what can one search tell us?
all the topics we look for are below. most of the lights are off: any given search contributes only a small, randomized slice of this vocabulary. in a one-loaded-die-rolling-among-four-fair-dice sort of way, randomly selected values may travel alongside or replace the closest match to your search, camouflaging the signal or replacing it with noise. because even the closest match isn't guaranteed to make the trip, neither we nor anyone else can say for certain which topic matched your search, or whether the intended signal is present at all.
considered → counted: five colored outlines show the attributes considered. bright text and solid outlines show which survived and were counted; dim text and dashed outlines show which were discarded.
- coffee
- tea
- breakfast
- lunch
- dinner
- bakery
- groceries
- takeout
- vegetarian food
- food trucks
- shared meals
- clothing
- shoes
- household goods
- hardware
- electronics
- books
- gifts
- secondhand goods
- locally made goods
- repair services
- recycling drop-off
- recycling pickup
- electronics recycling
- battery recycling
- glass recycling
- composting
- bulky-item disposal
- hazardous-waste disposal
- donation
- reuse
- public transit
- walking
- bicycling
- rideshare/taxi
- parking
- airport transportation
- intercity travel
- vehicle rental
- EV charging
- vehicle repair
- hotels
- short stays
- camping
- visitor information
- luggage storage
- public restrooms
- showers
- laundry
- drinking-water refill
- places to sit
live searches are checked locally using keyword rules and common synonyms. no external classifier receives your search. the illustration uses examples with one matching attribute.
one slice looks like noise. but the haystack reveals a shape.
each thin strip represents a slice taken from a search. use the slider to stack more slices and watch for a vertical density to emerge.
scrub through 1 → 10 → 100 → 500 → 2,300 slices.
example slices consider one matching topic and four random attributes, then independently include each with probability 65% if true or 25% if false. these results are generated from example data, this chart does not display live searches.