A daily YouTube news channel with no editing team
A daily news Short needs fresh stories, careful facts and a fast turnaround. We built a pipeline that does the whole job each morning, and that would rather post nothing than post something wrong.
The challenge
Automated news is easy to get wrong: stale stories, invented details, and robotic delivery. The channel needed real reporting discipline, applied automatically.
What we built
- 01
Collect
News feeds and community sources are gathered every morning.
- 02
Rank
A fast model picks the strongest stories of the last 24 hours.
- 03
Write with sources
A stronger model writes a three-story script, checking facts against the web as it writes, and names its sources.
- 04
Critique
A policy critic reviews the script and can rewrite it twice. If it still fails, nothing is posted and the reason is logged.
- 05
Render and publish
Voice, visuals and code-drawn charts are rendered, uploaded privately, then published on schedule with the sources pinned in a comment.
- 06
Reply
Viewer comments are classified and answered automatically.
Guardrails
- No story without a source, and the sources are spoken in the video and pinned in the comments.
- A hard daily budget. The supervisor pauses paid steps at the cap, while replying to viewers stays on.
- Rehearsal mode publishes privately so changes can be checked before they go live.
Results
- A daily Short published on schedule for about $0.75 a day.
- Every episode carries its sources, and episodes that fail the critic are held back rather than posted.
What we learned
Under the hood
More case studies
Want agents like these in your business?
Book a free 30-minute AI audit. We'll find the one workflow worth automating first, and what it would cost to run each month.