The neuron is a “structure/label” detector – it spikes on standalone technical or heading-style words (things like “decreasing order” in math, “Power”/“Energy” in a news heading, “org” in a URL, etc.), i.e. words that function as formal section labels, topic headings, or domain names.
