Yesterday's news, summarized by an LLM, in your inbox at 7am

You want a daily digest: pull twelve RSS feeds, send the contents to an LLM with a "summarize the interesting bits" prompt, email the result to yourself at 7am. Traditional path: a Python service somewhere: feedparser, an LLM client, an SMTP library, a cron entry, a server to run it on.

You schedule one hoody-curl request for 0 0 7 * * * against an exec endpoint. The script fans out twelve curl calls for the feeds, one to ai.hoody.com with the content as the prompt, one to your email provider's HTTP API. Every leg is persisted, retried on failure.

The whole pipeline is data. No service running. No process to monitor. Tomorrow's run will use the latest feeds without you touching anything.

The pipeline is three URLs in a cron expression.