Ok. I solved it but the second lemma is a bit ugly. maybe someone sees an improvement so we can improve that

thing
lemma1
there exists a quadratic non-residue

modulo

which is less or equal to
Proof:
let

be the smallest quadratic non-residue mod

. Now define

. Than

meaning

(obviouslly by substituting

we easily conclude

). therefore

must be a quadratic residue. we have (these are legendre symbols)
and since

is a non-residue also

must be.
but than because

is the smallest

which leads to
lemma2
there exists a quadratic non-residue

mod

(

) such that
Proof:
consider

(the smallest non-residue mod

).

by first lemma (obviously

for

)
if

than

. But if that is not the case than we consider numbers

. these numbers are all quadratic non residues mod

because legendre's symbol is multiplicative. and more. one of these numbers is between

and

. that is because

for suficiently large

and it's therefore imposible for the sequence

to miss this interval.
the inequality is equivalent to

. now we substitute

(where

) and we get

which is true for all

grater than

mening
now for the problem
assume the statement is false and

. therefore there exist

consecutive numbers mod

that are all quadratic residues (one of them may be zero). now consider the set

where

is such a non-residue as in lemma2.
because the legendre's symbol is multiplicative this now is a set of quadratic non-residues(one of them can be zero) that are at most

appart one from another.(if one is zero). we have

by definition of

. therefore it is imposible to pick

consecutive numbers such that none of them would be a non-residue starting between

and

.
however this interval has lenght of

which is grater or equal than

due to definition of

and lemma2. we have therefore covered the whole modul with non-residues that are at most

appart. it is therefore impossible to choose

consecutive numbers which do not contain a non-residue. Q.E.D
all that remains is to check all prime numbers up to

. which is quite a suferable task. i didn't check them myself but i trust the problem holds and hope someone will now come up with a lower bound than
