[kazinator]

He's designing an even better Lisp dialect in which lambda is spelled f (down a full 50% from fn!) and let bindings do not require any parentheses at all:

    (let a 1 b 2 c 3 (+ a b c)) -> 6.

Note how there can only be one form, otherwise it is ambiguous. But when you have two or more forms in a construct ("implicit progn"), all but the last are evaluated purely for their side effects (pun intended): it is imperative programming, which this new form of let nicely banishes.

[chubot]

Is this a joke or is there some update/successor to Arc?

[ChristianBundy]

This is a joke.

[lisper]

> Note how there can only be one form, otherwise it is ambiguous

Actually, that's not true. You can disambiguate multiple forms by noting that variables are atoms while forms are lists.

[kazinator]

Here is a problem: symbol macros. Symbol macro foo could expand to a form like (bar xyzzy) let cannot know whether foo is intended as a form (via expansion of to (bar xyzzy)) or whether it's intended as a new lexical variable shadowing the symbol macro foo. It doesn't matter whether let is a special form or macro: something at some point has to walk the code of let and make this decision, whether that is the expander for a let macro, or the code walker that traverses special forms in search of macros.

[lisper]

If you want foo to be interpreted as a form just wrap it in a progn.

(Personally, I think it's very bad practice to use a symbol macro that expands to anything other than a place, and it's certainly bad practice to define a symbol macro that expands to a form with side-effects. So this would never happen in my code. YMMV of course.)