Directions and methods under active investigation
75,118 per-item metric snapshots over 175 days
17 sources across 5 lenses; 6,118 items annotated
Real decisions as samples: the call, its basis, and what followed
When several source categories mention a topic at once, it is usually spreading
Nodes in different regions may see different recommendations — the difference is itself data
The ability to judge must come before the ability to optimise
Finished work is worth one visit; work in progress is worth returning to. None of these has a conclusion yet — some will hold, some will be overturned. The overturned ones matter too: they save the cost of trying again.