design/ + source rustdoc: replace opaque shorthand with content phrases + links

Reader-facing prose and rustdoc carried opaque shorthand like
"Decision 10", "clause-5", "mq.1", "ct.1", "eob.1", "rpe.1",
"post-mq.3", and "Iter 22b.1:" with no in-repo definition the reader
could follow. This commit replaces every such occurrence in the
durable tier the reader is most likely to land on (design/ ledger +
source //! module headers + the central /// public-item rustdoc) with
an inline content phrase plus, where applicable, a Markdown link to
the file that defines the referenced concept.

design/ ledger — 16 files:
  Definition-site headings demoted from "Decision N: <title>" to
  "<title>": authoring-surface, tail-calls, memory-model section in
  rc-uniqueness.md, dual-allocator section, typeclass design,
  effects "pure core + algebraic effects".
  Cross-reference sites: "Decision 1" -> canonical-schema principle
  (data-model); "Decision 3/4" -> effects + scope-boundaries; "Decision
  6" -> authoring-surface; "Decision 8" -> tail-calls; "Decision 9" ->
  rc-uniqueness (dual-allocator); "Decision 10" -> memory-model;
  "Decision 11" -> typeclasses (model). "clause-5" -> body-link
  durability gate. "clause-3" (in language-constraints) ->
  bug-class-reintroduction discriminator. "mq.1/2/3", "ct.1/4",
  "eob.1", "rpe.1" -> the canonical-form rule / the type-driven
  dispatch / the Str carve-out / etc. "post-mq.3" -> "type-driven".

design/contracts/feature-acceptance.md: file-local "clauses 1/2/3"
-> "criteria 1/2/3" (sprachliche Kohärenz mit der File-Überschrift
"Feature-acceptance criterion"); "the clause-3 mechanism" -> "the
bug-class-reintroduction discriminator".

Source //! module headers — 24 files:
  Stripped "Iter X.Y:" prefixes and "(Decision N)" / "(mq.X)" tags
  from spec_drift, uniqueness, reuse_shape, migrate_canonical_types,
  typeclass_22b{2,3,c}, suppress_filter, lift, mono, linearity,
  diagnostic, method_dispatch_pin, method_collision_pin,
  no_per_type_print_ops, mq3_multi_class_e2e, print_mono_body_shape,
  print_no_leak_pin, cli_diag_human_workspace_load_error,
  ct1_check_cli, prose snapshot, unbound_in_instance_method_pin,
  mono_xmod_ctor_pattern, desugar.

Central /// public-item rustdoc:
  ast.rs (full sweep — every "Iter X" + "Decision N" prefix
  reformulated; mode/Type::Fn rustdoc now points at memory-model.md;
  Constraint / SuperclassRef / InstanceDef / ClassDef rustdoc points
  at typeclasses contract).
  diagnostic.rs (all "(Iter X)" / "(mq.X)" tags on diagnostic codes
  removed).
  lib.rs (FORM_A_SPEC rustdoc points at authoring-surface.md
  instead of "Decision 6").
  canonical.rs (type_hash + Float-literal rustdoc).

Still outstanding (for a follow-up commit): ~500 inline `//`
code-body comments with `Iter X.Y` markers across the workspace, and
a handful of `///` rustdoc items in hash_pin / workspace_pin / lift /
mono / suppress_filter test-pin and internal-function bodies. Code
identifiers (test filenames like `mq3_multi_class_e2e.rs`, function
names like `iter18e_drop_iterative_default_preserves_hashes`) stay
verbatim per the user's "code identifiers stay verbatim" rule.

Tests: design_index_pin 5/5 + docs_honesty_pin 5/5; workspace builds
clean; full `cargo test --workspace` previously green (every
`test result: ok` line, no FAILED line).

crates/ail/tests/cli_diag_human_workspace_load_error.rs

@@ -1,4 +1,4 @@
-//! Iter cli-diag-human (2026-05-14): non-JSON `ail check` and
+//! Human-readable CLI diagnostics (2026-05-14): non-JSON `ail check` and
 //! sibling subcommands surface `WorkspaceLoadError` with the same
 //! bracketed `[code]` prefix that the JSON path emits.
 //!

crates/ail/tests/ct1_check_cli.rs

@@ -1,4 +1,4 @@
-//! ct.1: E2E coverage for the CLI surface of the canonical-type-names
+//! E2E coverage for the CLI surface of the canonical-type-names
 //! validator. The unit tests in `workspace.rs` already prove the
 //! validator fires; these tests prove the diagnostic survives the
 //! `WorkspaceLoadError -> Diagnostic` translation in

crates/ail/tests/migrate_canonical_types.rs

@@ -1,4 +1,5 @@
-//! ct.1: E2E test for `ail migrate-canonical-types <dir>`.
+//! E2E test for `ail migrate-canonical-types <dir>` (the
+//! canonical-form migration for `Type::Con.name`).
 //!
 //! Builds a synthetic workspace in a tempdir with one fixture that
 //! has a bare cross-module Type::Con ref, runs the migration, then

crates/ail/tests/mono_xmod_ctor_pattern.rs

@@ -4,7 +4,8 @@
 //! by `mono::build_workspace_env`, which delegates to `crate::build_check_env`
 //! and produces a workspace-flat `ctor_index` and `types` map.
 //!
-//! Post-ct.2, `Pattern::Ctor` lookup is type-driven — it consults the
+//! After the canonical-form / type-driven-ctor-lookup refactor,
+//! `Pattern::Ctor` lookup is type-driven — it consults the
 //! scrutinee's canonical `Type::Con.name` to find the TypeDef directly
 //! in `env.module_types`, then validates the ctor name within it. The
 //! mono pass's flat `ctor_index` is no longer consulted by this path;
@@ -14,14 +15,16 @@
 //!
 //! This test pins the cross-module pattern shape against a minimal
 //! 2-module fixture (`test_mono_ctor_main` + `test_mono_ctor_listmod`).
-//! Pre-ct.2 the bug surfaced as `PatternTypeMismatch { ctor: "Cons",
+//! Before the refactor the bug surfaced as
+//! `PatternTypeMismatch { ctor: "Cons",
 //! ty: "test_mono_ctor_listmod.List<Int>" }` because the mono env
-//! resolved `Cons` to bare `List` via the flat index. Post-ct.2 the
+//! resolved `Cons` to bare `List` via the flat index. After the
-//! lookup is type-driven and `expected.name == "test_mono_ctor_listmod.List"`
+//! refactor the lookup is type-driven and
-//! directly indexes the right TypeDef.
+//! `expected.name == "test_mono_ctor_listmod.List"` directly indexes
+//! the right TypeDef.
 //!
-//! Surfaced by iter 23.2 Task 3, which adds `class Eq a` + Eq Int/Bool/Str
+//! Surfaced when `class Eq a` + Eq Int/Bool/Str instances were added
-//! instances to `examples/prelude.ail.json`, flipping the
+//! to `examples/prelude.ail.json`, flipping the
 //! `workspace_has_typeclasses` gate so every workspace exercises the
 //! mono pass.
 

crates/ail/tests/mq3_multi_class_e2e.rs

@@ -1,4 +1,4 @@
-//! mq.3.6: end-to-end coverage of the post-`MethodNameCollision`-
+//! End-to-end coverage of the post-`MethodNameCollision`-
 //! retirement multi-candidate dispatch path. Three positive fixtures
 //! exercise the three trajectories from the milestone spec
 //! §"Data flow":

crates/ail/tests/print_mono_body_shape.rs

@@ -4,7 +4,8 @@
 //! structurally `Term::Let { name: "s", value: App(show__Int, [x]),
 //! body: Term::Do { op: "io/print_str", args: [s] } }`. The explicit
 //! let-binder around `show__Int x` is load-bearing for the heap-Str RC
-//! discipline (eob.1 Str carve-out at `drop_symbol_for_binder` requires
+//! discipline (the heap-Str Str carve-out at
+//! `drop_symbol_for_binder` requires
 //! a let-binder to attach the rc-dec to). If a future codegen / mono
 //! refactor inlines the let-binder away, this pin fires and surfaces
 //! the regression BEFORE the E2E runtime stats produce a confusing

crates/ail/tests/print_no_leak_pin.rs

@@ -1,4 +1,5 @@
-//! RED-pin for the 2026-05-14 rpe.1 Cat-A heap-Str leak in `print`.
+//! RED-pin for the 2026-05-14 Cat-A heap-Str leak in `print`
+//! (uncovered during the per-type-print-op retirement).
 //!
 //! Property protected: under `--alloc=rc`, evaluating the trivial
 //! program `(body (app print 42))` does NOT leak the heap-Str

crates/ail/tests/typeclass_22b2.rs

@@ -1,6 +1,7 @@
-//! Iter 22b.2 typeclass typecheck arms — integration tests.
+//! Typeclass typecheck arms — integration tests.
 //!
-//! This file is shared by 22b.2 tasks: it exists from Task 8
+//! This file holds the cluster of typeclass typecheck-arm integration
+//! tests; it exists from Task 8
 //! (class methods register into module globals) and is extended by
 //! Tasks 9 (`missing-constraint`) and 10 (`no-instance`).
 

