The neuron activates on programming‐language tokens—keywords, module/import directives, literals (like numbers or True/False), and code identifiers (e.g. “unsafe,” “jar,” “module,” “import,” numeric literals)—i.e. it spots bits of source‐code syntax.
