There are indefinitely many ways we might fit a curve to data, or generalize from finitely many cases to a universal rule that gives the right result in those given cases. For example, my actual usage of the '+' operator to date has not covered every possible case. So it's consistent with my past usage that the rule guiding this usage is not plus at all, but perhaps something like Kripkenstein's quus (which yields different results from plus for certain operations I've not yet performed).
Pettit suggests that meaning is irreducibly social, then, as it is only the stable dispositions of others that provides a check on our own actual conclusions (if I claim 21+22=44, say), establishing the possibility of error and thus meaningfulness. The normative standards we need to hold ourselves to account may be found by triangulating with others. Or so the argument goes.
However, it seems to me that all the work here is being done by the stable dispositions, and not the fact that they're held by other people. After all, if each individual has only a finite usage base to work from, the same is true of our community. We've added a few more points, perhaps, but there are still indefinitely many ways to fit a curve to the data. Nothing significant has been achieved by adding more people into the mix. Indeed, if another disagrees with my application of an operator or concept, we needn't always bother to triangulate at all -- we may simply conclude that we're following different rules, or employing different concepts!
So what really matters here is the stable dispositions (and, indeed, one's own stable dispositions). My use of '+' follows the rule for plus not quus because I have the appropriate dispositions. I am disposed to judge "21+22=44" unacceptable, upon sufficient reflection (and at first glance; but more complex examples may demand more by way of idealization). We can bind ourselves to norms, distinguish our actual vs. ideal judgments, and so other people play no essential role here.