What we are watching
A standing set of technologies, measured on a schedule. Every band is a share of its own universe, so a rising curve means the topic gained ground on everything else published that year.
How movement is scored. Every band reaches its own verdict and the badge takes the majority. A band only votes with the newest period it has measured, so a column four years behind the others sits the vote out. Three voting bands is the minimum; below that the badge reads unmeasured, and a tie reads steady. The line under each badge gives the count and the date of the oldest evidence behind it. Every column carries its own direction as well, so you can read a row across. Research often climbs while public attention falls, and that disagreement is usually where the story is.
Research attention is each technology’s share of all articles published that year — OpenAlex type:article,has_abstract:true on both the phrase and the denominator. Both halves require an abstract because the phrase search can only match one: counted against every article, a rise in how many papers carry an indexed abstract (58% in 2010, 80% in 2024) reads as rising attention for every topic at once. Unfiltered, OpenAlex also reports 34.2M works for 2026 against ~11M for every year 2015–2024, a bulk ingest of non-article record types that distorts recent years by up to 5×. Without any denominator at all, every topic rises, because science itself grows.
The two most recent years are provisional. A publication year keeps climbing for a year or two as indexing catches up, so those years are measured and drawn but never carry a change. When a value shifts between runs we record it as a revision. Trends need settled years.
A dash means we have no reading. It never means zero. A new technology reads unmeasured until the index next runs.