Brute force approach:
Write all terms in scientific notation, i.e. as $a\cdot 10^k$ for integer $k$ and $1\leq a<10$. Assume for contradiction that only finitely many terms of the sequence begin with $3$. In other words that only finitely many terms have $a\in[3,4)$ in our scientific notation.
That also means that only finitely many terms can have $a\in[6,8)$, as all such terms require the term before them to begin with $3$. Continuing with the same idea, we get that only finitely many terms have $a\in[1.2, 1.6)$. And so on.
It doesn't take too many of these intervals together to cover all of $[1,10)$, meaning there can only be finitely many terms, which is clearly a contradiction. The same argument holds for any other starting digit.