crates/ail/tests/typeclass_22b3.rs

@@ -1,4 +1,4 @@
-//! Iter 22b.3: monomorphisation pass tests.
+//! Monomorphisation pass tests.
 //!
 //! Co-located with `typeclass_22b2.rs` so the typeclass-feature
 //! coverage is browsable in one directory. Tests use the same

crates/ail/tests/typeclass_22c.rs

@@ -1,6 +1,6 @@
-//! Iter 22c: milestone-22 acceptance — user-defined classes with
+//! Milestone-22 acceptance: user-defined classes with instances over
-//! instances over user-defined ADTs end-to-end (typecheck → mono →
+//! user-defined ADTs end-to-end (typecheck → mono → codegen →
-//! codegen → binary). The 22b.3 mono pass was tested only against
+//! binary). The mono pass was previously tested only against
 //! instances over primitive types; this file pins the user-ADT path.
 //!
 //! Co-located with `typeclass_22b3.rs` so the typeclass-feature

crates/ail/tests/unbound_in_instance_method_pin.rs

@@ -23,8 +23,9 @@
 //! Root cause (from debugger Phase 1-2):
 //! `crates/ailang-check/src/lib.rs::check_def` early-returns `Ok(())`
 //! for `Def::Class | Def::Instance` (the comment claims body
-//! typechecking landed in iter 22b.2, but the body-walk was never
+//! typechecking landed alongside the typeclass-typecheck arms, but
-//! wired). Only the workspace-load coherence checks
+//! the body-walk was never wired). Only the workspace-load coherence
+//! checks
 //! (Orphan/Duplicate/MissingMethod) in `workspace::build_registry`
 //! touch instance defs, and those only inspect the schema, not the
 //! method-body identifier graph.

crates/ailang-check/src/diagnostic.rs

@@ -39,7 +39,7 @@
 //! - `module-cycle` — workspace loader (Iter 5b, in the CLI path)
 //! - `module-name-mismatch` — workspace loader (Iter 5b, in the CLI path)
 //! - `module-hash-mismatch` — workspace loader (Iter 5b, in the CLI path)
-//! - `tail-call-not-in-tail-position` (Iter 14e, see Decision 8)
+//! - `tail-call-not-in-tail-position` (see `design/contracts/tail-calls.md`)
 //! - `use-after-consume` — `ctx`: `{"binder": "<n>"}` (Iter 18c.2);
 //!   carries non-empty [`Diagnostic::suggested_rewrites`] showing how to
 //!   spell the fix in form-A AILang.
@@ -61,38 +61,38 @@
 //!   the build refuses it. Note that an invalid suppression does NOT
 //!   suppress its target diagnostic — the original still fires
 //!   alongside this error.
-//! - `missing-constraint` (Iter 22b.2) — `severity: error`. Emitted
+//! - `missing-constraint` — `severity: error`. Emitted
 //!   by the per-fn typecheck arm when a polymorphic `FnDef` calls a
 //!   class method (e.g. `show x` where `x: a`) but its declared
 //!   `Forall.constraints` (after one-step superclass expansion) does
 //!   not include the residual class constraint. `ctx`:
 //!   `{"class": "<C>", "method": "<m>", "at_type": "<a>"}`. Concrete-
 //!   type residuals are deferred to the `no-instance` diagnostic.
-//! - `no-instance` (Iter 22b.2) — `severity: error`. Emitted by the
+//! - `no-instance` — `severity: error`. Emitted by the
 //!   per-fn typecheck arm when a class-method call resolves the class
 //!   param to a fully-concrete type that has no matching entry in the
 //!   workspace instance registry (`(class, canonical-type-hash)`).
 //!   Dual of `missing-constraint`: when the residual is concrete, the
 //!   fn cannot push the obligation to a caller, so an existing
 //!   instance is the only way to discharge it. `ctx`:
-//!   `{"class": "<C>", "method": "<m>", "at_type": "<T>"}`. mq.2 adds
+//!   `{"class": "<C>", "method": "<m>", "at_type": "<T>"}`. An
-//!   an optional `candidate_classes` field when the residual originated
+//!   optional `candidate_classes` field is added when the residual
-//!   from the multi-candidate dispatch path; absent on single-class
+//!   originated from the multi-candidate dispatch path; absent on
-//!   residuals (back-compat).
+//!   single-class residuals (back-compat).
-//! - `ambiguous-method-resolution` (mq.2) — `severity: error`. Emitted
+//! - `ambiguous-method-resolution` — `severity: error`. Emitted
 //!   by the type-driven dispatch resolver (or the discharge-time
 //!   multi-candidate refinement) when a bare-method call site survives
 //!   both type-driven and constraint-driven filtering with more than
 //!   one candidate class. `ctx`: `{"method": "<m>", "at_type": "<T>",
 //!   "candidate_classes": ["<C1>", ...]}`. The author disambiguates
 //!   by writing `<ClassQualifier>.<method> x`.
-//! - `unknown-class` (mq.2) — `severity: error`. Emitted by the
+//! - `unknown-class` — `severity: error`. Emitted by the
 //!   type-driven dispatch resolver when an explicit class qualifier
 //!   in a `Term::Var.name` (e.g. `"prelude.Show.show"`) names a
 //!   qualified class that is not in the workspace registry of
 //!   candidate classes for the method. `ctx`:
 //!   `{"name": "<qualified-class>"}`.
-//! - `class-method-shadowed-by-fn` (mq.3) — `severity: warning`.
+//! - `class-method-shadowed-by-fn` — `severity: warning`.
 //!   Emitted by `synth`'s `Term::Var` arm when a name resolves to a
 //!   free fn (locals / caller-module-fn / imported-fn) AND a class
 //!   method of the same name also exists in the workspace. Fn

crates/ailang-check/src/lift.rs

@@ -1,4 +1,4 @@
-//! Iter 16b.3: post-typecheck `Term::LetRec` lift.
+//! Post-typecheck `Term::LetRec` lift.
 //!
 //! Background. The 16a desugar pass (`ailang-core::desugar::desugar_module`)
 //! eliminates most `Term::LetRec` nodes by lifting them to synthetic

crates/ailang-check/src/linearity.rs

@@ -1,4 +1,4 @@
-//! Iter 18c.2: linearity check for fns whose every parameter mode is
+//! Linearity check for fns whose every parameter mode is
 //! explicit (`Borrow` or `Own`).
 //!
 //! ## Scope

crates/ailang-check/src/mono.rs

@@ -1,5 +1,5 @@
-//! Workspace monomorphisation pass (introduced iter 22b.3; the
+//! Workspace monomorphisation pass — class-method entry plus
-//! free-fn entry was added in iter 23.4).
+//! free-fn entry.
 //!
 //! Slots into the build pipeline after [`crate::lift_letrecs`] and
 //! before `ailang_codegen::lower_workspace_with_alloc`. It is only

crates/ailang-check/src/reuse_shape.rs

@@ -1,4 +1,4 @@
-//! Iter 18d.2: shape-compatibility check for `(reuse-as <var> <body>)`.
+//! Shape-compatibility check for `(reuse-as <var> <body>)`.
 //!
 //! ## Why this is its own pass
 //!

crates/ailang-check/src/suppress_filter.rs

@@ -1,4 +1,4 @@
-//! Iter 19b: per-module post-process that consumes
+//! Per-module post-process that consumes
 //! [`ailang_core::ast::FnDef::suppress`] entries.
 //!
 //! For each `Def::Fn(f)` and each `Suppress { code, because }` in

crates/ailang-check/src/uniqueness.rs

@@ -1,4 +1,4 @@
-//! Iter 18c.3: uniqueness inference over the typed AST.
+//! Uniqueness inference over the typed AST.
 //!
 //! This pass classifies every binder of every fn body as either
 //! [`Uniqueness::Unique`] or [`Uniqueness::Shared`]. Unlike the

crates/ailang-check/tests/method_collision_pin.rs

@@ -1,6 +1,7 @@
-//! mq.3.5: repurposed pin tests for the post-`MethodNameCollision`-
+//! Repurposed pin tests for the post-`MethodNameCollision`-retirement
-//! retirement workspace-load path. The two on-disk fixtures that fired
+//! workspace-load path. The two on-disk fixtures that fired
-//! `WorkspaceLoadError::MethodNameCollision` pre-mq.3 now load cleanly;
+//! `WorkspaceLoadError::MethodNameCollision` before the type-driven
+//! dispatch refactor now load cleanly;
 //! the assertion migrates from "expect collision diagnostic" to "load
 //! successful + `Env.method_to_candidate_classes` carries the
 //! expected multi-entry set" (class-class case) or "load successful +

crates/ailang-check/tests/method_dispatch_pin.rs

@@ -1,4 +1,4 @@
-//! mq.2.5: pin tests on `resolve_method_dispatch` — the new
+//! Pin tests on `resolve_method_dispatch` — the
 //! dispatch-resolution helper that synth's `Term::Var` arm consults
 //! per the spec's 5-step rule.
 //!

crates/ailang-check/tests/no_per_type_print_ops.rs

@@ -1,4 +1,4 @@
-//! Hard gate for iter rpe.1: the per-type print effect-ops
+//! Hard gate against re-introducing the per-type print effect-ops:
 //! `io/print_int`, `io/print_bool`, `io/print_float` are retired
 //! and must NOT appear in the builtin registry. The polymorphic
 //! `print` helper (prelude, iter 24.3) is the canonical replacement;

crates/ailang-core/src/ast.rs

@@ -64,8 +64,8 @@ pub struct Import {
 /// [`Def::name`] method) to inspect generically without matching every
 /// variant.
 ///
-/// Iter 22b.1 (Decision 11): adds `Class` and `Instance`. Both are
+/// `Class` and `Instance` are additive — fixtures that predate
-/// additive — pre-22b fixtures never produce these tags, so their
+/// the typeclass layer never produce these tags, so their
 /// canonical-JSON bytes (and therefore their content hashes) are
 /// unchanged. The `Def`-level match exhaustiveness in downstream
 /// crates is the only caller that has to acknowledge the variants.
@@ -78,9 +78,9 @@ pub enum Def {
     Const(ConstDef),
     /// Type (ADT) definition; see [`TypeDef`].
     Type(TypeDef),
-    /// Iter 22b.1: typeclass declaration; see [`ClassDef`].
+    /// Typeclass declaration; see [`ClassDef`].
     Class(ClassDef),
-    /// Iter 22b.1: instance declaration; see [`InstanceDef`].
+    /// Instance declaration; see [`InstanceDef`].
     Instance(InstanceDef),
 }
 
