+Its functions behave the same as their L<Test::More> counterparts, except for the following differences :
+
+=over 4
+
+=item *
+
+Stringification isn't forced on the test operands.
+However, L</ok> honors C<'bool'> overloading, L</is> and L</is_deeply> honor C<'eq'> overloading (and just that one) and L</cmp_ok> honors whichever overloading category corresponds to the specified operator.
+
+=item *
+
+L</pass>, L</fail>, L</ok>, L</is>, L</isnt>, L</like>, L</unlike> and L</cmp_ok> are all guaranteed to return the truth value of the test.
+
+=item *
+
+L</like> and L</unlike> don't special case regular expressions that are passed as C<'/.../'> strings.
+A string regexp argument is always treated as a the source of the regexp, making C<like $text, $rx> and C<like $text, qr[$rx]> equivalent to each other and to C<cmp_ok $text, '=~', $rx> (and likewise for C<unlike>).
+
+=item *
+
+L</cmp_ok> throws an exception if the given operator isn't a valid Perl binary operator (except C<'='> and variants).
+It also tests in scalar context, so C<'..'> will be treated as the flip-flop operator and not the range operator.
+
+=item *
+
+L</is_deeply> doesn't guard for memory cycles.
+If the two first arguments present parallel memory cycles, the test may result in an infinite loop.
+
+=item *
+
+The tests don't output any kind of default diagnostic in case of failure ; the rationale being that if you have a large number of tests and a lot of them are failing, then you don't want to be flooded by diagnostics.
+
+=item *
+
+C<use_ok>, C<require_ok>, C<can_ok>, C<isa_ok>, C<new_ok>, C<subtest>, C<explain>, C<TODO> blocks and C<todo_skip> are not implemented.
+
+=item *
+
+L<Test::Leaner> depends on L<Scalar::Util>, while L<Test::More> does not.
+
+=back