Build an AI News Radar from RSS

Build an AI news radar by collecting public RSS and Atom feeds into one list, then ranking that list for each audience. A creator might want an open-source project to investigate, while an ecommerce team wants a workflow it can test this week. Keep the source link and selection reason beside every story so readers can decide what to open.

What makes a source worth adding?

Start with three to five public sources that can change a decision, such as product release notes, practical tutorials, or developer projects. Review three recent entries from each candidate. Can someone open the original item, tell when it was published, and explain which team would act on it? Record the feed address, topic, source weight, and intended audience. A working feed confirms you can collect its entries. Individual stories still have to earn a place in the digest.

SCTCamp's open-source collector requires a title and link and derives its feed item ID from the source name, title, and link. A story syndicated by two different sources can therefore produce two IDs. During editorial review, group those copies by the original publisher's URL where you can verify it. Keep publication and collection times distinct so older items fetched today don't look like breaking news. RSS 2.0 specification · Feed collector

How can two audiences rank the same story differently?

The public example adds points for source weight, matched terms, topic, and GitHub stars. Each matched strong term adds 3 points, each ordinary term adds 2, and each downranking term subtracts 2. Both example profiles set a minimum score of 8 and keep up to eight items, with at most three per source. The two profiles use different terms and topic weights. Scoring implementation · Profile configuration

Running the published scoring code on 14 fictional entries selects eight for each audience. Three rows show how the same inputs lead to different editorial choices.

Fictional sample	Editor score	Learner score	What happens
Agent task board	22	6	Makes the editor list and falls below the learner threshold
n8n support ticket template	11	23	Makes both lists and ranks first for learners
Airtable content scheduling automation	6	13	Makes the learner list and falls below the editor threshold

The editor profile rewards the task board's Agent and MCP terms. The learner profile rewards the ticket template's n8n, workflow, and template terms. When calibrating the weights for your team, take one useful item the rule missed and one weak item it ranked highly. Read their matched terms before you change the list. The 14 fictional inputs

How do you reproduce the digest?

Install the dependencies for the public SCTCamp repository and run uv run radar demo. It reads bundled fictional inputs and writes radar.json and an HTML report. The demo run uses no credentials or live feeds. You can inspect the output in the English browser demo first. To test real feeds, run uv run radar fetch --limit 3 to check the first three sources, then run uv run radar digest to produce a report. Repository instructions · English demo

Before publishing, open the top story for each audience and check that the summary stays within what the source shows. Review a high-scoring omission to see whether a source or item limit excluded it. Give each selected entry the event, why it matters for that audience, and a next action when one is justified. SCTCamp publishes its current briefs separately in a dated archive and RSS feed. SCTCamp brief archive · RSS feed