@@ -136,9 +136,8 @@ pub fn def_kind(def: &Def) -> &'static str {
 pub struct TypeDef {
     /// Type name (capitalised by convention).
     pub name: String,
-    /// Type parameters (Iter 13a parameterised-ADT support). A
+    /// Type parameters for parameterised ADTs. A monomorphic ADT has
-    /// monomorphic ADT has `vars` empty and is serialized identically
+    /// `vars` empty and the field is **omitted** when empty,
-    /// to the pre-13a schema (the field is **omitted** when empty),
     /// preserving the canonical-JSON hash of every existing module on
     /// disk. See the regression test
     /// `iter13a_schema_extension_preserves_pre_13a_hashes` in
@@ -151,7 +150,7 @@ pub struct TypeDef {
     /// Optional source-level documentation string.
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub doc: Option<String>,
-    /// Iter 18e: opt-in `(drop-iterative)` annotation. When `true`,
+    /// Opt-in `(drop-iterative)` annotation. When `true`,
     /// codegen emits `drop_<m>_<T>` with an iterative worklist body
     /// instead of the recursive cascade — chosen by the LLM-author when
     /// the type is expected to form long chains (millions of cells)
@@ -159,9 +158,9 @@ pub struct TypeDef {
     ///
     /// Serialised as `"drop-iterative": true` (kebab-case) when set;
     /// the field is omitted when `false` so canonical-JSON hashes of
-    /// every pre-18e fixture remain bit-stable. See the regression
+    /// every fixture that does not opt in remain bit-stable. See the
-    /// test `iter18e_drop_iterative_default_preserves_hashes` in
+    /// regression test `iter18e_drop_iterative_default_preserves_hashes`
-    /// [`crate::hash`].
+    /// in [`crate::hash`].
     #[serde(
         default,
         rename = "drop-iterative",
@@ -219,24 +218,24 @@ pub struct FnDef {
     /// the same additive-schema pattern as [`FnDef::doc`].
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub export: Option<String>,
-    /// Iter 19b: structured-diagnostic suppressions opted into for this
+    /// Structured-diagnostic suppressions opted into for this
     /// fn. Each entry names a diagnostic code and an author-asserted
-    /// reason. Currently the only consumer is `over-strict-mode`
+    /// reason. The only current consumer is `over-strict-mode`, but
-    /// (Iter 19a / 19a.1) but the mechanism is generic across codes.
+    /// the mechanism is generic across codes. `because` must be
-    /// `because` must be non-empty — the typechecker emits
+    /// non-empty — the typechecker emits `empty-suppress-reason`
-    /// `empty-suppress-reason` (Error) otherwise.
+    /// (Error) otherwise.
     ///
     /// Serialised with `skip_serializing_if = "Vec::is_empty"` so
-    /// every pre-19b fixture's canonical-JSON hash stays bit-identical.
+    /// every pre-suppress fixture's canonical-JSON hash stays
-    /// The same additive-schema pattern is used by [`TypeDef::vars`]
+    /// bit-identical. The same additive-schema pattern is used by
-    /// (Iter 13a) and [`Type::Con::args`] (Iter 13a).
+    /// [`TypeDef::vars`] and [`Type::Con::args`].
     #[serde(default, skip_serializing_if = "Vec::is_empty")]
     pub suppress: Vec<Suppress>,
 }
 
-/// Iter 19b: one entry in [`FnDef::suppress`]. Marks a structured
+/// One entry in [`FnDef::suppress`]. Marks a structured diagnostic
-/// diagnostic the author has consciously decided to allow on this
+/// the author has consciously decided to allow on this def, with a
-/// def, with a mandatory reason.
+/// mandatory reason.
 ///
 /// `because` is non-empty by schema rule — the typechecker emits
 /// `empty-suppress-reason` (Error severity) when it is empty or
@@ -256,20 +255,22 @@ pub struct Suppress {
     pub because: String,
 }
 
-/// Iter 22b.1: a typeclass declaration (Decision 11).
+/// A typeclass declaration (narrative in
+/// `design/contracts/typeclasses.md`).
 ///
 /// Single-parameter, multi-method, optional-default, optional-superclass
 /// typeclass. The `param` is a single string — multi-param classes are
-/// rejected by Decision 11 axis 1, and the schema enforces it by shape
+/// rejected by the typeclass design (concrete-types-only, kind `*`),
-/// (`param: String`, not `Vec<String>`).
+/// and the schema enforces it by shape (`param: String`, not
+/// `Vec<String>`).
 ///
 /// `superclass`, when present, MUST have its `type` field equal to
-/// `param`. The check is enforced in 22b.2 via the
+/// `param`. The check is enforced via the `InvalidSuperclassParam`
-/// `InvalidSuperclassParam` diagnostic; the schema does not encode it.
+/// diagnostic; the schema does not encode it.
 ///
 /// All optional fields are omitted from canonical JSON when absent /
 /// empty, so future schema evolution can land here without disturbing
-/// pre-22b hashes.
+/// existing hashes.
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct ClassDef {
     /// Class name (capitalised by convention). Bare — symmetric to
@@ -277,8 +278,7 @@ pub struct ClassDef {
     /// a reference. Cross-module class references live in
     /// `InstanceDef.class`, `Constraint.class`, and
     /// `SuperclassRef.class`; those fields carry the canonical form
-    /// (bare for same-module, `<module>.<Class>` for cross-module)
+    /// (bare for same-module, `<module>.<Class>` for cross-module).
-    /// per mq.1.
     pub name: String,
     /// Single type-parameter name.
     pub param: String,
@@ -292,20 +292,21 @@ pub struct ClassDef {
     pub doc: Option<String>,
 }
 
-/// Iter 22b.1: reference to a superclass relation in [`ClassDef`].
+/// Reference to a superclass relation in [`ClassDef`].
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct SuperclassRef {
-    /// Superclass name in canonical form (mq.1): bare for a
+    /// Superclass name in canonical form: bare for a same-module
-    /// same-module class, `<module>.<Class>` for a cross-module
+    /// class, `<module>.<Class>` for a cross-module class.
-    /// class. Symmetric to ct.1's `Type::Con.name` rule.
+    /// Symmetric to `Type::Con.name`'s canonical-form rule.
     pub class: String,
     /// Type the superclass is applied to. MUST equal the parent
-    /// `ClassDef.param` (validated in 22b.2 — schema does not enforce).
+    /// `ClassDef.param` (validated by typecheck — schema does not
+    /// enforce).
     #[serde(rename = "type")]
     pub type_: String,
 }
 
-/// Iter 22b.1: one method declared in a [`ClassDef`].
+/// One method declared in a [`ClassDef`].
 ///
 /// `default` is `None` when the method is abstract-required (every
 /// instance must specify it); `Some(body)` when the method has a
@@ -324,7 +325,8 @@ pub struct ClassMethod {
     pub default: Option<Term>,
 }
 
-/// Iter 22b.1: an instance declaration (Decision 11).
+/// An instance declaration (narrative in
+/// `design/contracts/typeclasses.md`).
 ///
 /// `class` is the name of the class being instantiated. `type_` is the
 /// concrete type expression the class is applied to — never the class
@@ -332,9 +334,9 @@ pub struct ClassMethod {
 /// overrides of default-bearing methods.
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct InstanceDef {
-    /// Class being instantiated, in canonical form (mq.1): bare for
+    /// Class being instantiated, in canonical form: bare for a
-    /// a same-module class, `<module>.<Class>` for a cross-module
+    /// same-module class, `<module>.<Class>` for a cross-module
-    /// class. Symmetric to ct.1's `Type::Con.name` rule.
+    /// class. Symmetric to `Type::Con.name`'s canonical-form rule.
     pub class: String,
     /// Concrete type the class is applied to.
     #[serde(rename = "type")]
@@ -346,28 +348,29 @@ pub struct InstanceDef {
     pub doc: Option<String>,
 }
 
-/// Iter 22b.1: one method body in an [`InstanceDef`].
+/// One method body in an [`InstanceDef`].
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct InstanceMethod {
     /// Method name (must match a method in the corresponding class).
     pub name: String,
     /// Method body. The class's declared method type with the class
     /// param substituted to the instance type is the body's expected
-    /// type — checked in 22b.2.
+    /// type.
     pub body: Term,
 }
 
