10:28 PM
@MatheinBoulomenos update on my proof
it's essentially different
so my new proof goes like this
Theorem: Let $F$ be a field and $f,g \in F[X]$ such that $g$ is irreducible. Let $K_1$ and $K_2$ be the splitting fields of $f$ and $g$ respectively, and let the roots of $g$ be $b_1, \cdots, b_n$. Further assume that we have a map $\varphi : F(b_1) \to K_1$. Then, $g$ splits in $K_1$.
Proof: consider the following diagram
where the first collection of maps $F \to F(b_j)$ are the usual maps
$K_{1j}$ is a copy of $K_1$
@loch maybe you would like to check my proof
the second collection of maps $F(b_j) \to K_{1j}$ is given by $\varphi \circ \sigma_j$ where $\sigma_j : F(b_j) \to F(b_1)$ is the usual map.
the third collection of maps $F(b_j) \to K_2$ are the usual maps
$B$ is the $F$-tensor product of the copies of $K_1$ quotient by a maximal ideal
$K_{1j} \to B$ is the canonical inclusions followed by a projection
$\Omega$ is $B \otimes_F K_2$ quotient some maximal ideal
Claim: for every $i$ and $j$, the image of the inclusion $K_{1i} \to B$ is a subset of the image of the inclusion $K_{1j} \to B$. Proof of claim: Note that $K_{1i} = F(a_1, \cdots, a_m)$ where $f(a_k) = 0$. Each generator $a_k$ is sent to a root of $f$, i.e. another generator, establishing the claim.
and as a result, $B$ is non-canonically isomorphic to $K_1$
Now, $K_2 = F(b_1, \cdots, b_n)$, and the image of $b_j$ in $\Omega$ is contained in the image of $B$ in $\Omega$, so the image of $K_2$ in $\Omega$ is contained in the image of $B$ in $\Omega$, so we have a map $K_2 \to B \cong K_1$, qed.