Execution stack responsibilities

High-level structure

1. Parsing (AST)
2. HIR (semantic IR)
3. CLIF lowering
4. Module layer
5. Execution backend (JIT thin / AOT primary)
6. Runtime/stdlib

Suggested responsibilities

beskid_analysis

• Parser + AST + source spans
• HIR lowering + type checking
• Semantic analysis rules + diagnostics

beskid_codegen

• HIR -> CLIF lowering
• FunctionBuilder integration
• Module abstraction (common for JIT/AOT)

beskid_runtime

• Builtins + allocation
• String/array primitives
• Panic/error handling

beskid_engine

• JIT thin driver / execution API (development-time)

beskid_aot

• AOT object emission and linker orchestration (production path)
• Runtime archive preparation and ABI/version validation

Artifact handoff boundaries

1. beskid_analysis outputs resolved, typed semantic data.
2. beskid_codegen consumes semantic data and outputs lowered backend artifact(s).
3. beskid_engine consumes lowered artifacts for in-memory execution only.
4. beskid_aot consumes lowered artifacts for object emission and final linkage.
5. beskid_runtime implements runtime ABI entrypoints used by both backends.

Forbidden coupling

• beskid_analysis must not depend on runtime or linker behavior.
• beskid_codegen must not encode backend-specific syscall policy.
• beskid_engine must not define production packaging behavior.
• beskid_aot must not redefine runtime ABI semantics.
• beskid_runtime must not own language-level parsing/type semantics.

Why this split

• Keeps analysis and execution concerns isolated.
• Avoids mixing runtime with analysis.
• Keeps JIT and AOT as backend variants over the same lowering + runtime ABI.
• Ensures platform-specific execution policy lives in runtime, not in backend drivers.

Expansion plan

• Start with minimal HIR + CLIF lowering for functions and literals.
• Introduce runtime builtins for strings/arrays.
• Keep JIT narrow as a fast feedback harness.
• Grow production capability through AOT and runtime ABI hardening.