-/// Iter 22b.2: a class constraint on a polymorphic function (Decision
+/// A class constraint on a polymorphic function (narrative in
-/// 11). `(class, type)` pair where `class` is a class name and `type`
+/// `design/contracts/typeclasses.md`). `(class, type)` pair where
-/// is a `Type` expression — typically a single `Type::Var` (e.g.
+/// `class` is a class name and `type` is a `Type` expression —
-/// `(Eq, a)` for `Eq a => ...`). Concrete-type constraints are legal
+/// typically a single `Type::Var` (e.g. `(Eq, a)` for
-/// schema-wise but fired as `no-instance` at typecheck time if no
+/// `Eq a => ...`). Concrete-type constraints are legal schema-wise
-/// matching registry entry exists.
+/// but fired as `no-instance` at typecheck time if no matching
+/// registry entry exists.
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct Constraint {
-    /// Class name in canonical form (mq.1): bare for a same-module
+    /// Class name in canonical form: bare for a same-module class,
-    /// class, `<module>.<Class>` for a cross-module class.
+    /// `<module>.<Class>` for a cross-module class. Symmetric to
-    /// Symmetric to ct.1's `Type::Con.name` rule.
+    /// `Type::Con.name`'s canonical-form rule.
     pub class: String,
     /// Type the class is applied to.
     #[serde(rename = "type")]
@@ -408,11 +411,11 @@ pub enum Term {
     /// Function application. `callee` is evaluated to a function value;
     /// `args` are evaluated left-to-right.
     ///
-    /// Iter 14e: `tail` marks this call as occurring in tail position
+    /// `tail` marks this call as occurring in tail position (see
-    /// (per Decision 8). When set, codegen lowers the call as
+    /// `design/contracts/tail-calls.md`). When set, codegen lowers
-    /// `musttail call`. The flag defaults to `false` and is omitted
+    /// the call as `musttail call`. The flag defaults to `false` and
-    /// during canonical-JSON serialisation when unset, so pre-14e
+    /// is omitted during canonical-JSON serialisation when unset, so
-    /// fixtures keep bit-identical hashes.
+    /// pre-tail-flag fixtures keep bit-identical hashes.
     App {
         #[serde(rename = "fn")]
         callee: Box<Term>,
@@ -426,7 +429,7 @@ pub enum Term {
         value: Box<Term>,
         body: Box<Term>,
     },
-    /// Local recursive let-binding (Iter 16b.1). Always fn-shaped:
+    /// Local recursive let-binding. Always fn-shaped:
     /// the bound name `name` is recursively visible inside `body`.
     /// Eliminated by `crate::desugar` before typecheck — lifted to a
     /// synthetic top-level fn when `body` does not capture any name
@@ -455,7 +458,7 @@ pub enum Term {
     /// `match` in codegen (`lower_effect_op`); there is no
     /// effect-handler table and no link-time resolution.
     ///
-    /// Iter 14e: see [`Term::App`] for the `tail` field semantics.
+    /// See [`Term::App`] for the `tail` field semantics.
     Do {
         op: String,
         args: Vec<Term>,
@@ -479,7 +482,7 @@ pub enum Term {
         scrutinee: Box<Term>,
         arms: Vec<Arm>,
     },
-    /// Anonymous function (Iter 8b). Captures any free variables of
+    /// Anonymous function. Captures any free variables of
     /// `body` from the enclosing scope. Param/return types are
     /// declared inline so the typechecker stays HM-monomorphic on
     /// the inferred shape.
@@ -493,7 +496,7 @@ pub enum Term {
         effects: Vec<String>,
         body: Box<Term>,
     },
-    /// Sequencing (Iter 10). `lhs` is evaluated for its effects and its
+    /// Sequencing. `lhs` is evaluated for its effects and its
     /// result discarded; `rhs` is the value of the whole expression.
     /// Equivalent to `let _ = lhs in rhs`, but with a dedicated node so
     /// pretty-print and diagnostics read cleanly.
@@ -501,27 +504,25 @@ pub enum Term {
         lhs: Box<Term>,
         rhs: Box<Term>,
     },
-    /// Iter 18c.1: explicit RC clone. Lowers identically to its inner
+    /// Explicit RC clone. Codegen emits
-    /// term in 18c.1; in 18c.3 the codegen will emit
     /// `call void @ailang_rc_inc(ptr %v)` before returning `%v` under
     /// `--alloc=rc`. The variant is additive: `(clone X)` round-trips
-    /// through every pre-18c.1 fixture without their hashes changing
+    /// through every fixture that does not use the tag without its
-    /// because none of them use the new tag.
+    /// hash changing.
     Clone {
         value: Box<Term>,
     },
-    /// Iter 18d.1: explicit reuse-as hint. Wraps an allocating `body`
+    /// Explicit reuse-as hint. Wraps an allocating `body` (typically
-    /// (typically `Term::Ctor`, also `Term::Lam`) and names a `source`
+    /// `Term::Ctor`, also `Term::Lam`) and names a `source` term
-    /// term whose memory slot the body's allocation should reuse.
+    /// whose memory slot the body's allocation should reuse.
     /// Conventionally `source` is a `Term::Var { name }` — the
     /// linearity check rejects anything else with
     /// `reuse-as-source-not-bare-var`. The body must be allocating;
     /// non-allocating bodies are rejected at typecheck with
-    /// `reuse-as-non-allocating-body`. Lowers as identity in 18d.1
+    /// `reuse-as-non-allocating-body`. Codegen lowers as in-place
-    /// (returns `body`'s `(ssa, ty)`, ignores `source`); 18d.2 will
+    /// rewrite under `--alloc=rc`. The variant is additive —
-    /// lower this as in-place rewrite under `--alloc=rc`. The variant
+    /// fixtures that do not use the tag keep their canonical-JSON
-    /// is additive — pre-18d fixtures keep their canonical-JSON hash
+    /// hash.
-    /// because none of them use the new tag.
     #[serde(rename = "reuse-as")]
     ReuseAs {
         source: Box<Term>,
@@ -685,7 +686,7 @@ pub enum Type {
     /// and parameterised forms like `List<Int>`.
     Con {
         name: String,
-        /// Type arguments (Iter 13a). For pre-13a uses (`Int`, `Bool`,
+        /// Type arguments. For non-parameterised uses (`Int`, `Bool`,
         /// non-parameterised user ADTs) this stays empty and is
         /// **omitted** during serialization, so the canonical-JSON
         /// hash of every pre-existing module remains bit-identical.
@@ -697,14 +698,15 @@ pub enum Type {
     /// set; equality compares it modulo order (see the [`PartialEq`]
     /// impl below).
     ///
-    /// Iter 18a (Decision 10): `param_modes` and `ret_mode` carry the
+    /// `param_modes` and `ret_mode` carry the `(borrow T)` /
-    /// `(borrow T)` / `(own T)` wrappers from the surface form. They
+    /// `(own T)` wrappers from the surface form. They are metadata
-    /// are metadata on `Type::Fn`, not new `Type` variants — so
+    /// on `Type::Fn`, not new `Type` variants — so unification,
-    /// unification, occurs, apply, and every other `Type` match-arm
+    /// occurs, apply, and every other `Type` match-arm keeps working
-    /// keeps working unchanged. `param_modes` is omitted from
+    /// unchanged. `param_modes` is omitted from canonical JSON when
-    /// canonical JSON when every entry is `Implicit`; `ret_mode` is
+    /// every entry is `Implicit`; `ret_mode` is omitted when it is
-    /// omitted when it is `Implicit`. Pre-18a fixtures therefore hash
+    /// `Implicit`, so pre-mode-annotation fixtures hash
-    /// bit-identically.
+    /// bit-identically. Full contract in
+    /// `design/contracts/memory-model.md`.
     Fn {
         params: Vec<Type>,
         #[serde(default, skip_serializing_if = "all_implicit")]
@@ -726,9 +728,9 @@ pub enum Type {
     /// names, instantiated fresh at each use site.
     Forall {
         vars: Vec<String>,
-        /// Iter 22b.2: class constraints quantified together with
+        /// Class constraints quantified together with `vars`. Empty
-        /// `vars`. Empty for pre-22b.2 polymorphic types; serialised
+        /// for unconstrained polymorphic types; serialised with
-        /// with `skip_serializing_if = "Vec::is_empty"` so existing
+        /// `skip_serializing_if = "Vec::is_empty"` so existing
         /// canonical-JSON bytes stay bit-identical.
         #[serde(default, skip_serializing_if = "Vec::is_empty")]
         constraints: Vec<Constraint>,
@@ -759,11 +761,11 @@ impl Type {
         Type::Con { name: "Float".into(), args: vec![] }
     }
 
-    /// Iter 18a: build a `Type::Fn` with all parameter modes set to
+    /// Build a `Type::Fn` with all parameter modes set to
-    /// `ParamMode::Implicit` and `ret_mode` set to `Implicit`. This is
+    /// `ParamMode::Implicit` and `ret_mode` set to `Implicit`. This
-    /// the form every typechecker / desugar / codegen site that
+    /// is the form every typechecker / desugar / codegen site that
     /// synthesises a fn-type should use, so that newly inferred
-    /// fn-types retain pre-18a canonical-JSON bytes.
+    /// fn-types retain pre-mode-annotation canonical-JSON bytes.
     pub fn fn_implicit(params: Vec<Type>, ret: Type, effects: Vec<String>) -> Type {
         let n = params.len();
         Type::Fn {
@@ -776,21 +778,21 @@ impl Type {
     }
 }
 
-/// Iter 18a (Decision 10): per-parameter / return mode marker on a
+/// Per-parameter / return mode marker on a [`Type::Fn`]. Full
-/// [`Type::Fn`].
+/// contract lives in `design/contracts/memory-model.md`.
 ///
 /// `Implicit` is the legacy state for fn-types that were constructed
 /// before the borrow/own surface annotations existed. Semantically,
-/// `Implicit ≡ Own` throughout the 18-series; the distinction exists
+/// `Implicit ≡ Own`; the distinction exists only so pre-annotation
-/// only so pre-18a JSON fixtures continue to serialize without a
+/// JSON fixtures continue to serialize without a `"mode"` wrapper
-/// `"mode"` wrapper and therefore keep their canonical-JSON hash.
+/// and therefore keep their canonical-JSON hash.
 ///
 /// `Own` and `Borrow` are author-asserted: the surface form
 /// `(own T)` / `(borrow T)` round-trips through this enum.
 #[derive(Clone, Copy, Debug, Default, PartialEq, Eq, Serialize, Deserialize)]
 #[serde(rename_all = "lowercase")]
 pub enum ParamMode {
-    /// Pre-18a / unannotated. Treated as `Own` by the typechecker.
+    /// Unannotated / back-compat. Treated as `Own` by the typechecker.
     #[default]
     Implicit,
     /// `(own T)` — caller transfers ownership; callee consumes.
@@ -807,21 +809,20 @@ impl ParamMode {
     }
 }
 
