This SICP lecture shows some of the power of Lisp by building a symbolic differentiator in a very small amount of code, and is also the first one to introduce the idea that code is data and data is code (watch from the beginning to 29:00 — at 13:00 he starts talking about using Lisp as the data format):

Here is the first symbolic differentiator he writes (but, in clojure)

    (defn atom? [x]
        (not (instance? clojure.lang.IPersistentCollection x)))

    (defn const? [expr val] (and (atom? expr) (not= expr val)))
    (defn same-var? [expr val] (and (atom? expr) (= expr val)))
    (defn arith? [expr sym] (and (not (atom? expr)) (= (first expr) sym)))
    (defn sum? [expr] (arith? expr ‘+))
    (defn prod? [expr] (arith? expr ‘*))

    (defn a1 [expr] (frest expr))
    (defn a2 [expr] (first (rrest expr)))
    (defn m1 [expr] (frest expr))
    (defn m2 [expr] (first (rrest expr)))

    (defn deriv [expr val] (
        cond
            (const? expr val) 0
            (same-var? expr val) 1
            (sum? expr)
                (list ‘+ (deriv (a1 expr) val) (deriv (a2 expr) val))
            (prod? expr)
                (list ‘+ (list ‘* (m1 expr) (deriv (m2 expr) val))
                    (list ‘* (m2 expr) (deriv (m1 expr) val))
                )
            (:else) nil
        )
    )

The things to note:

1. Quote syntax, which uses ‘
2. first and rest which are car and cdr. Clojure also supports combinations like frest (cadr) and rrest (cddr)
3. In clojure, extra parens are usually not necessary — look at how cond is implemented to use variable arguments rather than a single list argument (which would require extra parens)
4. My first usage of Java interop (using instance? to find out if an expression is atomic or a collection)