ParadoxV5

> > `` def self.of: (E) -> instance # RBS::UnknownTypeName (it’s not inferred from `MyWrapper[E]`) `` > > How about letting class methods pick up generics? Would this rather be...

> The type you want would be this? > > ``` > def self.of: [E, T < MyWrapper[E]] (E) -> T > ``` Almost. The strength of `instance` is that...

Some APIs use `nil` as integer infinity, others rëuse `Float::Infinity`. But yeah, `nil` folks would need something else for negative infinity.

How about allowing constants to be used as literals, not just those of `Float`? 🤔

This doesn’t implode: ```ruby def y = yield y { :y } ``` ```rbs class Object def y: () { () -> void } -> void end ```

This doesn’t implode: ```ruby class P def initialize = yield end class C < P def initialize = super() { p self } end ``` ```rbs class P def initialize:...

Some of the commands I’d explicitly write `steep …` rather than `rake steep:…`, such as ``` rake steep:langserver # Run steep langserver rake steep:version # Run steep version rake steep:watch...

> I'll let others chime in here, but I'd just say I agree that for the most part I would continue to just use the `steep` binary directly for CLI...

[Someone in Ruby Discord](https://ptb.discord.com/channels/518658712081268738/1143338778384146574/1167237039134085130) asked if we can write “truthy”, i.e. exact opposite of `false | nil`. Would that in the RBS core be more preferred?

*Comment moved from #256* > * Allows skipping `nil`-able attributes > > ``` > # @type var x: { name: String, email: String? } > x = { name: "foo"...