static typing doesn't appear to be an effective tool for encoding intent

Static typing is about as effective at encoding intent as intention revealing names. That is to say, if you are diligent and clever in how you structure and name things, a system can be very intention revealing.

HN user jerf has a method of revealing intent through types in golang, which he calls something like microtyping. Basically, everything has its own type, based on its domain semantics. So if you're using a float64 to store time deltas, you don't just use the float64. Instead, you base a TimeDelta type on float64.

    type TimeDelta float64

Then golang allows you to attach methods to that type, which can then provide more opportunities for intention revealing. Also, this prevents a programmer from mistakenly using a float64 representing width and shoving it into a time calculation. Is this as ironclad as Eiffel Design by Contract or Haskell typing? No. However, I've been using it myself, and I can affirm that yes, the compiler will warn you about some semantic errors if you do it this way.

The big problem with the Intention Revealing Names pattern, is that there is no enforcement and that it is subjective. For this reason, it's likely to fall apart in a large enough project that lives long enough.