Haskell — Lazy, Pure, and Influential Far Beyond Its Adoption
Headline
A committee formed at FPCA ‘87 in Portland decided the fragmented world of non-strict pure FP needed a shared language (Wikipedia: Haskell). Haskell 1.0 shipped on April 1, 1990 — “a date had to be chosen” (Peyton Jones, A History of Haskell: Being Lazy With Class). Named after logician Haskell Curry. Not a joke, despite the release date.
The three ideas that shaped everything
- Purity. No side effects in functions. Every effect — I/O, state, exception — lives inside a monadic type. This makes equational reasoning sound.
- Type classes. Wadler and Blott’s principled ad-hoc polymorphism, later everywhere: Rust
trait, Swiftprotocol, Scalagiven, Go’s compromise interfaces. - Laziness as default. Every value is a thunk until forced. This was the unifying commitment of the committee; it also produces the space-leak class of bugs that haunts every large Haskell system.
What Mo takes
-
Purity as a type-level property. Not the whole language, but the effect region around pure code — d15-effects-via-capabilities is Haskell’s monadic segregation done via capabilities and handlers instead of return types.
-
Type classes / traits. Some form of principled ad-hoc polymorphism is likely; Rust-style traits are the direct descendant Mo prefers.
17 Sep 2026: built —
traitandimplare in the language (pick 15). -
The reasoning discipline. Reason about programs as if they were mathematical values. This is exactly what d03-source-carries-its-evidence asks agents to do.
What Mo refuses
- Laziness by default. Space leaks, opaque runtime behavior, no predictable performance — all disqualifying for a language agents ship into production. Strict evaluation (d10-immutable-by-default; link corrected 17 Sep 2026).
IOas a monad. Monadic effect stacks are elegant but hard to compose in agent-written code. Algebraic effect handlers with capabilities are the successor (koka).- The committee culture of experimentation. Haskell’s motto — “avoid success at all costs” — is beautiful and unshippable. Mo optimizes for adoption.
The lasting lesson
Haskell is the language whose ideas everyone else adopted. Type classes, Hindley-Milner extensions, GADTs, monad transformers, lens libraries, software transactional memory, QuickCheck property testing — all Haskell exports. But Haskell itself remains niche because the substrate (laziness, purity as language default) is too demanding.
Mo’s read: steal the ideas; steal them onto a strict, capability-flavored substrate that ships.
Two Peyton Jones quotes worth internalizing
“Haskell’s designers were united most strongly by laziness.” — the whole language is a bet on one axis.
“There will always be a suspicion that if Haskell had made more of an effort to be practical it would have been less influential.” — the trade Mo refuses to make.
Sources
Related
- ml — the strict-eager branch of the same family
- koka — algebraic effects on an ML/Haskell substrate
- elixir — pattern matching and immutability without the monads
- d15-effects-via-capabilities
- d10-immutable-by-default