Could you explain how the cosmos/tendermint is different from other interchain projects such as…

This is a big question.
Ark is similar in many things, no wonder it also mentions Tendermint in its whitepaper. If you read the whitepaper you’ll see that Ark wants to do more than Cosmos wants to do. There is also IPFS to be included, Smart Contracts like in Ethereum and the Stuff that Cosmos has as well. The most important bit might be SmartBridges, in Cosmos it is called IBC and it only allows for token transfer between blockchaenins. In Ark you can listen to trigger functions, it is not very concise in the whitepaper, but it sounds like you can send any information between blockchains. You did not mention Polkadot, but this is the big difference between Polkadot and Cosmos. But let’s stick to Ark, it wants to do a lot, the whitepaper makes many references, but is not as clearly written and thought through as it is in the Cosmos whitepaper. Finally this is reflected in the progressing speed of the whole project. There is much left to be done in Ark and whenever I check back after some month, I have the impression not much has changed. Also there is another good reason why Cosmos does not try to be a virtual machine based blockchain, Ark tries to, but this is a long topic and not easily explained. Maybe in the future I will write an article about that.

Let’s look at the others you mentioned: AION, ICON, Wanchain. All of their ICOs came after the Cosmos ICO. Ark is obviously influenced by Tendermint / practical byzantine fault tolerance, these three are influenced by Cosmos additionally. Still, there are notable differences. AION has Proof-of-Intelligence, which we don’t like because we are real idiots and for us there is nothing to proof. PoI is actually PoW but instead of classical cryptographic hashes there are neural nets to be trained. I think this is not a good idea. I have worked with neural nets in the last years and maybe someday I will wake up and think “WAIT! this is a great idea”, but right now I’d say, different work same shit. Also I don’t understand why predefined neural nets are used. There is not much to be learned there. If it was for finding different topologies, maybe using genetic algorithms, stuff like hyperNEAT, I’d say ok, this work might discover efficient neural networks. But still in such a case, I’d say it is a waste of time. Neural networks are in its infancy still. The nets of today are outdated in 3 years. It’s cool to optimize these nets, do speech recognition and apply it to self-driving cars. But it is very hard to make Proof-of-Work bring useful progress into this field. This research must be supervised and guided, but this is not possible with the constraints of a blockchain. AION is also a blockchain based virtual machine, I think this not as effective as Cosmos’ approach, but it is not that bad. The whitepaper is clearly written and the team looks good. I don’t understand why AION does not have a higher valuation, because it is actually the blockchain starter pack:

ICON has a weak whitepaper in my opinion. So much text about what they want to do, what products will/can be built and many passages of “we have this loopchain and it will solve everything”, “and then we connect this and that”, it reads more like a business plan. All the important bits are mentioned, but not connected in a sensible way. This is further reinforced by the fact that they compare themselves to Ethereum, EOS and Bancor, as well as their direct aim to build a DEX. Don’t get me wrong, these are great projects, but not the right ones to compare to, if you want to “hyperconnect the world”. Also I don’t understand why the fee is fixed to 0.01 ICX, I think it should adjust dynamically, but even if there is a good reason, why isn’t it written in the whitepaper? It just says that’s the fee to prevent DDoS, ok great, but why?

Wanchain in contrast has a clearly written whitepaper, that gets satisfiably deep into technicals. It is not super innovative, I’d consider it more like an Ethereum++, but it looks solid. Also it references to other papers in scientific style, of the competitors mentioned here, I like this the most. Nonetheless it is again a blockchain based virtual machine and I think this is not the best way for interoperability.

To sum it all up:
The common difference is Cosmos is not a virtual machine, the others are or let’s say will be virtual machines, once they go live / move from ERC20 to mainnet. To give at least a short introduction, what this difference means, look at it this way:
A blockchain based virtual machine is a computer program running on a world computer. All nodes agree on the same computation. In contrast Cosmos only agrees on a state. Therefore you don’t have to synchronize and agree on the specific operations happening in the virtual machine, you just have to agree on the result of your computation. That is the reason why the Cosmos/Tendermint combo is compatible with all programming languages and Ethereum and all copy cats of Ethereum have their own programming language or at least a deterministic variant of another programming language, for example Haskell. Therefore these blockchains are turing complete and Cosmos is not. For ethereum it is not a problem, if someone implements an infinite loop into a smart contract. At one point gas will be empty and the loop will stop. If the Cosmos Team implements an infinite loop (accidently) then Cosmos stops to work. Wait, cannot anybody implement a smart contract? No. In Cosmos you don’t build smart contracts, you build application specific blockchains, so if your blockchain is halted because your nodes are stuck in endless loops that’s your problem. 
That’s the view of a techie. The investor’s view is very easy:
The Cosmos team does not hype and shill the fuck out of their product. The other projects reference to Tendermint and Cosmos. Easy decision where to invest.