Finally No Database, No Problem.
So, I tried to come up with some eloquent post introducing this AWS-GCP-esq alternative, but I give up lol. Just check this out:
TLDR; They're building an all-in-one compute platform. Applications are deployed in "canisters", which is analogues to a container.
Canisters run an "actor", which is analogues to a program. Canisters can communicate with one another by emitting messages - which is exposed to the dev as just calling a canister function (so you get PubSub out of the box even).
One insanely cool part is your applications memory page is persisted - Meaning No databases! Just stash your data in the best data model. You can even update your code, while maintaining your memory page.
Additionally, canisters can be loaded up with static assets, meaning you can deploy a canister to host your frontend - which then can just out of the box talk to your backend canister.
Right now, the project is still pretty early on in development, but Its 100% useable and the developer experience is really good.
So, there are a bunch of drawbacks as well. Get calls are silly fast, but update calls are a bit slow since its essentially reaching consensus on a gossip network. You can't make external API calls since your application needs to be deterministic, but you could use an oracle system publishing data into the canister. Same thing with randomness, however your code is private, so you can just use a DRNG algo with a large seed keyset (which allows updating).
Definitely check it out!
Are you still playing with this or have you moved onto something else?