> conditions must be verifiable on the blockchain

And that's where it all falls to bits. Without trusted oracles there is no way to get real world information onto the blockchain.

There are people working on this problem, with some progress and success. Also, reducing the contract problem to the oracle problem is an important advancement, and even if you end up simply trusting oracles, you've made a substantial improvement, imo.

There's a familiar pattern here.

1. Person points to the fact that all existing blockchain implementations are wildly impractical, or logically impossible, or are just more convoluted and confusing versions of the current system with blockchain layered on top.

2. People point out that "there are people working on this problem" and not to worry.

3. A year or two passes.

4. Person points to the fact that all existing blockchain implementations are wildly impractical, or logically impossible, or are just more convoluted and confusing versions of the current system with blockchain layered on top.

Or maybe you're just not very familiar with the state of the art? Token curated registries and Schelling-voting prediction markets can be and are being used as oracles effectively. Oracles are being used all over the place with many different architectures - some centralized, some decentralized. People are iterating on solutions as we speak, and many of them are operational and functioning smoothly as we speak.

When I say these things are 'being worked on', I don't mean they are abstract, purely theoretical ideas being discussed on github threads and dueling whitepapers. Many of them actually exist and are being used to safeguard funds and enforce contracts right now. I say it's being worked on because I don't think we have achieved the final vision of exactly the best ways of doing these things yet, it's an evolving process. Do not mistake that for a lack of progress.