Typical Computing

Typical Programming System

Engineering Guarantees For AI-Written Software.

Typical is a programming language + compiler + environment for codebases where quality is critical and AI is the author.

fn distanceSquared x is int · y is int is int
xSquared=squarex
ySquared=squarey
returnxSquared+ySquared
Typical source.

The Problem

Plausible output is not an engineering guarantee.

Models can produce convincing code. But some systems need stronger evidence that the code is correct, secure, and consistent with what was requested.

Current process

  1. Request
  2. Generated code
  3. Tests
  4. Review
  5. Hope

Typical

  1. Request
  2. Constraints
  3. Compiler
  4. Evidence

The Approach

Put guarantees beneath the model.

Better models and agent setups help. Typical moves part of the reliability problem into the language, compiler, and environment—where constraints can be checked deterministically.

01

Claims

Keep requested intent connected to the implementation.

02

Proofs

Make selected guarantees compiler-enforced.

03

Sound types

Prevent invalid states from hiding behind escape hatches.

Familiar Above. Rust Below.

Familiar APIs. Native foundations.

Typical’s standard library follows familiar JavaScript and Node.js APIs. Typical programs compile to readable Rust and remain compatible with the Rust ecosystem.

fn distance_squared(x: i64, y: i64) -> i64 {
	let x_squared = square(x);
	let y_squared = square(y);

	x_squared + y_squared
}
fndistanceSquaredx is int · y is intisint
xSquared=squarex
ySquared=squarey
returnxSquared+ySquared

Drag to compare. No JavaScript runtime underneath.

Why This Project

Two decades of code-interface and language design.

Built by Paul Young, creator of Stylizer—an early visual CSS editor used by developers worldwide. Later work explored programming languages, type systems, and ontology verification. Typical brings those threads together.

StylizerTruthTypical
Documentation GitHub Follow on X