Comment by davi
16 years ago
I'm curious about how you evaluate prospective algorithm changes. Do you roll out a change you think should work and then monitor, or do you have a corpus of e.g. flame wars you test against?
16 years ago
I'm curious about how you evaluate prospective algorithm changes. Do you roll out a change you think should work and then monitor, or do you have a corpus of e.g. flame wars you test against?
There's a repl on the server, and I test tweaks on the live site.
Do you have a dashboard of some kind to visualize your tests effects? I am curious about the broad kind of insight one could have with such a programmatic access to this community. Perhaps I missed an essay about HN/REPL?