Eitan Chatav

> Seriously, this `gradientDescent` code was not intended for serious use I didn't realize that :-/ I've been trying to make serious use of it.

@barak : Do you know what the best way to calculate `argmin`s of convex objective functions in Haskell is?

I've found this library [optimization](https://hackage.haskell.org/package/optimization) which seems designed to fit with `ad` and `linear`.

How do you want to capture `happend` for both `NP` and `NS` in a class? ```Haskell append_NP :: NP f xs -> NP f ys -> NP f (xs ++...

Using the identity `Conversion` you can define operators like: ``` func >*< (left: Parser, right: Parser) -> Parser func >|< (left: Parser, right: Parser) -> Parser ```

That's another idea, use `Builder` instead of `ByteString`. Sadly, I don't really know much about performance stuff :-/

Hasql has [some benchmarks](https://github.com/nikita-volkov/hasql/blob/master/profiling/Main.hs) that might be worth looking at.

I dance around my laptop and chant to improve performance. It's more art than science.

Does `postgres-wire` provide a similar interface to `postgresql-libpq`? I'd need things like the `Result` type and `exec`, `execParams`, `execPrepared`, and `prepare` functions. Would you want to see it replace `postgresql-binary`...

> Would the idea behind this be more towards checking schemas or generating them? Both hopefully, I agree it would be awesome to get there.