Watch my sources

Reads each source marked Yes in the list sources_list_id (signal-sources when left out; made with its fields name, address, feed, kind, watching and the rest when missing) and keeps every item it has not kept before as a row in items_list_id (signals when left out). The first read of a source keeps what it lists as a baseline and brings nothing in; later reads keep what is new, ask your own AI once for all of them whether each moves a pursuit directly or carries an insight from elsewhere for a pursuit or a content format, and hand only those to your AIs. The rest are set aside. A page is watched by whether its words changed; an X account needs a feed address; a YouTube channel is read from its feed or, when YouTube refuses that, its videos page. A source that answers 'not changed' costs nothing. Reads at most 25 sources and weighs at most 20 items a run; the rest wait for the next. Sends nothing outward. Returns read (sources read), kept (rows added), moves ([{ title, link, pursuit }]), set_aside, waiting (new items not yet weighed), problems ([{ name, problem }]) and sources_waiting.

What it asks for

• sources_list_id (Optional, words): The list of places to watch (signal-sources when left out)
• items_list_id (Optional, words): The list the items are kept in (signals when left out)
• most_per_source (Optional, number): How many of a source's newest items to look at each read (10 when left out)

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.

Say to your AI: “Use Trillion to watch the places I follow (any site's feed, a YouTube channel, a blog, a changelog, a page) and keep each new item as a row, weighed against what I am pursuing; only what moves a pursuit is handed to my AIs”

What your AI calls: run_tool {"tool_id":"tool:watch-sources"}

• Connect your AI