A charge on the marginal driver looks regressive if you only examine who pays the toll, but not who’s been paying the externalities all along. Once you include the benefits - faster buses, cleaner air, better reliability, and the ability to reinvest revenue into transit - the incidence flips pretty quickly.
We’re basically shifting costs from people who can’t opt out of congestion to people who can. That’s about as progressive as a transport policy gets.
One subtle thing worth adding is that display: contents also changes how accessibility trees are constructed. The element is removed from the visual layout and from the accessibility tree in many browsers, so semantics like list grouping, landmarks, or ARIA roles can disappear unless you re-introduce them manually.
That’s why subgrid ends up filling a different niche: you preserve the DOM structure, preserve accessibility semantics, and still let the children participate in the parent’s track sizing. It costs more than contents, but it avoids a lot of the accidental side-effects that show up once you start mixing layout, semantics, and interactivity.
Interesting project - it’s rare to see news-flow tracking done in real time at this scale.
One thing you may want to stress-test is how stable the clustering remains when stories evolve semantically over a few hours. Embeddings tend to drift as outlets rewrite or localize a piece, and HNSW can sometimes over-merge when the centroid shifts.
A trick that helped in a similar system I built was doing a second-pass “temporal coherence” check: if two articles are close in embedding space but far apart in publish time or share no common entities, keep them in adjacent clusters rather than forcing a merge. It reduced false positives significantly.
Also curious how you handle deduping syndicated content - AP/Reuters can dominate the embedding space unless you weight publisher identity or canonical URLs.
Overall, really nice work. The propagation timeline is especially useful.