Reuben

> Ah, true. I guess it's implemented as `State` for efficiency reasons, but its interface only allows `Writer` operations. That's right. I'm still surprised that there isn't a Writer implementation...

I also had that thought re. SequentialT, and attempted to replace with FreeT and a catamorphism, but discovered that my understanding was not good, because the "folding" in the coroutine...

I recall that (possibly in the paper?), the authors of monad-bayes were aware that `ListT` is not a valid monad transformer, but not that their semantics does not require it...

There is also the Church transformed `FreeT` (https://hackage.haskell.org/package/free-5.1.10/docs/Control-Monad-Trans-Free-Church.html) but that is not making use of the `Applicative` instance. Is the lawlessness an actual problem here? That is, does the package...

I was mostly afraid of meddling with the semantics when I made changes to the library, but there are a couple of things that did change, that are perhaps worth...

Ah, I share your interest in investigating this. I'll have a read of the paper too. I've also thought about doing what I suggest above, but have never thought through...

My preference would be to drop the notebook.

That's fine with me. Especially now that rhine-bayes exists, the streaming stuff is not so important. I think the docs may mention the tui, which uses mcmcP, but other than...

Fine with me, although if it requires more than a little work, it seems to me like not a good use of time, since monad-bayes is unlikely to have many...

That is fair to say, and I don't oppose that.