Matthew Yglesias writes:

[W]e often seem to be assuming that the equilibrium state in Afghanistan is a stable, centralized government. Thus either Karzai and the United States will win by creating a stable, centralized, Karzai-run state or else the Taliban will win by creating a stable, centralized, Taliban-dominated state. The evidence, however, suggests that the equilibrium state is for Afghanistan to be run by rival gangs of warlords configured in a complicated panorama of shifting alliances. Any “winning” coalition is likely to collapse under the weight of victory.

To put this in a different context: the administration is considering whether a victorious Taliban would invite al-Qaeda back into Afghanistan. (Leave aside for a moment the question of whether al-Qaeda would accept any such offer.) But keeping with Yglesias’s framework, there’s little reason to believe "victory" for the Taliban is more of a qualitative shift than a quantitative increase. To quote myself:

 …Taliban elements have free reign in many areas of southern and eastern Afghanistan, yet the administration is contending that there are only an estimated 100 al-Qaeda operatives in Afghanistan.

Now, even a Taliban fulsomely backed by Pakistani intelligence couldn’t take control of the entire country, so it’s unlikely that any Taliban “victory” scenario is going to end up with the Taliban and its affiliates running all of Afghanistan, and accordingly the administration will have to ask whether quantitative gains in Taliban control would lead either al-Qaeda or the Taliban to change the current dynamic. What’s more, it has to ask whether, say, the upcoming Pakistani military offensive in Waziristan would make al-Qaeda’s senior leadership look for safer haven.