My take from the presentation is that the function of coin laundering needs to be incorporated into the client itself.
His example blogger with a single published donation address should be giving out new addresses for each donor if he cares about income/spending anonymity.
His client software should be monitoring his spending habits and intelligently managing his pool of addresses to ensure that any time he spends Bitcoins at least one address has enough to initiate a single sender transaction.
In addition the heuristics should have a lot of tunable parameters so that the software can automatically adjust its behavior to make wallet reshuffling look statistically similar to the other transactions in the block chain.
All this can be done by the client without the user needing to worry about the details. The user would only need to make a few high-level decisions like how much he was willing to pay in fees if it was the case wallet reshuffling incurred them and what to do in the case of a large unexpected purchase in which no single address had enough Bitcoins to handle it alone (wait for a certain amount of time for reshuffling or else accept a permanent reduction in anonymity). Users who where especially interested in anonymity could take the time to tell the client ahead of time about planned purchases to give it time to make those funds available via an appropriate single address.