The signal model
Every interaction pays a different rate. A project's signal score is a weighted sum of its views, clicks, likes, and comments — weighted toward the actions that mean someone actually engaged, not just looked. It's computed once overnight, not on every page load, so the number you see might be a few hours behind the most recent clicks — that's by design, not a bug.
Clicks
40%
Likes
25%
Comments
15%
Views
20%
Clicksweigh highest — someone opened the link and left the board to look at the actual project. That's the strongest signal a project earned attention.
Likes come next: a lightweight but deliberate vote. Comments take real effort to write, but they're rarer, so they carry less total weight despite the higher bar per action. Viewsare the easiest to rack up and count for the least. You can comment more than once on a project — that's still a real discussion — but only your first comment on each project adds to its signal, and anything a reviewer hides for being spam doesn't count. A flagged IP's activity is excluded only after a reviewer confirms abuse.
Raw counts are normalized before weighting. Your project's views, clicks, likes, and comments are each compared only to other projects published in the same calendar month — its "cohort" — not to every project ever shipped. Counts are compared with the 90th percentile when the cohort has at least 10 projects, or the median for smaller cohorts. The normalized value is count divided by count plus that benchmark, so each extra interaction adds less than the previous one without a hard cap at the benchmark.
Scoring window: activity counts from publication through the end of that calendar month. Scores update nightly during the month and are final after it ends. This means an early-month project has more days to gather activity than one published near month-end. Signed-in owners and contributors cannot add signal to their own project (they can still like it, but the like does not count toward the score). Anonymous visits are not tied to an owner by IP because shared campus networks can include unrelated people.
Review is separate from ranking. A reviewer checks that a project is real, coherent, and not spam or plagiarism before it ever reaches the board — signal only decides order among projects that already passed that bar.