is one command:

docker compose up -d