I like this Idea. The only problem I see is this: if I'm trying to write code that is very generic and abstract, how does the compiler know if the update

> a { fr = 5 }

is targeting a field fr of the record a, or a variable fr, which is in scope and "points to" a first-class field. The difference depends on the record in question, so the code would work differently depending on the context. I would think it would have to be something like 

> a { :fr = 5 } 

or something else syntactically distinct from current record update syntax.

With this and a few more conveniences on record syntax, lenses could go away. For example, I'd love to see a "lambda update" syntax. For example instead of:

> setName n r = r {name = n}

we'd write

> setName n = \{name = n}

I'd also like to see an "Update field by" syntax. Instead of 

> addMr r = r { name = "Mr. " ++ (name r) }

we'd write 

> addMr r = r { name => ("Mr. "++) }

or combining the previous 2:

> addMr = \{name=>("Mr. "++)}

feels very terse and "Haskelly" to me.

Regards,

--J Arthur