V is smart but is not a serious coder.
This is a correct assessment. There are a legion of factors that go into a talented software developer. Some come from experience, others from knowledge of computer science and some are personality characteristics like discipline and curiosity. Vitalik is like the Poincaré of our field. He's brilliant and capable of doing anything, but rarely focuses for a long period of time on a single problem like Andrew Wiles did. The end result is a thousand starting points that others pick up and finish.
If the environment is well structured, then this can be a tremendous asset. If not, then it can be a curse. Really there needs to be serious, highly focused people working on picking up the threads and finishing them. With their level of capitalization, I suspect this will be done.
In respect to the yellow paper, it was surprising to me that Gavin didn't include Vitalik as a Co-Author. And frankly he had an opportunity to work with others like Neal Koblitz to enhance it. It's a rather moot point as the proof of concepts have drifted from the yellow paper spec and the ecosystem is now no longer just about some scripting language on a blockchain, but rather a combination of a completely new browser, a secure, distributed communication system called whisper and finally a file sharing system called swarm. Why this isn't being done with Bittorrent, WebRTC and Open Peer is absolutely beyond me and also why it requires entirely new programming languages like Solidity is also strange.
Furthermore, there seems to be limited focus on certain things that absolutely have to be done correctly. For example, Ethereum needs a proper consensus algorithm and has the money, talent and time to develop one, yet this seems to be one of the least areas of public project focus. My idea was to develop it like AES was via an incentivized contest (BTW DPOS could have entered). Yet now it seems to be an in house affair. The problem is that distributive consensus over an untrusted public channel is one of the hardest problems in computer science, which is why Leslie Lamport won a turing prize.
Another area is interoperability with legacy ecosystems. The Google Dart team is a great example of the enormous pain one must endure to attempt to move an ecosystem into a better equilibrium. Everyone agrees that maintaining javascript code, frameworks, testing and DOM issues are a damn mess. There are so many partial fixes from Angular to Meteor and backbone. Yet when Google creates a beautiful new language, a wonderful GUI library (Polymer with material design) and it compiles into fast JS with polyfills for everything, it's treated like a redheaded stepchild with a body odor problem. Thus even when you have something awesome, developer and consumer migration to it is slow and painful.
I would have loved to have the entire software stack just be a chrome app that would be a single click install like CryptoCat and gradually move consumers to a custom framework built on top of CurveMQ and utilizing the work of Breach.cc; however, they are starting with the hardest sell first without a single viral app to push it, which brings me to the last point. Developer experience.
There isn't an IDE specifically for DAPPs. There aren't specialized libraries and reusable modules like the node packages. There aren't tutorials explaining how to build a 101 pedagogical application or smart contract best practices. It's young days. Whoever works hardest here will likely be the dominant platform for DAPP development. Currently the 1.0 space is gaining better traction than the 2.0 due to ventures like GEM and Chain. Honestly, build a browser based chrome APP and connect it to a blockchain style GIT. This could be done in weeks to beta and months to 1.0. It could have code academy like tutorials built in and would focus the dev community towards whatever APIs the App's maintainers thought important. It's always about beautiful, rich experiences.