Overall, I'd say it's a step forward, but not something that quite goes deep enough to make a repeat of the crisis highly unlikely. Forcing lenders to hang on to some of the loans they originate will
force motivate them to do a better job vetting borrowers, which was obviously a problem during the boom; it's good for the government to have "resolution authority" for non-banks so we don't get into a Lehman-style situation where you can't shut down a failing institution in an orderly way; all things equal, it's better if someone (apparently the Fed) is keeping an eye on how all the risks in the financial system interact with one another, rather than just looking at whether individual companies are sufficiently hedged; and what's not to like about a consumer financial products regulator, given that the profitability of a lot of financial products rests on their being completely opaque to the average consumer. I also like the idea of raising capital standards for banks and regulating derivatives by forcing them onto clearinghouses and exchanges, both of which can reduce the amount of leverage in the system and make big flameouts less likely. But there isn't a ton of detail about that stuff yet.
I guess my biggest concern is that systemic risk just isn't nearly as easy to spot as it sounds in these discussions of a systemic risk regulator. We're now aware of the way derivatives contracts can link financial institutions together in a deadly chain, so that previosuly healthy institutions end up teetering when the company on the other end of a contract suddenly melts down. But the problem with financial markets is that the pace of innovation always outstrips our ability to understand the consequences. (Recall that derivatives were initially seen as a way to dampen risk in the financial system.) It's not at all clear to me that the systemic risk regulator will be able to pre-emptively identify future financial products that, when used in certain far-from-initially-obvious ways, lead to greater system-wide risk. Nor is it clear to me that, even if the regulator were able to identify these products and their uses, banks, hedge funds, etc. would listen. The very qualities that make the danger of a new financial product non-obvious (say, it appears to lower risk while doing the opposite) also make it extremely tough to persuade the users of that product to rein it in.
Which is why I'd like to have seen a bit more structural change. To be sure, it's not always easy to say what that structural change should be--when you advocate breaking up institutions so they're not too big too fail, people respond that the problem isn't so much size as interconnectedness*, which brings us back to some of the limitations I mentioned in the previous graf. (These skeptics tend to point to examples like Long-Term Capital Management, which was a small-ish, if highly-leveraged, hedge fund when it imploded in 1998 and nearly brought down the financial system.) But I do think limits on the size of institutions would get you something, assuming the limits were such that behemoths like Citigroup and Bank of America fell on the wrong side of them. If nothing else, smaller financial institutions are generally a bit easier to keep track of, as their balance sheets are generally less convoluted, which isn't nothing if we're trying to spot systemic risks.
*I should point out that clearinghouses can mitigate the interconnectedness problem, since they require you to settle up your derivatives contracts in a central place, rather leaving intact a series of bilateral arrangements, which is where the links in the chain arise.
Update: Long story short, my big concern here is epistemological. I'm skeptical of our ability to anticipate risks and recognize dangerous concentrations of them before it's too late. I feel like the systemic risk regulator requires too much of that kind of knowledge, and that some more structural fixes--breaking up large companies, limiting interconnectedness when possible, certainly limiting leverage (the last of which Obama is proposing in various forms)--are more comforting. That's because, even if they're not perfect, the knowledge requirements are much lower.
Update II: I should add, as Krugman notes, that one of the worthier reform ideas is to extend bank-style regulation to non-bank institutions whose collapse could threaten the financial system. Again, definitely a step in the right direction. My point is just that, in an ideal world, we wouldn't have many institutions--whether they're banks or hedge funds or insurance companies or whatever--whose collapse could threaten the financial system. And in a less ideal world we'd at least have fewer of them.