It's tuned towards discovery, so if you search a topic you'll get results for smaller, new repos that do something around that topic. We deliberately hard downranked common repos but it's also 2 months out of date now since that was to test the waters and we didnt set up recrawling at the time. That said we shared your concerns and have changed things up with how we are approaching it for the beta
So we didn't pick specific repos, but we crafted a function based on some metrics we we're using from Github that had a sharp drop off after a certain point and basically -any- super common or well known repo would be down ranked based on those metrics.
I can appreciate your thought on that but we're not necessarily geared towards the most common per se (though this might be me misunderstanding what you mean) as we have experienced multiple times the most common result being wrong or outdated and the way things are now it takes a long time for those to slide out of the rankings.
We've been asking around for a while now to flesh out what our actual thought is and the description for the problem we're solving right now is "information staleness", you search for something and it leads you to a reddit post but that's outdated by 5 years and then you wind up actually having to do a deep dive and it turns out there was actually a more accurate post from a year ago but it just hasn't crept up to the top yet because everything references the 5 year old post.
With our alpha we actually think we went too much in the other direction we focused on it all being super super new but the reality is there is nuance between different topics for what timeframe information decays in, if that makes sense, and now we're for the beta trying to strike a better balance.
Right, it makes sense. What I end up doing a lot of times, is manually filtering by "last year". It's good to give preference to more recent results. Thanks for the explanation
