True that if it is open source then it would be distributed. However, it is not autonomous or perpetually tangible, which I think is very important. The only way to obtain these valuable features is to create a replicating state machine for these items recorded by Steem, but that means the use of Steem becomes unnecessary.
The value of on-chain execution is that your asset is preserved by a replicating state machine and thus perpetually tangible. Ethereum caused the intangible to become tangible. Soft consensus does not do this, but rather it only serves as a back up receipt of authenticity.
I suppose soft consensus is good if you want everyone running their own servers, or design a system that rewards people for running servers, but then the cost would go up and it would be no different from on-chain costs. In this way, I could see something being built on top of IPFS. However, I still argue that smart contracts are better, because you don't have to run your own server.
Thanks for sharing your comment. :)
I think the main difference between the two models is scalability. Smart contracts very difficultly scale well. If you look at the most scalable proposals of smart contracts at the moment all of them shard the public network and execute the state machines on a reduced subset which in the end also results in reduced decentralization.
So, either you have a model which doesn't scale at all (Ethereum) or you got a model which scales decently (Algorand) but is less decentralized and easier to attack, or you got a model which scales very well (Steem - "soft consensus") but you need to run it yourself and people have to verify things externally before trusting the values.
I did some comparisons in terms of price and ease in my last paper, and executing things on Steem via soft consensus was ridiculously cheap even if you add the server renting costs together.