Fix $n$ and let $P^n$ be projective $n$-space. Let $S = k[x_0, \dots, x_n]$. Set $A^0 = \bigoplus_{d \ge 0} H^0(P^n, \mathcal{O}(d))$ and $A^n = \bigoplus_{d < -n} H^n(P^n, \mathcal{O}(d))$.
I have been told that it is a "well-known" result that $A^0 \oplus A^n$ has an $A_\infty$-algebra structure that extends the usual algebra structure on $A^0$ and so that using the higher multiplication maps the algebra is generated by $H^0(P^n, \mathcal{O}(1))$ and $H^n(P^n, \mathcal{O}(-n-1))$.
Question: Is there a reference or can someone sketch a proof of the statement above?
There is probably a more general statement involving a scheme $X$ and an ample line bundle $L$ on $X$, though I'd already be interested in this special case.
Some thoughts: I have tried to realize this by putting a dga structure on the standard Cech complex (as in the proof of Theorem III.5.1 of Hartshorne's Algebraic Geometry book) and it looks like it works, but there is a problem with grading shift. To be precise, we have a complex $C^\bullet$ so that $C^d = \bigoplus_{I \subset \{0,\dots,n\}} S[x_I^{-1}]$ where $x_I = x_{i_0}\cdots x_{i_d}$ and $I$ ranges over all subsets of size $d+1$. Then there is an obvious multiplication structure on $C^\bullet$ but with a shift: $C^d \otimes C^e \to C^{d+e+1}$.
So it doesn't quite work (and also shifting the indexing does not resolve the problem because then I would be removing the algebra structure on $A^0$) but it seems that this shift by 1 is a common occurrence (like the Whitehead product on the homotopy groups of a space).
