Trust & sources

Hot ranking

score = sources × 2−age_hours ÷ 6

The board's "Hot" tab ranks incidents by this score, computed in SQL at query time. Sources counts non-removed timeline rows. Age counts from the last timeline touch. Latest order stays the default tab. Rider timeline rows pass the moderation gate before they land on a timeline.

Line and station facets

• Lines and stations come from the rail catalog behind the transit page.
• The line facet pins, it does not filter: incidents carrying a bridged rider report (a threads-rider source) that passed the review gate, from that line, move to the top. Every other incident stays on the board below them — nothing is hidden.
• The station facet narrows the board to that station's incidents; clearing it restores the full board.
• An unknown line or station returns an error naming the accepted values. A bad facet value never reads as "no results".

Sources and reliability

• Every source timeline row carries a reliability class, computed from stored fields only: platform news or official renders news/official; any other platform with a handle renders named handle; no handle renders anonymous social. Classes are static and ordered news/official > named handle > anonymous social. No numeric score is rendered anywhere.
• The corroboration chip counts what the timeline already shows: corroborated · N platforms · M sources, plus a news marker when a news/official row exists. One source renders "single source" — never padded. Zero sources render no chip.
• Corroboration is display-only. It never feeds status or settlement; weekly settlement stays the only status setter.
• Transit rider reports are review-gated before render: only review-passed rows join incident timelines. The station pages state the gate on every page; this section documents it as a trust layer.