Most news briefings tell you what was published. This one tells you what changed, and whether it matters.
It reads about 120 sources across AI labs, research, tech and business press, markets, central banks, policy, geopolitics and science. Then it keeps only stories that pass a simple test: did something actually change, and would knowing it change a decision or how you understand something?
What you get
A tabbed HTML page you can read in about five minutes:
- Today's Brief: the five to seven things worth knowing.
- AI Technical and AI Business: new capabilities, and who's gaining or losing ground.
- Markets, Economics & Policy, Geopolitics, Culture & Science: what moved and why.
- For You (optional): the few stories that matter for your own work, based on a short context file you fill in once.
A market strip sits on top, and a watchlist, patterns and takeaways sit at the bottom. Quiet sections say "No major signal today" instead of filling space.
How it works
The split between code and model is what keeps it cheap. Python fetches the feeds, removes duplicates, groups coverage of the same event, and renders the page. The model only reads a ranked shortlist and makes the editorial calls: what made the cut, what it means, how confident to be. An earlier version had the model write the whole page and cost roughly five times as much per run.
It also remembers what it covered, so a story only comes back when something new happened.
Try it
- "Morning brief"
- "Daily digest, last 72 hours" after time away
- "Morning brief, focus on markets"
Install
In Claude Code, copy the folder into your skills directory:
git clone https://github.com/ecemkaraman/ai-skills.git
cp -r ai-skills/skills/daily-digest ~/.claude/skills/
Then run pip install -r ~/.claude/skills/daily-digest/requirements.txt and add your email to sources.yaml, which some government data feeds require. This one runs Python scripts on your machine, so it works best in Claude Code.