Challenge 3: Why solve(X is 1 + 1) Fails, and How to Fix It — Possible Solution
====================================================================

-- Why does solve(X is 1 + 1) fail? --
--
-- X is 1 + 1 matches none of solve/1's first four specialized
-- clauses (it isn't true, a conjunction, a disjunction, or a
-- negation), so it falls through to the generic final clause:
-- solve(A) :- clause(A, Body), solve(Body). This tries
-- clause(X is 1 + 1, Body) -- looking up a USER-DEFINED clause whose
-- head matches is/2. But is/2 is a built-in predicate, implemented
-- directly by the Prolog engine itself, not defined via ordinary
-- clauses the way parent/2 or grandparent/2 are. clause/2 can only
-- see clauses that actually exist in the loaded program's database,
-- and no such clause exists for is/2 -- so clause(X is 1 + 1, Body)
-- fails immediately, and solve/1 fails right along with it, without
-- ever performing the addition.

-- A fix: special-case built-ins before the generic fallback --

solve(true) :- !.
solve((A, B)) :- !, solve(A), solve(B).
solve(G) :- predicate_property(G, built_in), !, call(G).
solve(A) :- clause(A, Body), solve(Body).

    ?- solve(X is 1 + 1).
    X = 2.

-- Explanation of the fix --
--
-- The new third clause checks predicate_property(G, built_in) --
-- true for predicates like is/2 that the Prolog engine itself
-- implements, rather than something clause/2 could ever look up.
-- When that check succeeds, the clause cuts and simply calls G
-- directly via Prolog's own call/1, handing the goal straight to the
-- real engine instead of trying to reinterpret it through clause/2.
-- This has to be placed BEFORE the generic clause/2-based fallback,
-- so built-ins are recognized and handled before solve/1 ever
-- attempts (and fails) the clause/2 lookup that doesn't apply to
-- them.

WHY THIS WORKS AS AN ANSWER
------------------------------
This traces the exact failure back to clause/2 having no entries for
a built-in predicate, then supplies a genuinely working fix using
predicate_property/2 and call/1, placed correctly ahead of the
generic fallback clause, directly resolving the chapter's own
warn-box gap.