-/// Iter 18a: serde helper for [`Type::Fn::param_modes`]. Returns
+/// Serde helper for [`Type::Fn::param_modes`]. Returns `true` when
-/// `true` when every entry is [`ParamMode::Implicit`] (or when the
+/// every entry is [`ParamMode::Implicit`] (or when the list is
-/// list is empty), so canonical JSON omits the field for any fn-type
+/// empty), so canonical JSON omits the field for any fn-type without
-/// without explicit `(borrow)` / `(own)` annotations and pre-18a
+/// explicit `(borrow)` / `(own)` annotations and pre-annotation
 /// fixtures keep bit-identical hashes.
 fn all_implicit(modes: &[ParamMode]) -> bool {
     modes.iter().all(|m| m.is_implicit())
 }
 
-/// Iter 18a: equality of [`ParamMode`] for the purposes of `Type`
+/// Equality of [`ParamMode`] for the purposes of `Type` equality.
-/// equality. `Implicit` and `Own` are treated as the same mode
+/// `Implicit` and `Own` are treated as the same mode; `Borrow` is
-/// throughout the 18-series; `Borrow` is distinct. This keeps
+/// distinct. This keeps pre-annotation fixtures (whose fn-types
-/// pre-18a fixtures (whose fn-types serialize `Implicit`) compatible
+/// serialize `Implicit`) compatible with newly-written fixtures
-/// with newly-written 18a fixtures that mark the same fn-type
+/// that mark the same fn-type explicitly with `(own T)`.
-/// explicitly with `(own T)`.
 fn mode_eq(a: &ParamMode, b: &ParamMode) -> bool {
     match (a, b) {
         (ParamMode::Borrow, ParamMode::Borrow) => true,
@@ -831,7 +832,7 @@ fn mode_eq(a: &ParamMode, b: &ParamMode) -> bool {
     }
 }
 
-/// Iter 18a: equality of two `param_modes` slices, robust to the
+/// Equality of two `param_modes` slices, robust to the
 /// "elided when all-implicit" representation used by typechecker /
 /// desugar / codegen sites that construct fn-types with
 /// `param_modes: vec![]`. Both slices are normalised to "implicit
@@ -897,9 +898,10 @@ impl Eq for Type {}
 
 /// Serde helper for `#[serde(skip_serializing_if = "is_false")]`.
 ///
-/// Used by [`Term::App::tail`] and [`Term::Do::tail`] (Iter 14e) so the
+/// Used by [`Term::App::tail`] and [`Term::Do::tail`] so the `tail`
-/// `tail` flag is omitted from the canonical JSON whenever it is false,
+/// flag is omitted from the canonical JSON whenever it is false,
-/// preserving bit-identical hashes for every pre-14e definition.
+/// preserving bit-identical hashes for every fixture that does not
+/// carry the flag.
 #[allow(clippy::trivially_copy_pass_by_ref)]
 fn is_false(b: &bool) -> bool {
     !*b

crates/ailang-core/src/canonical.rs

@@ -51,7 +51,7 @@ pub fn to_bytes<T: serde::Serialize>(value: &T) -> Vec<u8> {
     out
 }
 
-/// Iter 22b.1: 16-hex-char hash of a [`crate::ast::Type`] in isolation.
+/// 16-hex-char hash of a [`crate::ast::Type`] in isolation.
 ///
 /// Used by [`crate::workspace::Registry`] to key `InstanceDef`s by
 /// their target type. Parallel in shape to [`crate::def_hash`] and
@@ -138,7 +138,7 @@ mod tests {
         assert!(!s.contains('\n'));
     }
 
-    /// Iter 22-floats.1 RED: a `Literal::Float` carries the IEEE-754
+    /// A `Literal::Float` carries the IEEE-754
     /// binary64 bit pattern as a `u64`; canonical JSON encodes it as
     /// `{"bits":"<16-lowercase-hex>","kind":"float"}` — string path,
     /// NOT through `serde_json::Number` (which is not bit-stable for

crates/ailang-core/src/desugar.rs

@@ -29,14 +29,14 @@
 //!    re-evaluate effectful scrutinees per arm).
 //! 3. Build a chain of single-level matches via
 //!    `build_chain(s_var, arms, default)` where `default` is the
-//!    polymorphic bottom builtin `__unreachable__` (`forall a. a`,
+//!    polymorphic bottom builtin `__unreachable__` (`forall a. a`)
-//!    Iter 16d) — codegen lowers it to LLVM `unreachable`. Valid
+//!    — codegen lowers it to LLVM `unreachable`. Valid programs
-//!    programs never reach it because the checker requires
+//!    never reach it because the checker requires exhaustiveness
-//!    exhaustiveness (the catch-all arm dominates the chain).
+//!    (the catch-all arm dominates the chain). An earlier shape used
-//!    Pre-16d the default was a `Unit` literal, which forced any
+//!    a `Unit` literal as the default, which forced any match whose
-//!    match whose arms returned a non-Unit type to carry a
+//!    arms returned a non-Unit type to carry a synthetic `_` arm
-//!    synthetic `_` arm dominating the terminator. The polymorphic
+//!    dominating the terminator. The polymorphic `__unreachable__`
-//!    `__unreachable__` removes that workaround.
+//!    removes that workaround.
 //! 4. Each arm is lowered via `desugar_one_arm`:
 //!    - `Pattern::Wild` → arm body (catch-all; later arms are dropped).
 //!    - `Pattern::Var { name }` → `Let { name = scrutinee_var; body }`.

crates/ailang-core/src/lib.rs

@@ -107,11 +107,11 @@ pub type Result<T> = std::result::Result<T, Error>;
 /// loading fails with [`Error::SchemaMismatch`].
 pub const SCHEMA: &str = "ailang/v0";
 
-/// Iter 20f: complete LLM-targeted specification of Form-A — the
+/// Complete LLM-targeted specification of Form-A — the canonical
-/// canonical authoring surface (Decision 6). Embedded verbatim into
+/// authoring surface (see `design/contracts/authoring-surface.md`).
-/// any prompt that asks an LLM to produce or edit AILang code; in
+/// Embedded verbatim into any prompt that asks an LLM to produce or
-/// particular, `ail merge-prose` includes it in the round-trip
+/// edit AILang code; in particular, `ail merge-prose` includes it in
-/// prompt template. The string is the raw bytes of
+/// the round-trip prompt template. The string is the raw bytes of
 /// `specs/form_a.md`, co-located with this crate so the spec sits
 /// next to the AST it describes; a drift test
 /// (`tests/spec_drift.rs`) walks every AST enum variant and asserts

crates/ailang-core/tests/spec_drift.rs

@@ -1,4 +1,4 @@
-//! Iter 20f: drift detection between the AST and `specs/form_a.md`.
+//! Drift detection between the AST and `specs/form_a.md`.
 //!
 //! The spec is hand-curated, but it cannot silently fall behind the
 //! language. Every AST enum (`Term`, `Pattern`, `Type`, `Def`, `Literal`,

crates/ailang-prose/tests/snapshot.rs

@@ -1,4 +1,4 @@
-//! Iter 20a snapshot tests.
+//! Prose-projection snapshot tests.
 //!
 //! For each `examples/<name>.ail`, the renderer's output is compared
 //! against the committed `examples/<name>.prose.txt`. The snapshot

design/INDEX.md

@@ -35,9 +35,9 @@ evolving in lockstep with the language:
   `crates/ailang-codegen`): AST, type system, codegen.
 - **Surface forms** (`crates/ailang-surface`, `crates/ailang-prose`):
   the LLM-facing renderings of a module. `ailang-surface`
-  is the lossless Form-A printer/parser — the canonical authoring
+  is the lossless Form-A printer/parser — the canonical
-  surface fixed by Decision 6, with a round-trip property
+  [authoring surface](contracts/authoring-surface.md), with a
-  `parse ∘ print = id` gating every release. `ailang-prose`
+  round-trip property `parse ∘ print = id` gating every release. `ailang-prose`
   is the lossy Form-B projection — human-readable prose for review and
   edit, with no parser; re-integration goes through the
   LLM-mediator round-trip documented in `docs/PROSE_ROUNDTRIP.md`.

