Mistral-large is probably the best large model for practical purposes at this point.
What makes you say that? I have not performed my own comparison, but everything I have seen and read suggests that GPT4 is king, currently.
Yes, I don't know how I forgot to mention that Iceshrimp and Sharkey both have Mastodon compatible APIs - so all the same apps work (mostly).
Based on your requirements, I would suggest looking at one of the Firefish / CalcKey forks. They are ideal for single user or small instances and they support s3 compatible object storage out of the box.
I would recommend looking at Sharkey or Iceshrimp. Both are under very active development and have very responsive developers if you need support.
If you would like to check out an example, Ruud (of mastodon.world and lemmy.world) set up an instance of Sharkey at (you guessed it) sharkey.world.
Another vote here for Fastmail. I also like Posteo, Mailbox and mxroute, but these are not as fully featured - which may be perfect for you if you're after email only. What I really like about Fastmail is that on top of being a customer-focused business (rather than a customer is the product business), they offer a really snappy web interface with excellent search - and they are extremely compliant with email standards, building everything on JMAP.
I do not like Proton or Tutanota. I have used both, including using Proton as my main email account for the past two years. I do believe they are probably the best when it comes to encryption and privacy standards, but for me it's at far too much cost. Encrypted email is almost pointless - the moment you email someone who isn't using a Proton (or PGP encryption), then the encryption is lost. Or even if they just forward an email to someone outside your chain. I would argue that if you need to send a message to someone with enough sensitivity to require this level of encryption, email is the wrong choice of protocol.
For all that Proton offer, it results in broken email standard compliance, awful search capability and reliance on bridge software or being limited to their WebUI and apps. And it's a shame, because I really like the company and their mission.
Honestly, for any large scale project in Python, Pydantic makes it bearable. We use Python heavily at work (and I'd argue we shouldn't be for the projects we're working on...), and Pydantic is the one library we're using that I wouldn't be without. Precisely because it allows us to inject some of these static typing concepts and keeps us honest, and our code understandable.
Yes! The concepts are intertwined.
I think the key take away, for me, is to lean heavily into your type system and allow that to do some of the heavy lifting. Accept that something like a username is not a string, but a subtype of a string (this has to be true if any validation is required, otherwise you'd just accept any valid string).
It's one of my favourites. Something I revisit every couple of years.
Goodness, what a choice to make. They are both excellent, and you should of course read both. Personally, I would start with Hyperion.
A seemingly unpopular opinion, but Christian Bale's Batman is my favourite live action version of the character.
Celebrities, politicians and businesses will be more likely to show up on the platform, if that's your jam.
When corporations inevitably arrive to the platform, we can use it to shame them into offering a decent service after they ignore our calls and emails.
Interesting. That's not something I've heard about until now, but something I'll surely look into.