Ok I got the part about using some repository that (magicallly) 
secures semantics, the more misterous part is what actually 
gives this thing computational capabilities. Does it sit in an
isolated world and operate only on things within the repository
without an ability for a user to reference anything that 
lies outside of the realm of logos?
The above in short: how and what does this thing compute?

Some of examples you gave suggest this to be a kind of a platform 
for interactions where actors seek agreenment in a dialogical 
manner, but what is it that they are bargaining for? 

Also why did you decide to base the interface on natural language 
- the least viable medium for describing algorithmic actions with? 
What does it give you that other approaches - if you have 
considered any - dont?

Hey,

Thanks for asking. I'm not sure if this deserves a dedicated thread, but
no problemo, I'm happy to answer these questions.

>Ok I got the part about using some repository that (magicallly) secures
>semantics, the more misterous part is what actually gives this thing
>computational capabilities.

There is nothing inherently magical about semantics in Logos.

If you were to look at the (heavily outdated) spec which can be found
over at https://aletheia.icu/~badt/logos/spec/ you would see soon enough
that logos is cathartic in a way; at least to people preoccupied with
philosophical questions about the nature of computation, I know it is.

In logos, we primarily talk of sense and nonsense.

Whereas certain utterances constitute meaningful propositons, others—
simply serve as a necessary filler to the logical form that allows it to
stand out among other logical utterances.

There's no 'repository', too.

Discourses in logos are completely modular; you are free to apply them
whenever appropriate with respect to the exact domain you're working
with.

What you get always is a pretty conservative AST, which is what you run
in a virtual machine for it to be computed.

>but what is it that they are bargaining for?

The actors are bargaining for state.

The only way to directly influence others in logos is by means of
interactive bargaining scheme, which maps 100% to what the smart
contract guys are doing in the blockchain space.

In fact, I wonder if logos could be used as a blockchain frontend.

>Also why did you decide to base the interface on natural language
>the least viable medium for describing algorithmic actions with?
>What does it give you that other approaches - if you have
>considered any - dont?

I don't believe in types.

Not that they're useless at what they represent, but rather because I
have a very specific vision of applied computer science and how it's
bound to evolve during the next couple years.

GPT-3 has already prooved that primitive language models can build some
impressive understanding of language provided the corpus contains loads
of data.

>the least viable medium for describing algorithmic actions

I disagree with this violently. Under the hood, logos allows to express
algorithms concisely (in the end of the day) it's just not the preferred
way to do so.

Don't forget that soon enough we'll be able to overfit arbitrary models
given the domain of their application is known and this is a plus, as it
would potentially allow to translate logos more easily.

-badt