How Intel60 works

Every 30 minutes, a short automated run turns new security news into 60-word cards. It takes about a minute, and readers see the result the next time they open the app.

FREE costs nothing the only step that costs money (about $3–5 a month)
1

⏰ Timer goes off FREE

cron-job.org tells GitHub to start a run every 30 minutes. GitHub's own schedule runs too, as a backup.

2

⚙️ GitHub Actions starts a run FREE

GitHub lends a temporary cloud computer for about a minute and runs pipeline/brief.py.

3

📰 Read the news FREE

Downloads the latest headlines from 8 RSS feeds, plus CISA's list of actively exploited vulnerabilities. About 150 items per run.

BleepingComputerThe Hacker NewsSecurityWeek Krebs on SecurityDark ReadingThe Record CyberScoopSANS ISCCISA KEV
4

🧹 Filter out what's not new FREE

Drops stories it has already processed, stories older than 48 hours, and ads, webinars and promos.

Usually only 0–5 new stories are left. If there are none, the run stops here and costs $0.
5

🤖 Claude writes the cards

New stories go to Claude Haiku (Anthropic) in batches of up to 8. For each story it returns:

Headline (≤10 words)60-word summary Severity: critical / high / infoCategory CVE IDsDuplicate check
About $0.005 per call. Each story is summarized once for everyone, so more readers don't cost more.
6

🔍 Add vulnerability facts FREE

For cards that mention CVEs, it looks up public data. New CVEs are checked again every 6 hours, because scores often arrive late.

CVSS score (NVD)Exploit chance (FIRST EPSS)"Exploited" tag (CISA KEV)
7

💾 Save and publish FREE

cards.json stores all current cards, the app's whole database. state.json is the robot's memory: what it has seen, feed health and spending. Both are saved to GitHub, and the app is republished.

Then three things happen in parallel:

🌐 The app updates

GitHub Pages serves the new cards. Readers open the app, which downloads cards.json and shows one card per screen.

Everyone

🔔 Phone alerts

Critical stories and high-severity watchlist matches go out through ntfy, plus a 7 AM morning brief. Silent from 11 PM to 6 AM.

News topic subscribers

🔧 Owner reports

Failed runs, AI errors, feeds that go down, the Monday spending report, the monthly total, and a budget alert over $8.

Only you

📱 Inside the app

  • Swipe up for the next card
  • Today / Yesterday dividers and filters
  • Severity strip and 3-bar threat meter
  • Text shrinks so every card fits the screen
  • Works offline and updates itself

🛟 When something goes wrong

  • An AI reply can't be read: retried once immediately
  • Still failing: the story is retried on the next run
  • A feed is down: the others keep working
  • A run crashes: alert to you, then "recovered"
The key idea: the expensive work (Claude) happens once per story on GitHub's computer. Readers only download the finished result, which is why 1 reader and 1,000 readers cost the same.