I think this is worth reiterating:

For example, Clojure’s cond works in similar fashion, although the convention there is to use :else instead of true, e.g.,:

(let [x 11]
  (cond
    (< x 2)  "x is less than 2"
    (< x 10) "x is less than 10"
    :else  "x is greater than or equal to 10"))

And :else also works in Elixir:

iex(6)> x = 11
11
iex(7)> cond do
...(7)>   x < 2  -> "x is less than 2"
...(7)>   x < 10 -> "x is less than 10"
...(7)>   :else  -> "x is greater than or equal to 10"
...(7)> end
"x is greater than or equal to 10"

true, :else, "default", etc. are all truthy and therefore serve the purpose.

Personally, I kinda like :else -> or :default -> but the Elixir community appears to have settled on true -> so I’ll use that.

I understand that you are only referring to a mental model - but as you know in Elixir a literal if is actually implemented a case expression, while cond is it’s own special form (so if taken literally this could be confusing).

Essentially the first expression that evaluates to a truthy value “wins” and consequently the expression to the right of that -> is evaluated and becomes the value of the cond expression. In a sense the cond expression is enclosing a sequence of condition -> expression constructs.

Why can’t we use _?

cond by design

Raises an error if all conditions evaluate to nil or false .

so

it may be necessary to add a final always-truthy condition (anything non- false and non- nil )

_ would imply that any value is OK.

In the end it makes the design of cond simpler
(in the true sense of the word - if ... else if ... else .. end is more familiar).

true -> expression

is no different from the other

condition -> expression

before it - i.e. there is no need for a special else part inside cond.

Beacause _ is only valid in a match and cond is only expanding to a bunch of nested ifs.