Skip to main content

AILANG Design Axioms

AILANG is defined less by individual features than by a small set of axioms. These axioms constrain the design space and explain why certain features exist — and why others are deliberately absent.

They are intended to be stable over time, even as syntax and implementation evolve.

Axiom 1: Determinism Is a Semantic Invariant​

If a program's execution cannot be replayed, it cannot be trusted.

AILANG programs must have deterministic meaning by default. Any source of nondeterminism (time, randomness, IO, scheduling) must be:

  • explicit
  • typed
  • localized
  • observable in traces

Determinism is not an optimization or a runtime promise; it is part of the language's semantics.

See: Philosophical Foundations for the deeper motivation.

Axiom 2: Execution Must Be Replayable and Auditable​

Execution is a first-class artifact.

AILANG treats execution traces as:

  • inspectable
  • serializable
  • replayable
  • suitable for verification and debugging

A program that cannot explain why it produced an outcome is incomplete.

Axiom 3: Effects Are Real and Must Be Legible​

Side effects are not a failure of abstraction; they are a property of reality.

AILANG does not hide effects. It makes them explicit, typed, and machine-readable.

Effects define what a computation is allowed to do, not just what it computes.

See: Effects Reference for implementation details.

Axiom 4: Authority Must Be Explicit​

No program should have implicit access to the world.

AILANG requires that authority — to read data, write data, consume resources, or interact with external systems — is:

  • explicitly declared
  • statically visible
  • constrained by capability and budget

Ambient authority is treated as a design error.

Axiom 5: Verification Should Be Local, Bounded, and Automatable​

Verification is most useful when it is cheap and frequent.

AILANG favors:

  • local properties over global proofs
  • bounded checks over unbounded reasoning
  • machine-driven verification over human ceremony

Verification is a mode of execution, not a separate discipline.

Axiom 6: Concurrency Must Not Destroy Meaning​

Parallelism must not introduce ambiguity.

AILANG rejects concurrency models where program meaning depends on:

  • scheduling
  • timing races
  • hidden shared state

Concurrency is permitted only when its semantics remain deterministic and replayable.

Axiom 7: Machines Are Primary Readers​

AILANG assumes that machines are the primary consumers of code.

The language prioritizes:

  • semantic compression
  • decidable structure
  • toolability and analysis
  • clarity of intent over stylistic expressiveness

Human readability is preserved, but never at the expense of machine reasoning.

See: Why AILANG? for the AI-first motivation.

Axiom 8: Syntax Is a Liability​

Every syntactic feature increases the semantic surface area.

AILANG adopts syntax only when it:

  • reduces ambiguity
  • improves analyzability
  • lowers total system complexity

Expressiveness without semantic payoff is rejected.

Axiom 9: Cost Is Part of Meaning​

Computation has cost.

AILANG treats cost — in time, memory, IO, or external calls — as part of a program's semantics, not an incidental runtime detail.

Programs should make their resource implications visible to tools and agents.

See: Design Lineage § Capability Budgets.

Axiom 10: The Language Must Compose With Itself​

Features must compose without creating semantic blind spots.

AILANG rejects features that:

  • break reasoning when combined
  • require "special cases" in the mental model
  • behave differently under analysis than under execution

Composability is a correctness property.

Axiom 11: Failure Must Be Representable​

Failure is not exceptional; it is informative.

AILANG treats failure as:

  • structured
  • typed
  • inspectable
  • composable

Unstructured exceptions are avoided.

Axiom 12: The Language Is a System Boundary​

AILANG is not just syntax; it is a boundary between:

  • intent and execution
  • authority and action
  • possibility and permission

Crossing that boundary must always be explicit.

Closing Note​

These axioms are intentionally constraining.

They exist to make AILANG:

  • predictable for machines
  • inspectable for humans
  • safe for autonomous execution
  • suitable for long-lived, evolving systems

Any feature proposal that violates an axiom must justify why the axiom itself should change.

See Also​