I wonder what's the computational power of untyped linear lambda calculus?
Pretty sure it's not Turing complete, I don't think you can write a fixpoint combinator or any other kind of recursor without breaking linearity
@julesh All untyped linear λ-calculus terms can be given a simple type. Just do Hindley-Milner type inference, and nothing can go wrong without multiple occurrences of a variable. (Conor taught me this.)