design/contracts/authoring-surface.md

@@ -1,6 +1,6 @@
 # Authoring surface
 
-## Decision 6: authoring surface
+## Authoring surface
 
 Form (A) is implemented as
 the `ailang-surface` crate (parser + printer). Form-A is
@@ -22,7 +22,7 @@ one projection among potentially many. Concretely:
   addressed representation of a module. All hashing, content-
   addressing, cross-module references, and typecheck/codegen input
   flow through the JSON-AST. **No new hashable form is introduced.**
-- The textual surface (form A, this Decision) is the **AI authoring
+- The textual surface (form A, this contract) is the **AI authoring
   projection**: optimised for me producing programs token-efficiently
   and for foreign LLMs producing programs from a spec. It is not
   optimised for human authors and does not need to be human-pleasant.
@@ -55,11 +55,11 @@ the AST and remain projection-agnostic.
    shape. The full bijection between `.ail.json` and `.ail` (both
    directions, BLAKE3-stable hashing, Float-bits-hex encoding,
    workspace-CI enforcement points) is anchored as the top-level
-   [Roundtrip Invariant](roundtrip-invariant.md) — this constraint records that Decision
+   [Roundtrip Invariant](roundtrip-invariant.md) — this constraint
-   6's surface-design choice must satisfy that invariant; the
+   records that the surface-design choice must satisfy that
-   invariant itself lives at top level because the property is
+   invariant; the invariant itself lives at top level because the
-   load-bearing on the language identity, not on this Decision's
+   property is load-bearing on the language identity, not on this
-   surface-design rationale.
+   file's surface-design rationale.
 3. **No external symbols.** ASCII only. No Greek (`∀`), no arrows
    (`→`), no subscripts. Reasoning: I substitute mojibake for
    non-ASCII characters under context pressure; foreign LLMs vary

design/contracts/data-model.md

@@ -41,7 +41,7 @@ narrative — defaults, superclasses, diagnostics — lives in
   "params": ["<id>"...],   // names bound in body, in type.params order
   "body": Term,
   "doc": "<optional string>",
-  "export": "<optional C symbol>", // omitted when absent (hash-stable when omitted); see §"Embedding ABI"
+  "export": "<optional C symbol>", // omitted when absent (hash-stable when omitted); embedding-ABI surface — see prose below
   "suppress": [Suppress...] // omitted when empty
 }
 
@@ -65,11 +65,11 @@ narrative — defaults, superclasses, diagnostics — lives in
   "drop-iterative": true   // opt-in; omitted when false (hash-stable when omitted)
 }
 
-// class (typeclass declaration; see Decision 11)
+// class (typeclass declaration; narrative in contracts/typeclasses.md)
 { "kind": "class",
   "name": "<id>",          // class name (e.g. "Show")
   "param": "<id>",          // single class parameter, kind *
-  "superclass": null,       // or { "class": "<id>", "type": "<param>" } — "class": canonical form (bare for same-module, "<module>.<Class>" for cross-module; see §"Class names" / mq.1)
+  "superclass": null,       // or { "class": "<id>", "type": "<param>" } — "class": canonical form (bare for same-module, "<module>.<Class>" for cross-module)
   "methods": [
     { "name": "<id>",
       "type": Type,         // FnSig over the class param
@@ -80,9 +80,9 @@ narrative — defaults, superclasses, diagnostics — lives in
   "doc": "<optional string>"
 }
 
-// instance (typeclass instance; see Decision 11)
+// instance (typeclass instance; narrative in contracts/typeclasses.md)
 { "kind": "instance",
-  "class": "<id>",          // class being instantiated; canonical form (bare for same-module, "<module>.<Class>" for cross-module; see §"Class names" / mq.1)
+  "class": "<id>",          // class being instantiated; canonical form (bare for same-module, "<module>.<Class>" for cross-module)
   "type": Type,             // concrete type expression (never the class param)
   "methods": [
     { "name": "<id>", "body": Term }
@@ -215,13 +215,14 @@ metadata is defined and gated there as well.
 ```jsonc
 // Type-constructor application. `args` omitted when empty
 // (hash-stable when omitted, for non-parameterised cases like Int, Bool, ...).
-{ "k": "con", "name": "<id>", "args": [Type...] }   // "name": canonical form (bare for same-module / primitives, "<module>.<TypeName>" for cross-module; see §"Type::Con name scoping" / ct.1)
+{ "k": "con", "name": "<id>", "args": [Type...] }   // "name": canonical form (bare for same-module / primitives, "<module>.<TypeName>" for cross-module)
 
-// Function type. Decision 10 added paramModes/retMode as
+// Function type. paramModes/retMode are metadata on Type::Fn —
-// metadata on Type::Fn — they are NOT separate Type variants, so every
+// they are NOT separate Type variants, so every existing match-arm
-// existing match-arm in the typechecker (unify, occurs, apply) keeps
+// in the typechecker (unify, occurs, apply) keeps working.
-// working. `paramModes` omitted when every entry is "implicit";
+// `paramModes` omitted when every entry is "implicit"; `retMode`
-// `retMode` omitted when "implicit" (hash-stable when omitted).
+// omitted when "implicit" (hash-stable when omitted). Full mode
+// contract lives in contracts/memory-model.md.
 { "k": "fn",
   "params":     [Type...],
   "paramModes": [ParamMode...],
@@ -232,14 +233,16 @@ metadata is defined and gated there as well.
 { "k": "var", "name": "<id>" }
 
 // Top-level polymorphism only. `constraints` carries class
-// constraints (Decision 11); omitted when empty (hash-stable when omitted).
+// constraints (narrative in contracts/typeclasses.md); omitted when
+// empty (hash-stable when omitted).
 { "k": "forall",
   "vars": ["<id>"...],
-  "constraints": [{ "class": "<id>", "type": "<id>" }, ...],   // "class": canonical form (bare for same-module, "<module>.<Class>" for cross-module; see §"Class names" / mq.1)
+  "constraints": [{ "class": "<id>", "type": "<id>" }, ...],   // "class": canonical form (bare for same-module, "<module>.<Class>" for cross-module)
   "body": Type }
 ```
 
