\subsection{Finishing the finite ranges}\label{sec:finite} The infinite estimates start at \(10^8\); below it we use the exact membership sieve and complete target scans. For $c\in\{2,4\}$ put \[ H_c=\left\lfloor\frac{10^8-c-1}{6}\right\rfloor. \] Index the possible summands by $6j+1$, $0\le j\le H_c$, and set $\varepsilon_j=1$ unless a prime $p\equiv5\pmod6$, $p\le9999$, divides $6j+1$. For such a divisor set $\varepsilon_j=0$; negative indices also have value zero. This is an exact membership test for $\AA$. Each indexed integer is odd and prime to $3$. If it contains an inert prime factor, the total number of inert factors counted with multiplicity is even, since the integer is $1$ modulo $3$. At least one of these factors is at most the square root, which is less than $10000$. For a prime $p\equiv5\pmod6$, precisely the indices $(5p-1)/6+rp$, $r\ge0$, are removed. \begin{proposition}\label{prop:finite-certificates} For every target $n=6k+2<10^8$, there is an $i\le35$ with $2i\le k$ and $\varepsilon_i\varepsilon_{k-i}=1$. For every target $n=6k+4<10^8$, there is a $j\le33$ with $3j\le k$ and $\varepsilon_j\varepsilon_{k-3j}=1$. \end{proposition} \begin{proof}[Computer-assisted proof] The standard-library program described in Appendix~\ref{app:verifier} constructs the exact membership sieve, checks the small candidates by direct divisor tests, and requires a witness at every target index. Its two complete scans give the following results: \begin{center} \begin{tabular}{lrrr} \toprule family & targets & largest target & small summand bound\\\midrule $\AA+\AA$ & $16{,}666{,}667$ & $99{,}999{,}998$ & $6i+1\le211$\\ $\AA+3\!\cdot\!\AA$ & $16{,}666{,}666$ & $99{,}999{,}994$ & $6j+1\le199$\\ \bottomrule \end{tabular} \end{center} Neither scan has an uncovered target. The witnesses give, respectively, \[ 6k+2=(6i+1)+(6(k-i)+1),\qquad 6k+4=(6(k-3j)+1)+3(6j+1). \] All terms are positive members of the required sets. Separate NumPy implementations reproduce every membership flag and witness; their normalized digests agree with the standalone verifier. The integer arithmetic and receipt conventions are specified in Appendix~\ref{app:verifier} and the supplement. \end{proof} Proposition~\ref{prop:finite-certificates}, the infinite-range argument in Section~\ref{sec:infinite}, and the initial multiplicative reduction prove Theorem~\ref{thm:main}. The finite witness menu cannot work for every target: \begin{proposition}\label{prop:no-finite-menu} For either $c=1$ or $c=3$, and every finite $F\subset\AA$, there are infinitely many positive $n\equiv c+1\pmod6$ with $n\notin\AA+(c\!\cdot\! F)$. \end{proposition} \begin{proof} The case $F=\varnothing$ is immediate. Otherwise choose distinct odd inert primes $q_f$, one for each $f\in F$. There are arbitrarily many such primes: for any finite collection including $2$, the integer three times their product minus one has an inert prime factor outside the collection. The Chinese remainder theorem gives an arithmetic progression of solutions to \[ n\equiv c+1\pmod6,\qquad n\equiv cf\pmod{q_f}\quad(f\in F). \] Choose its positive terms larger than $c\max F$. Then $n-cf>0$ is divisible by the forbidden prime $q_f$, so it is not in $\AA$, for every $f\in F$. \end{proof} Thus the finite witness bounds $211$ and $199$ cannot be replaced by fixed global bounds of the same kind.