Waffled between putting this here and putting this in the Stupid Questions thread:

Why is the default assumption that a superintelligence of any type will populate its light cone?

I can see why any sort of tiling AI would do this - paperclip maximizers and the like. And for obvious reasons there's an inherent problem with predicting the actions of an alien FAI (friendly relative to alien values).

But it certainly seems to me that a human CEV-equivalent wouldn't necessarily support lightspeed expansion. Certainly, humanity has expanded whenever it has the opportunity - but not at its maximum speed, nor did entire population centers move. The top few percent of adventurous or less-affluent people leave, and that is all.

On top of this, I ... well, I can't say "can't imagine," but I find it unlikely that a CEV would support mass cloning or generation of humans (though if it supports mass uploading, then accelerated living might produce a population boom sufficient to support luminal expansion.) In which case, an FAI that did occupy as much space as possible, as rapidly as possible, would find itself spending resources on planets that wouldn't be used for millenia, when it could instead focus on improving local life.

There is, of course, the intelligence-explosion argument, but I'd think even intelligence would hit diminishing marginal returns eventually.

So to sum up, it seems not unreasonable that certain plausible categories of superintelligences would willingly not expand at near-luminal velocities - in which case there's quite a bit more leeway in the Fermi Paradox.