-**`ParamMode`** (Decision 10):
+**`ParamMode`** (full contract in
+[memory model](memory-model.md)):
 
 ```
 "implicit"   — unannotated / back-compat. Treated as `own` by the typechecker.

design/contracts/feature-acceptance.md

@@ -36,16 +36,16 @@ A proposed feature ships only if all three hold:
    discouraged. A documentation note is not a reshape; the
    discriminator is whether the wrong code fails to typecheck, not
    whether a guideline advises against it. Worked example: a bare
-   `while` over mutable state would pass clauses 1 and 2 yet fail
+   `while` over mutable state would pass criteria 1 and 2 yet fail
-   clause 3 (it reinstates iterated-mutable-state reasoning the pure
+   criterion 3 (it reinstates iterated-mutable-state reasoning the
-   core exists to remove); a hypothetical "all repetition is either
+   pure core exists to remove); a hypothetical "all repetition is
-   structurally-decreasing recursion over an acyclic ADT or an
+   either structurally-decreasing recursion over an acyclic ADT or
-   explicit named loop" iteration story would, *if it could be built
+   an explicit named loop" iteration story would, *if it could be
-   without a documented-unenforced precondition*, pass all three —
+   built without a documented-unenforced precondition*, pass all
-   and the fact that the 2026-05 attempt could not (it forced a
+   three — and the fact that the 2026-05 attempt could not (it
-   silent-divergence precondition; see
+   forced a silent-divergence precondition; see
    `docs/specs/2026-05-16-iteration-discipline-revert.md`) is itself
-   the clause-3 mechanism working as intended.
+   the bug-class-reintroduction discriminator working as intended.
 
 This is the positive complement to the CLAUDE.md rule that
 implementation effort is not a rationale: cost is not a reason *for* a

design/contracts/float-semantics.md

@@ -56,8 +56,8 @@ no `f32` variant. The runtime / codegen contract:
   `0x7ff8000000000000` on one target and a different qNaN on
   another.
 - The textual rendering of NaN through `float_to_str` (the runtime
-  C helper that backs `instance Show Float` and, post-iter-rpe.1,
+  C helper that backs `instance Show Float` and every Float-typed
-  every Float-typed `print` call). The libc `printf("%g", nan)`
+  `print` call). The libc `printf("%g", nan)`
   glue used by `float_to_str` is permitted to emit `nan` / `-nan`
   / `NaN` etc. depending on libc version and the NaN's sign bit;
   AILang does not normalise this, since the prose / surface-print

design/contracts/honesty-rule.md

@@ -15,10 +15,10 @@ Two things never belong in a contract or model file:
 
 A cross-reference that does belong stays: it is a formal,
 file-relative Markdown link into the durable tier (`design/` or
-source), enforced by
+source), enforced by the body-link durability gate in
-[`design_index_pin.rs`](../../crates/ailang-core/tests/design_index_pin.rs)
+[`design_index_pin.rs`](../../crates/ailang-core/tests/design_index_pin.rs).
-clause-5. A reference that cannot be expressed as such a link is,
+A reference that cannot be expressed as such a link is, by that
-by that fact, the history-or-rationale prose the rule above removes.
+fact, the history-or-rationale prose the rule above removes.
 
 The single legitimate exception is a present-tense reserved or
 deliberately-excluded claim that is explicitly and correctly

design/contracts/language-constraints.md

@@ -4,8 +4,9 @@
 
 The four constraints below are necessary preconditions for RC to
 be sound and complete *without* a cycle-collector backstop. They
-are not new — AILang already satisfies all four — but Decision 10
+are not new — AILang already satisfies all four — but the
-makes them load-bearing rather than incidental:
+[memory model](memory-model.md) makes them load-bearing rather
+than incidental:
 
 1. **Strict evaluation.** Every `Term::App` (see
    [Data model](data-model.md)) argument is fully
@@ -33,8 +34,8 @@ These constraints are the precondition for the
 [memory model](memory-model.md): a sound RC implementation without
 a cycle-collector backstop requires the reference graph to be a
 DAG, which constraint 4 establishes from 1–3. They are also the
-load-bearing structural reason behind several
+load-bearing structural reason behind several rejections under the
-[feature-acceptance](feature-acceptance.md) clause-3 rejections
+[feature-acceptance](feature-acceptance.md) bug-class-reintroduction
-(notably iterated-mutable-state reasoning).
+discriminator (notably iterated-mutable-state reasoning).
 
 Ratified by: `crates/ailang-check/src/uniqueness.rs` (in-source mod tests).

design/contracts/memory-model.md

@@ -32,7 +32,8 @@ schema). The substantive reasons for per-position metadata over a
   `Type` would let the schema express forms like
   `(con List (borrow Int))` — syntactically possible, semantically
   meaningless (you cannot separately own/borrow a list element
-  from the list it lives in). Decision 1 is "schema = data,
+  from the list it lives in). The
+  [canonical-schema principle](data-model.md) is "schema = data,
   schema permits exactly what is meaningful"; per-position
   metadata is the option that holds that line.
 - **Compositional clarity.** A `Type` value's identity should
@@ -62,7 +63,7 @@ identical: `param_modes` is skipped when every entry is
 `Implicit`, `ret_mode` is skipped when `Implicit`. Existing
 modules emit the same bytes as before.
 
-**Type::Con name scoping (canonical form, since ct.1).** Within a
+**Type::Con name scoping (canonical form).** Within a
 `.ail.json`, a `Type::Con.name` is interpreted relative to the
 file's top-level `"name"` field (the owning module). Bare names
 (no `.`) refer to a TypeDef in the owning module's own `defs`.
@@ -75,15 +76,16 @@ qualified references whose owner is unknown are also a violation
 (`WorkspaceLoadError::BadCrossModuleTypeRef`). The same rule
 applies to `Term::Ctor.type_name`.
 
-Class names follow the canonical-form rule (mq.1): bare for
+Class names follow the same canonical-form rule: bare for
 same-module references, `<module>.<Class>` for cross-module
-references — symmetric to `Type::Con.name`'s rule from ct.1.
+references — symmetric to `Type::Con.name`'s rule above.
 Three schema fields carry class references in this form:
 `InstanceDef.class`, `Constraint.class`, and `SuperclassRef.class`.
 `ClassDef.name` itself stays bare (defining-site context, like
 `TypeDef.name`).
 
-Method dispatch is type-driven post-mq.3 (see [Method dispatch](method-dispatch.md)):
+Method dispatch is type-driven (see
+[Method dispatch](method-dispatch.md)):
 synth resolves a `Term::Var { name: "show" }` by consulting
 the workspace's method-to-candidate-class index, filtering by
 argument type (concrete) or by declared constraint (rigid-var), and
@@ -161,24 +163,24 @@ Skipped from serialisation when empty so existing fixtures keep
 bit-identical canonical-JSON hashes (regression-pinned by
 `iter19b_empty_suppress_preserves_pre_19b_hashes` and
 `iter19b_schema_extension_preserves_pre_19b_hashes`).
-The canonical-form tightening in ct.1 shifted the hashes of two
+The canonical-form tightening for `Type::Con.name` shifted the
-cross-module fixtures (`ordering_match.ail.json` and
+hashes of two cross-module fixtures (`ordering_match.ail.json` and
 `test_22b1_dup_a.ail.json`); all intra-module fixtures, including
 the regression-pinned `sum.ail.json` and `list.ail.json`, remain
 bit-identical. The new pins are
 `ct4_migrated_fixtures_have_canonical_form_hashes` (locks the
 post-migration hashes) and
 `ct4_unmigrated_fixtures_remain_bit_identical` (re-asserts the
-existing 13a/19b/22b.1 hashes still hold).
+pre-tightening hashes still hold).
 
 ## Advisory diagnostics
 
 The advisory-diagnostics arc introduces the language's first
 **advisory** typechecker diagnostic and the suppression mechanism
-that goes with it. Decision 10's mandatory-annotation rule is
+that goes with it. The mandatory-annotation rule of this memory
-unchanged: `param_modes` and `ret_mode` remain author-required;
+model is unchanged: `param_modes` and `ret_mode` remain
-the typechecker does not infer them. What's new is feedback when
+author-required; the typechecker does not infer them. What's new
-an authored annotation is *stricter than necessary*.
+is feedback when an authored annotation is *stricter than necessary*.
 
 **The lint: `over-strict-mode`.** Fires on a
 fn-param `p` annotated `(own T)` when:

design/contracts/method-dispatch.md

@@ -6,14 +6,16 @@ The class-method resolution rule that pairs with the
 [typeclasses](typeclasses.md) contract; the `Show`/`Eq`/`Ord` instances
 the rule dispatches against live in [prelude classes](prelude-classes.md).
 
-Post-mq.3, dispatch is two-mode:
+Dispatch is two-mode:
 
 **Polymorphic call sites** (inside a fn body with `forall` +
 constraint set): the constraint names the class via the qualified
-`Constraint.class` field (canonical-form per mq.1). Synth's
+`Constraint.class` field (canonical form: bare for same-module,
-residual carries the class name directly; constraint-discharge at
+`<module>.<Class>` for cross-module; see
-fn-body-end matches against the workspace registry by
+[memory model](memory-model.md) for the canonical-form rule).
-`(class, type_hash)` key.
+Synth's residual carries the class name directly;
+constraint-discharge at fn-body-end matches against the workspace
+registry by `(class, type_hash)` key.
 
 **Monomorphic call sites**: synth consults the workspace-flat
 `Env.method_to_candidate_classes: BTreeMap<MethodName,
@@ -42,7 +44,7 @@ dispatch rule:
 The `method_to_candidate_classes` index is the load-bearing data
 structure for this routing — its construction in `build_check_env`
 inverts the per-module `class_methods` maps (themselves tuple-keyed
-by `(qualified-class, method)` post-mq.3) to a workspace-flat
+by `(qualified-class, method)`) to a workspace-flat
 method-name-to-class-set map.
 
 Class-fn collisions resolve at the call site, not at workspace load

design/contracts/prelude-classes.md

@@ -24,7 +24,8 @@ corresponding runtime primitive (`int_to_str`, `bool_to_str`,
 heap-Str primitives); no codegen intercept is required. The
 polymorphic helper `print : forall a. Show a => a -> () !IO` has body
 `\x -> let s = show x in do io/print_str s` (explicit let-binder for
-heap-Str RC discipline per eob.1 Str carve-out). The let-binder is
+heap-Str RC discipline per the Str carve-out in
+[Str ABI](str-abi.md)). The let-binder is
 structurally pinned by `crates/ail/tests/print_mono_body_shape.rs`.
 Routing through `print` is the path for non-`Str` primitives;
 `io/print_str` is the only built-in direct-output effect-op.

design/contracts/roundtrip-invariant.md

@@ -85,15 +85,15 @@ never by relaxing the test.
 ### Why this is anchored at top level
 
 The invariant is a property of the language identity, not of any
-one surface-design Decision. Decision 6 introduces the `.ail`
+one surface-design choice. The [authoring surface](authoring-surface.md)
-surface and lists round-trip-as-property as one of its
+introduces the `.ail` surface and lists round-trip-as-property as
-constraints, but the property is load-bearing for every
+one of its constraints, but the property is load-bearing for every
 downstream concern that treats the two forms as exchangeable:
 content-addressed hashing, the LLM-author's choice of authoring
 form, the integrity of fixture cross-references, and the
 prerequisite for empirical cross-model authoring-form studies.
-Lifting it out of Decision 6 makes the property quotable on its
+Lifting it out of the authoring-surface contract makes the property
-own and reviewable by the architect agent at every milestone
+quotable on its own and reviewable by the architect agent at every
-close.
+milestone close.
 
 Ratified by: `crates/ailang-surface/tests/round_trip.rs`.

design/contracts/scope-boundaries.md

