Hasen points to an article by Rick Pildes,and thinks its very important and will be influential. Pildes suggests replacing primaries with an IRV general election.
This is not the worst imaginable system. It lets republicans prefer republicans, democrats prefer democrats, independents to vote for who they like,and the mechanism sorts it out till a winner is a declared. I tried to think about what are the possible flaws. I'm not one of those political scientists expert at applying game theory to election procedures. But here's what I've come up with. Pildes' system will encourage each faction to coalesce around a front runner. Because when one party has a front runner and the other party has a hotly contested race, the front-runner is more likely to pick up enough votes from independents to reach a majority, on some round of balloting. I could be wrong about this, but that seems to be how it would shake out.
That provides an incentive for parties to develop mechanisms to find a front runner, and these mechanisms might be less democratic in process.
Here in Indiana, the parties hold a slating convention, at which candidates can buy onto a slate by donating 10% of a year's salary. If more than one person buys in, the convention decides between them. Voting is done by precinct captains, who are usually controlled by a county chair. The slated candidate usually but not always wins at the primary. Indiana's slating convention is an example of the "smoke-filled room" method. Both times I've run as Republican in Indiana, I was not the slated candidate. In my first race neither I or my opponent was slated. This time the slated candidate ended up not filing and I was unopposed at the primary.
The jaybird primary, a privately held whites-only primary found unconstitutional in one of the White Primary cases from 1940s Texas, is another example of a mechanism to determine a front runner before the official election.
Primaries were one of the great reforms of the progressive era, 1900-1920. The goal was to reduce the power of the bosses and empower the rank and file party member.
Various plans to eliminate or re-tinker the primary system, including Pildes' proposal, blanket primaries or top-two run-off systems, risk undercutting this progressive goal of democratizing the party nomination process.
I agree with Pildes (and others) that IRV is a sensible way to run elections and has cost savings over a run-off method. But I do not agree that his system eliminates the need for primaries altogether. He would need to make a stronger case that the cost savings outweigh the value of the democratic processes lost in the shuffle if we eliminate the primary in exchange for an IRV general election.
Maybe one of the game theory types can comment on my concerns.

what i got done today - not much.
paid credit card bill cashed check
nned to write thanks you card for mom
need to call credit card company.
washed dishes. made stuff from garden, gardened, met w caretaker,
called jasper, wrote kendle, faxed jasper,
harvested parsley seeds.
to do: make comic.
make list.
plan campaign.
write that demfundraiser guy, he's on facebook. met in toranto.
weds - auto parts. downtown gop hq.
call jasper re $1700.
check for studies. make comic.
call arabian
proofread brief.