Thanks. Will check out “Email para newsletters” next and fix it.

On the influence map: I’m still trying to make it genuinely useful and easier to read. Right now I’m tweaking a few knobs, mainly the time-decay [exponential decay, λ=0.15] and the way communities are detected, to see what gives the graph a better shape. Ideally, the clusters should start reflecting different generations of developers, with old-timers in one part of the graph and newer gen in another. If that works, someone like sipa should show up as a bridge between those groups [high betweenness centrality].

You’ll also notice the shaded box on the graph. That is meant to capture expertise clusters, basically grouping people by what they tend to talk about. For that, I’m building topic fingerprints and comparing them using similarity scores [cosine similarity, Louvain detection]. I think we still need to be smarter about how we assign expertise, because the top contributors will likely show up across multiple areas rather than fitting neatly into one bucket.

The harder part is that these expertise clusters do not line up neatly with the main network, which is built from who replies to whom [reply graph, PageRank-weighted]. So the challenge now is how to show both layers without making the whole thing confusing. Maybe that means switching between views, or maybe using color and layout in a better way.

Let’s see if I can make the graph genuinely insightful. Otherwise, it’s just a profile map of all devs.