My AI newsletter costs $0.01/user/day to run. Getting users is the actual hard part.

Been building Daily Dose as a side project for the past several months alongside my day job in investment research. It's a personalized morning email where you toggle on the sections you want (markets, sports, geopolitics, local news, even anime), pick your teams and tickers, set a delivery time, and every morning Claude writes you a fresh brief from live data.

Where it stands: 41 paying/trial users, 4,000+ briefs delivered, and ~70% of users open and read it every single day.

A few things I didn't expect:

The AI cost is a rounding error. Each brief costs about a penny to generate. Everyone asks about AI costs and it's genuinely nothing for these briefs. At $4.49/mo the unit economics were never the question.

Retention solved itself, distribution didn't. Users who make it past week one basically don't churn because the brief becomes part of the morning routine. But getting people to the front door is brutal.

Explicit config beats "learning" algorithms. I don't infer what you like, but you tell me your exact tickers, teams, and city, and brief #1 is fully personalized.

Happy to answer anything about the stack (Next.js + Railway + Postgres + Anthropic API), the prompt architecture, or what it's like running a cron job that emails real customers at 7:30am and cannot fail.