The Ethereum project's fundamental goal of providing a distributed computing system, paid for using the cryptocurrency generated by the computations on the platform is a really interesting concept.
Smart contracts, however, have some key limitations, one of which was demonstrated by the attack on the DAO.
I anticipate in the future that we'll end up with smart contracts that are congruent with gnu tools. Where we currently manage a project by holding a hash of the tool available for download on a trusted resource (the organisation's website) - we could then develop these single purpose tools as smart contracts. Organisations themselves could be defined in a smart contract, and that contract would maintain a version history of 'approved smart contracts.'
To this end, you could then 'safely' chain smart contracts together in an ecosystem similar to the original gnu utils.
I'm sure the economics of ethereum don't enable this to be economically viable on the public live network, but it'd be interesting to see how far away the Ethereum project is from this becoming a reality.
Be interesting to know your thoughts.