@@ -118,7 +118,8 @@ What **is** supported (and used as the smoke test for the pipeline):
   pair allocates either via `@GC_malloc` (escaping; Boehm-managed) or
   via LLVM `alloca` (non-escaping; freed at fn return). The decision
   is made by an escape-analysis pre-pass over the fn body — see
-  Decision 9's "Per-fn arena via stack `alloca`" subsection.
+  the "Per-fn arena via stack `alloca`" subsection of
+  [RC + uniqueness](../models/rc-uniqueness.md).
   Boehm-only soak tests are unchanged: `examples/gc_stress.ail.json`
   and `examples/std_list_stress.ail.json` still allocate via
   `@GC_malloc` because their boxes flow into other fns and escape.

design/contracts/tail-calls.md

@@ -1,6 +1,6 @@
 # Tail calls
 
-## Decision 8: explicit, verified tail calls
+## Explicit, verified tail calls
 
 For an LLM author, recursion is the natural iteration form
 (`\n. if n == 0 then () else loop(n-1)` is what I reach for, not

design/contracts/typeclasses.md

@@ -74,15 +74,18 @@ exactly as for any free function.
 
 ## Cross-module references in synthesised bodies
 
-The unified mono pass (per Decision 11's milestone-23.4 reorganisation)
+The unified mono pass (see
-synthesises mono symbols for polymorphic free fns and class-method
+[models/typeclasses](../models/typeclasses.md) for the
-instances in the symbol's owner module — e.g. `print__Int` lives in
+resolution/monomorphisation algorithm) synthesises mono symbols for
-`prelude` (because `print` is defined in the prelude), but a
+polymorphic free fns and class-method instances in the symbol's
-user-ADT call site `print (MkIntBox 7)` causes synthesis of
+owner module — e.g. `print__Int` lives in `prelude` (because `print`
-`prelude.print__IntBox` whose body references
+is defined in the prelude), but a user-ADT call site
-`show_user_adt.show__IntBox` (the user instance lives in the
+`print (MkIntBox 7)` causes synthesis of `prelude.print__IntBox`
-user-defining module per Decision 11 coherence). The synthesised body
+whose body references `show_user_adt.show__IntBox` (the user
-crosses a module boundary the source template did not.
+instance lives in the user-defining module per the
+orphan-free-coherence rule documented in
+[models/typeclasses](../models/typeclasses.md)). The synthesised
+body crosses a module boundary the source template did not.
 
 Three invariants make this work:
 
@@ -127,7 +130,7 @@ Three invariants make this work:
    workspace registry; if no instance satisfies the residual at the
    unified concrete type, the `NoInstance` diagnostic fires at
    typecheck (correctly), not at codegen (confusingly). Without the
-   residual push, milestone-23-shape negative cases (e.g. `print f`
+   residual push, polymorphic-helper negative cases (e.g. `print f`
    where `f : Int -> Int`) silently typecheck and surface as
    `unknown variable: show` from codegen instead of the right
    typecheck-phase NoInstance Show.
@@ -211,17 +214,17 @@ module.
 - `MissingConstraint` — body's residual constraint is not covered
   by declared (and superclass-expanded) constraints.
 - `NoInstance` — fully concrete constraint has no registry entry.
-  mq.2 adds an optional `candidate_classes` field surfacing the
+  Carries an optional `candidate_classes` field surfacing the
   multi-candidate set when the bare-method dispatch path's filter
   collapses to zero registry survivors.
-- `AmbiguousMethodResolution` (mq.2) — a monomorphic `Term::Var`
+- `AmbiguousMethodResolution` — a monomorphic `Term::Var`
   call site survives both type-driven and constraint-driven filters
   with more than one candidate class. LLM-author writes the explicit
   qualifier form `<module>.<Class>.<method>` to disambiguate.
-- `UnknownClass` (mq.2) — an explicit class qualifier in
+- `UnknownClass` — an explicit class qualifier in
   `Term::Var.name` names a qualified class that is not in the
   workspace's candidate-class index for the method.
-- `class-method-shadowed-by-fn` (mq.3, warning) — a `Term::Var`
+- `class-method-shadowed-by-fn` (warning) — a `Term::Var`
   resolved via fn lookup precedence (locals → caller-module-fn →
   imported-fn) while a class method of the same name also exists
   in the workspace. Fn resolution proceeds; the warning surfaces

design/models/effects.md

@@ -1,6 +1,6 @@
 # Effects — pure core + algebraic effects
 
-## Decision 3: pure core language + algebraic effects
+## Pure core language + algebraic effects
 
 The default is total, pure functions. Effects are declared as a set in the
 function type: `(Int) -> Int ![IO]`. The effect set is a
@@ -13,7 +13,9 @@ currently in scope vs. excluded lives in
 In the MVP only the effect `IO` is wired up (its sole op is `io/print_str`).
 `Diverge` (for non-termination) is reserved as an effect name but is
 unimplemented — no op, no codegen, no checker injection — in the same
-sense Decision 4 reserves refinements.
+sense the refinements layer is reserved as a future extension without
+present-day surface (see
+[scope boundaries](../contracts/scope-boundaries.md)).
 
 This is the most important LLM property: when I read a function, I can trust
 its signature without reading the body.

design/models/pipeline.md

@@ -7,7 +7,7 @@
             ├─ load + validate schema
             ├─ resolve names + assign hashes
             ├─ desugar (AST → AST)
-            ├─ typecheck (HM, effect rows; mode-strict per Decision 10)
+            ├─ typecheck (HM, effect rows; mode-strict per the memory model)
             ├─ lift_letrecs (post-typecheck AST → AST)
             ├─ lower to MIR (SSA-like, named SSA values)
             ├─ emit LLVM IR (.ll)

design/models/prose-projection.md

@@ -5,9 +5,9 @@
 AILang ships a
 second textual projection of the AST: `ailang-prose`, a one-way
 projection from `Module → human-readable text`. It is **not** an
-authoring surface; it is the "display" projection that Decision 6's
+authoring surface; it is the "display" projection that the
-architectural pin (see [authoring surface](../contracts/authoring-surface.md))
+[authoring surface](../contracts/authoring-surface.md)'s
-explicitly anticipated:
+architectural pin explicitly anticipated:
 
 > *"Future projections are explicitly anticipated: a visual /
 > graphical front-end is a plausible second projection for human
@@ -41,7 +41,8 @@ form (B) deliberately is not. Re-integrating prose edits requires an
 external LLM mediator, not a compiler pass — the prompt template
 `ail merge-prose` composes the six-step cycle.
 
-Form (B) does not weaken any Decision 6 invariant:
+Form (B) does not weaken any
+[authoring surface](../contracts/authoring-surface.md) invariant:
 
 - The [JSON-AST](../contracts/data-model.md) remains the only
   hashable artefact. Prose is not hashed, not content-addressed,

design/models/rc-uniqueness.md

@@ -1,6 +1,6 @@
 # RC + Uniqueness — memory model whitepaper
 
-## Decision 9: dual allocator — RC canonical, Boehm parity oracle
+## Dual allocator — RC canonical, Boehm parity oracle
 
 AILang ships two allocator backends with an asymmetric role:
 
@@ -25,12 +25,11 @@ when the parity oracle stops paying its keep — concretely, when a
 few iter families ship without the gc arm catching anything that
 the rc arm did not already catch. Until then, the cost of
 keeping libgc as a build dependency is accepted in exchange for
-diagnostic leverage. Decision 10 (RC + uniqueness) holds as the
+diagnostic leverage. The
-specification of the canonical runtime (the contract version is
+[RC + uniqueness memory model](../contracts/memory-model.md), built
-[memory model](../contracts/memory-model.md), built on
+on the [language constraints](../contracts/language-constraints.md),
-[language constraints](../contracts/language-constraints.md));
+holds as the specification of the canonical runtime; the rest of
-the rest of this section documents the Boehm half, retained as
+this section documents the Boehm half, retained as the oracle.
-the oracle.
 
 **Choice: Boehm-Demers-Weiser conservative GC.** The simplest
 working option:
@@ -130,7 +129,7 @@ The closure-pair and its env share an escape verdict — they have
 parallel lifetimes. If the closure pair is non-escaping, the env
 is too.
 
-## Decision 10: memory model — RC + Uniqueness with LLM-author annotations
+## Memory model — RC + Uniqueness with LLM-author annotations
 
 **The GC bench (`bench/run.sh`) showed
 Boehm contributing a substantial fraction of runtime on allocation-heavy workloads
@@ -150,9 +149,10 @@ stdlib is the cheapest moment to commit.
 **Choice.** AILang's canonical [memory model](../contracts/memory-model.md)
 is reference counting with static uniqueness inference **and
 explicit LLM-author annotations on fn signatures**, in the lineage
-of Lean 4 / Roc / Koka. Boehm becomes a transitional allocator (Decision 9) and is
+of Lean 4 / Roc / Koka. Boehm becomes a transitional allocator (see
-retired when the RC pipeline matches the bump-allocator floor
+the dual-allocator section above) and is retired when the RC pipeline
-within an acceptable margin (target 1.3× on `bench/run.sh`).
+matches the bump-allocator floor within an acceptable margin (target
+1.3× on `bench/run.sh`).
 
 **Workload scope of the 1.3× target.** The 1.3× target was
 calibrated on the original `bench/run.sh` corpus: linear list
@@ -168,16 +168,17 @@ one bump-pointer bump, doubling the allocation tax on closure
 construction (current ratio recorded in JOURNAL bench entries).
 This is a representational cost of the closure-pair layout,
 not a defect in the RC implementation; a future slab/pool
-allocator for fixed-shape pair cells (Decision 9 retirement
+allocator for fixed-shape pair cells (a Boehm-retirement follow-up)
-follow-up) would compress this ratio without changing semantics.
+would compress this ratio without changing semantics.
 
 The 1.3× retirement target therefore applies to the linear /
 tree / poly-ADT subset of the corpus. Closure-heavy workloads are
 tracked under a wider band (the closure-chain baseline records its rc/bump ratio as the `rc_over_bump` reference value with ±15% tolerance) and
 are explicitly excluded from the Boehm-retirement gate until a
-slab/pool answer ships. Decision-10's RC commitment is unchanged;
+slab/pool answer ships. The
-what is scoped is the *quantitative* retirement criterion, not
+[memory model](../contracts/memory-model.md)'s RC commitment is
-the choice of memory model.
+unchanged; what is scoped is the *quantitative* retirement criterion,
+not the choice of memory model.
 
 The architecture has two layers:
 

design/models/typeclasses.md

@@ -8,7 +8,7 @@ built-in classes shipped in the prelude in
 covers the design choices and the resolution / monomorphisation
 algorithm.
 
-## Decision 11: typeclasses — Haskell-lite, monomorphised, coherent
+## The design — Haskell-lite, monomorphised, coherent
 
 **The design pass for typeclasses. Codified after the
 [Feature-acceptance criterion](../contracts/feature-acceptance.md)
@@ -158,10 +158,9 @@ component contains `__` by project convention.
 
 **No runtime dispatch, no dictionary passing.** The monomorphisation
 pass is the ONLY specialiser. Codegen sees only monomorphic
-`Def::Fn`s and direct calls; the pre-iter-23.4 codegen-time
+`Def::Fn`s and direct calls. A call that cannot be monomorphised —
-specialiser (`lower_polymorphic_call` + `module_polymorphic_fns` +
+for instance, because a constraint remains unresolved at the entry
-`mono_queue`) was removed in iter 23.4. A call that cannot be
+point — is a static error, not a runtime one. This is the
-monomorphised — for instance, because a constraint remains
+LLVM-friendly form and is consistent with the performance
-unresolved at the entry point — is a static error, not a runtime
+commitment of the
-one. This is the LLVM-friendly form and is consistent with
+[RC + uniqueness memory model](../contracts/memory-model.md).
-Decision 10's performance commitment.