Challenge 3: What the Cut in symbol/2 Actually Prevents — Possible Solution
====================================================================

row_demo.pl (same row_symbols/3, row//4, symbol//2 as the chapter):

row_symbols(Queens, Row, Symbols) :-
    length(Queens, N),
    phrase(row(Queens, Row, 1, N), Symbols).

row(_, _, C, N) --> { C > N }, [].
row(Queens, Row, C, N) -->
    { C =< N, nth1(C, Queens, R) },
    symbol(R, Row),
    { C1 is C + 1 },
    row(Queens, Row, C1, N).

symbol(Row, Row) --> !, [q].
symbol(_, _) --> [e].

Backtracking query directly into symbol//2, bypassing row_symbols/3's normal
single call, to see every reading it's actually capable of producing:

    ?- phrase(symbol(3, 3), Out).
    Out = [q] ;
    Out = [e].      % <-- only reachable WITHOUT the cut

-- Explanation --
--
-- With the cut present, phrase(symbol(3, 3), Out) only ever reports
-- Out = [q], correctly, and only once -- the cut commits the moment
-- Row and the queen's own R unify successfully, so the second clause
-- (symbol(_, _) --> [e].) never even gets a chance to run for that
-- same call.
--
-- If the cut were removed, symbol(3, 3) would ALSO match the second
-- clause on backtracking, since symbol(_, _) unifies with any two
-- arguments regardless of their values -- producing a second, extra
-- result: Out = [e]. That's a genuinely WRONG reading of that
-- particular cell -- column 3's queen really is in row 3, so
-- reporting it as empty (e) is factually incorrect, not just a
-- redundant repeat of the same correct answer.
--
-- Because removing the cut would change WHICH results this predicate
-- can produce -- adding a genuinely incorrect one, not just an
-- inefficient duplicate of the correct one -- this is a RED cut by
-- prolog1-8's own test, not a green one.

WHY THIS WORKS AS AN ANSWER
------------------------------
This backtracks directly into symbol//2 (rather than only calling it
once through the normal row_symbols/3 path) to surface the exact
extra result the cut suppresses, then applies prolog1-8's own
green-vs-red test explicitly to justify classifying it as a red cut.