The modelling of value is a prerequisite to the most interesting use cases.

Governance (DAO's, Ardent United, Aragon, etc.)

That's "interesting" as opposed to "useful", because The DAO is most famous for being hacked.

Logistics and proving provenance. Proof of location.

Prediction Markets

Global computer and storage system

Git?

I didn’t realize git was Turing complete

"Turing complete" wasn't part of GP's suggestion; my reply therefore still applies. On top of that, Ethereum's "turing complete" relies on an unsafe, unproven and likely broken programming language.

And I didn't realise the secrets I gave to my notary were out in the open. Hint: they're not, and that's legally enforceable. The blockchain is nothing because data is [legally] [1] nothing. You might as well just use public key cryptography and release your data over BitTorrent, like Wikileaks did. It'll get distributed based on supply & demand, and using BitTorrent is still cheaper & efficient than blockchain.

[1] At least, in the EU.

With the wonders of cryptography, you don't necessarily need to put secrets out in the open to notarize something on the blockchain. You can quite easily use a one way hash on the secret document, allowing you to verify the authenticity and the timestamp, but preventing someone from reversing the operation and seeing your secret information.

Turing completeness was sort of assumed with "Global computer". Neither myself nor the poster mentioned Ethereum at all. He just mentioned "global computer" as one application of blockchain that isn't a cryptocurrency. I could have just as easily been talking about NEO, which uses a number of languages, including C# and Java. Further, while Ethereum does frequently use a new language (Solidity), it doesn't rely on it. Ethereum is a virtual machine, and the first implementation of the virtual machine was Solidity. Others have been made as well, its turing complete after all.

A notary

Opentimestamps

Supply chain provenance