Then there are problems elsewhere in your code; you didn’t share the full thing.
For example if you did not follow my instruction to replace map with doseq, then your swap! would never have been called, for the reasons I explained above.
It will be easier to do that if you follow my advice and move the string/split inside of parse-input, then you only need a single doseq call over the instruction data, rather than worrying about how to map over the splitting and then feed the results into the doseq.
A great excuse to learn about reduce:
user> (let [input ["forward 5"
"down 5"
"forward 8"
"up 3"
"down 8"
"forward 2"]]
(reduce (fn [eax s]
(let [[dir amount] (clojure.string/split s #"\s+")]
(condp = dir
"forward" (update eax :x + (Long. amount))
"down" (update eax :y + (Long. amount))
"up" (update eax :y - (Long. amount)))))
{:x 0 :y 0}
input))
{:x 15, :y 10}
Some questions stemming from the order of arguments passed to the update function, and the commutativity of subtraction - worth of some tests and further exploration.