My feelings are very similar. Ethereum does a ton of things right -- GHOST for fast blocks, Merkle root of balance tree instead of transaction tree, nonce-base

My feelings are very similar. Ethereum does a ton of things right -- GHOST for fast blocks, Merkle root of balance tree instead of transaction tree, nonce-based double-spend protection instead of UTXO gymnastics, simple transaction semantics, fixed reward instead of decreasing reward. But then on top of that there's this Turing-complete language that basically makes every client hugely complex while enabling use cases which are little more than zero-sum games. It's totally possible that Ethereum will be wildly successful and overtake Bitcoin -- the advantages above are significant. But I have to imagine that just as non-standard transactions in Bitcoin never took off, so the contract support in the future will be seen as a slightly embarrassing anti-feature that continues to exist but is increasingly rarely used, like 402 "Payment Required" or any other similar remnants.