Is there a family $$\{f_\alpha:\alpha<\kappa\}\subseteq 2^\omega$$ of size $\mathfrak c$ or $\omega_1$ such that for any $n\in\omega$, $\{\alpha_i:i<n\}\subseteq \kappa$, and $n\times n$ binary matrix $\textbf M\in 2^{n\times n}$, there exists $k\in\omega$ such that $$(f_{\alpha_i} (k+j))_{ij}=\textbf M.$$
Hopefully I explained it well. It is easy to picture in your mind:
(1) think of the binary sequences ($f_\alpha$'s) going from left to right
(2) pick $n$ of them ($n$ rows) and let a matrix $\textbf M$ be given as above
(3) now imagine an $n\times n$ box sliding over the selected sequences
I want the $f_\alpha$'s to be chosen (beforehand) so that if I slide the box enough, I get $\textbf M$.
If it is too hard to get uncountably many, what about countably many? It seems like you could use some arithmetic to stagger the finite binary sequences or something.
One can at least get countably infinitely many by a straightforward recursive construction. Suppose that for some $n,m_n\in\omega$ we’ve defined $f_k(i)$ for $k<n$ and $i<m_n$ so that:
For $i<n$ let $f_n(i)=1-f_i(i)$. Let $f_n(n)=0$ and $f_n(n+1)=1$. Let $m_n'=\max\{m_n,n+2\}$. If $m_n<n+2$, let $f_k(i)=0$ for $k<n$ and $m_n\le i<n+2$; if $m_n>n+2$, let $f_n(i)=0$ for $n+2\le i<m_n$. We’ve now defined $f_k(i)$ for $k\le n$ and $i<m_n'$.
For $2\le d\le n+1$ let $\mathscr{F}_n(d)$ be the family of finite subsets of $n+1$ that contain $n$ and have cardinality $d$. For each such $d$, each $S\in\mathscr{F}_n(d)$, and each $\mathbf{M}\in{^{d\times d}2}$ in turn extend by $d$ the domains on which the $f_k$ for $k\le n$ are defined so that (2) is satisfied for $S$ and $\mathbf{M}$. This is a finite number of finite extensions; when they have all been performed, there is some $m_{n+1}\in\omega$ such that $f_k(i)$ is defined for all $k\le n$ and all $i<m_{n+1}$ in such a way that (1) and (2) hold with $n$ replaced by $n+1$. Thus, the construction goes through to $\omega$.
Added:
Theorem $3$ of R. Engelking and M. Karłowicz, Some theorems of set theory and their topological consequences, Fund. Math. $57$ ($1965$), $275$-$285$, which I present in slightly different form, allows the construction of a family of $2^\omega$ functions from a countably infinite set to itself that code something similar to what you want without actually giving what you want.
Now let $I={^{<\omega}2}$, the set of finite binary sequences; $|I|=\omega$, so by the theorem there is an independent family $\mathscr{F}=\{f_\xi:\xi<2^\omega\}\subseteq{^II}$ of cardinality $2^\omega$. Let $n\in\Bbb Z^+$, $\{\xi_0,\ldots,\xi_{n-1}\}\subseteq 2^\omega$ with $\xi_0<\ldots<\xi_{n-1}$, and $\mathbf{M}\in{^{n\times n}2}$. Each row of $\mathbf{M}$ is a binary sequence of length $n$ and hence an element of $I$, so let the rows be $i_0,\ldots,i_{n-1}$. Then there is a $j\in I$ such that $f_{\xi_k}(j)=i_k$ for each $k<n$.
Indexing $I$ by $\omega$ allows us to view the functions $f_\xi$ as sequences of finite binary sequences, but as yet I see no way to modify this to get actual binary sequences.
Because it may be of use in thinking about the problem, I include a brief sketch of the proof of the theorem above: