Bill Easterly argues that development is about creating problem-solving systems, and not just individual solutions or interventions for specific problems (e.g. malaria, access to finance, etc.). The whole post is worth reading in full, but here is a snippet:
Development happens thanks to problem-solving systems. To vastly oversimplify for illustrative purposes, the market is a decentralized (private) problem solving system with rich feedback and accountability. Democracy, civil liberties, free speech, protection of rights of dissidents and activists is a decentralized (public) problem solving system with (imperfect) feedback and accountability. Individual liberty in general fosters systems that allow many different individuals to use their particular local knowledge and expertise to attempt many different independent trials at solutions. When you have a large number of independent trials, the probability of solutions goes way up...
...Direct solutions to problems (say, using aid programs) still may be worthwhile as benefiting a lot of people. But a long list of many such solutions is not development; development is the gradual emergence of a problem-solving system.
Easterly makes a great point about the need to create dynamic systems to foster development. But I think he unfairly relegates aid to the narrow category of "direct solution". It is of course intended to be a direct solution, but certain interventions may help create the larger problem-solving system that Easterly espouses. To give one example, David Roodman argues that microfinance (and related initiatives like M-PESA in Kenya) may very well have this kind of transformative impact. Aid interventions ought to be judged not solely by their immediate efficacy but also by their ability to create problem-solving systems.
The messy part, of course, is how measure this kind of impact -- and, as Giulio Quaggiotto has already pointed out on the PSD Blog, development institutions tend not to deal very well with this kind of messiness.