I Made GitHub Build My AI News Briefing — EVERY MORNING.
Every morning, GitHub reads 17 AI news feeds and turns them into one page before I finish breakfast.
I built this open-source AI news aggregator with GitHub Actions, Python, RSS/Atom feeds, Git history, and GitHub Pages—no app server or database. I let it run for over 2 months, followed one real OpenAI story through the full pipeline, and looked at what worked, what broke, and what I’d improve next.
LIVE DAILY DIGEST
https://zmcddn.github.io/gitops-ai-ne...
SOURCE CODE
https://github.com/zmcddn/gitops-ai-news
IN THIS VIDEO
Why I stopped opening the same AI blogs, feeds, and news sites every morning
How GitHub Actions collects 17 feeds and builds a daily shortlist
How freshness, source weights, keywords, deduplication, and 14 days of URL memory affect the ranking
How one real OpenAI story moves from selected JSON to a daily digest, Git commit, and public page
What 71 generated editions reveal about the useful parts and the rough edges
What failed during three scheduled runs and how I triggered the workflow manually
What I would change in the next version
The bigger idea goes beyond AI news. The same pattern could power a product-release monitor, research digest, security bulletin, competitor tracker, dependency report, or any recurring workflow where you want the inputs, decisions, and outputs to remain visible.
What would your version watch overnight?
CHAPTERS
00:00 Breakfast Without the Ten Tabs
00:17 When AI News Became a Second Job
01:27 Follow One Headline from Feed to Front Page
03:13 Why This Story Ranked #1
04:54 Can You Trace an AI Summary Back?
05:24 The Page I Actually Use Every Morning
06:36 What 71 Mornings of AI News Reveal
08:11 The Three Mornings It Broke
09:12 What I'd Fix Next
09:53 What Would Yours Watch Overnight?
---
RECORDED SNAPSHOT AND CURRENT NOTE
The video was captured using the project state through August 12, 2026. The live site has continued updating since the recording.
GitHub Models was retired on July 30, 2026. All the editions after that, including the August 12 edition shown in this video, used the project’s deterministic renderer rather than agent summary. All other functions operational (i.e. feed collection, ranking, versioned output, manual workflow, Git history, and GitHub Pages deployment).
---
Technically Carson tests AI tools and workflows on real tasks—showing the outputs, trade-offs, rough edges, and human judgment that still matter.
Subscribe for practical AI-builder field tests with visible receipts instead of hype.
#GitHubActions #AIAutomation #Python