As some know, my project Swarm(.fund) was inspired by the Ethereum whitepaper but ultimately implemented on the Bitcoin metacoin Counterparty. It included crowdfunding, decentralized voting, de applications, and a fund that would take positions in individual projects. Not all of have been incredibly successful so far...
Here are some things that I wish we had done better and that I think the DAO is doing extremely well at:
(1) Highly engaged, highly technical audience that can intelligently debate key issues
(2) Deployed on a stable production environment that has full EVM capability and a good associated developer community
(3) Focused on small projects that will return capital in the near term (instead of high risk deferred return)
(4) Trans-national deployment environment
(5) Transparent allocation of funds
(6) Truly autonomous. No team. Individuals can remove funds if they don't like what is happening.
I think this is an excellent combination in that even if some initial amount of funds does not go into high yield products, the DAO can hopefully iterate on its expenditures and put them into things that have better yield. Also if the first few votes go into questionable allocations money can be withdrawn by other DAO members.
The part I'm most excited about is that it can dynamically evolve based on the input with economic incentives that are aligned with this DAO structure evolving.
Here are some additional key learnings from my own time launching proto-DAOs, including Swarm(.fund) and the many projects that attempted to launch on our Ethereum inspired solution (which was deployed in a Bitcoin metacoin context) :
(1) DAOs have an extremely difficult time entering into any binding contract with any non-DAO entity
(2) DAOs being extra-legal have virtually no way of enforcing any contractual agreements. This means that verbal or other agreements are likely to be ignored.
(3) Loose distributed organizational models have rapid association and rapid disassociation and are typically not suitable for projects which need sustained effort over a long period.
(4) Collaborative decision making methods are highly prone to empower the loudest voices who will often hijack the process and lead to suboptimal outcomes
(5) Collaborative allocation of funds typically disincentives the best people who are capable of investing on their own so you end up with more less capable people
(6) Most p2p businesses that work have a platform as middleman which is able to take a profit. Fully distributed platforms without a middleman often do not have sufficient profit potential.
(7) Highly technical people who have no intention of scamming anyone often have no appreciation of the many details it takes to make a business work (i.e. assessing market size, acquiring of paying customers, etc.), and frequently get excited by unrealistic plans
(8) Regulators typically move slowly but when they do move they use a very heavy hammer. Many projects have been killed in their second or third year by regulators (last night at the Ethereum Silicon Valley meetup we hosted Sand Hill exchange: http://www.ibtimes.com/sand-hill-exchange-regulators-stop-illegal-trades-silicon-valleys-new-wall-street-1971876)
(9) All of these things are new and don't cleanly fit into existing categories so it is very easy for misunderstandings to occur.
I also wrote this article on the "The Competitiveness of Distributed Organizations" (https://docs.google.com/document/d/1AnwwNNMomg7sRoR6d0dQjZwvJ3PfBIxW3Vq9fm6K4I8/edit), which includes some insightful comments by Vitalik at the end.
Key learnings were originally posted on the Ethereum subreddit (https://www.reddit.com/r/ethereum/comments/4kpl08/lessons_for_the_dao_from_the_protodao_world/)
Very happy to be a stakeholder in this project and hope I have something to contribute.
 The DAO