I came across the following proof as part of studying introductory linear algebra.
Let $$\vec u = \begin{bmatrix}1 \\ 0 \\ 0 \end{bmatrix}, \vec v = \begin{bmatrix} 0 \\ 1 \\ 0 \end{bmatrix}, \vec w = \begin{bmatrix}1 \\ 1 \\ 0 \end{bmatrix}. $$ Prove $\text{span} \{ \vec u, \vec v \} = \text{span} \{ \vec u, \vec v, \vec w \}. $
Let $\vec r \in \text{span} \{ \vec u, \vec v, \vec w \}. $ By definition, $$\vec r = \alpha \vec u + \beta \vec v + \gamma\vec w$$ for some $\alpha, \beta, \gamma\in \mathbb{R}. $ Since $\vec w = \vec u + \vec v$, we see that $$\vec r = \alpha \vec u + \beta \vec v + \gamma(\vec u + \vec v) = (\alpha + \gamma) \vec u + ( \beta + \gamma) \vec v \in \text{span} \{ \vec u, \vec v \}. $$ Hence, $\text{span} \{ \vec u, \vec v, \vec w \} \subseteq \text{span} \{ \vec u, \vec v \}. $
Conversely, let $\vec s \in \text{span} \{ \vec u, \vec v \}, $ so that, by definition, $$ \vec s = a \vec u + b \vec v = a \vec u + b \vec v + 0 \vec w $$ for some $a, b \in \mathbb{R}$. This means $\boldsymbol{ \vec s \in \text{span} \{ \vec u, \vec v, \vec w \} }, $ and therefore $ \boldsymbol { \text{span} \{ \vec u, \vec v \} \subseteq \text{span} \{ \vec u, \vec v, \vec w \} } $.
We conclude that $\text{span} \{ \vec v, \vec u\} = \text{span} \{ \vec u, \vec v, \vec w \} $.
My problem is with the part of the proof that I've emphasized in bold text.
The definition of span of a set of a vectors is that it's the set of all possible linear combinations of those vectors.
But $\vec s = a \vec u + b \vec v + 0 \vec w$ does not represent all specific linear combination of those three vectors, because we are multiplying $\vec w$ by the constant $0$. So how does it follow that $\vec s \in \text{span } \{\vec u, \vec v, \vec w \}$?
(An idea I had while writing this question is that maybe it is because $\vec w = \vec v + \vec u$, and we can always find some linear combination of $\vec v, \vec u$ that gives $0$.)
Notice that $$s=a\vec u+b\vec v=(a-k)\vec u +(b-k)\vec v+k\vec w\in\operatorname{span}(\vec u,\vec v,\vec w)$$
But really whether you chose $k=0$ or $k=1$ or even $k=\frac{1372\ln(\pi)}{\sqrt[3]{\sin(666)}}$ does not matter, all are linear combinations of the three vectors.
In the same way, $\vec w=0\vec u+0\vec v+1\vec w\in\operatorname{span}(\vec u,\vec v,\vec w)$, and you are not bothered with the two coefficients equal to $0$.
Remember that you need to prove $\operatorname{span}(\vec u,\vec v)\subset\operatorname{span}(\vec u,\vec v,\vec w)$, which means for any linear combination of $\vec u,\vec v$ there exists at least one linear combination of $\vec u,\vec v,\vec w$ which is the same vector.
So LHS you have "any" and RHS you have "at least one".