Known limitations

0.9 is a stable core, not the whole language. Notably:

• No generics. Overloading (ad-hoc, exact-type dispatch) is the only polymorphism; there are no type variables — which is why the matchers are compiler-provided rather than written in .qn. The module system is minimal (core.io/core.test built-ins + file-path imports).
• Closures are monomorphic. Lexical capture works end-to-end (= by value / := by reference; see Closures), including recursion of non-capturing nested functions, capture across nesting levels, and capturing-then-calling another closure. A closure can also be passed to a function-typed parameter and called there. Deferred (each needs the closure’s type threaded through inference): capturing a polymorphic value, generic closures, and returning a closure from a function.
• Overloaded and top-level function names are not first-class values. A closure is passed as a lambda literal or a named closure binding; passing a top-level function or an overloaded name as a value is not yet supported.
• Sum-type payloads mixing types across variants aren’t unified yet. Distinct payload types per slot across variants (a position that is Num in one variant and Text in another) is deferred; the payload set (Num/Text/Bool/$ and a named record, consistent per position) works.
• A named-composite sum payload must be a record, and a record field cannot yet be a named composite. A variant may carry a named record (Post(Body)), but not another named sum; and a record field is still limited to built-in types and arrays (a { inner :: Inner } field of a user type is a deferred follow-up).
• Concurrency is partly built. The model is locked; the fiber scheduler, reactor, @sleep, and the deferred-value primitives (@readStdin, @tcpRequest) run. Remaining for 1.0: overlap as a showcase, deferred composites, further @ primitives (file), and multicore M:N.