110K Tests. 14 Targets. 207 Proofs. The Hard Numbers.

No Marketing Fluff

This post is pure data. No adjectives. No hand-waving. Every number is reproducible. Every test is in the repository. Clone it. Run it. Verify it yourself.

Test Suite: 110,227 Tests

Level 1 — Unit tests: Individual monomer verification. Each of the 128 monomers (64 core + 64 extended) tested with boundary values, zero, max, overflow. Every atomic operation, verified in isolation.

Level 2 — Integration tests: EVA composition chains. Sequential, parallel, conditional operators. Type flow verification across compositions. The places where most compilers silently break.

Level 3 — Exhaustive monomer tests: 25,000+ tests. Every monomer tested with every valid input in its domain. Not sampling. Not statistical coverage. Exhaustion. Every single value.

Level 4 — Cross-backend tests: 50,000+ tests. Same PCD compiled to Rust, JS, Python, C, Go, WASM, BIR. Outputs compared bit for bit. 100% consistency across all backends.

Level 5 — Self-compilation tests: The compiler compiles itself. Gen0 output must equal Gen1 output, byte for byte. True fixpoint. The ultimate proof that the compiler is correct.

Level 6 — Lifter tests: 515 tests across 10 input languages (JS, TS, Python, Rust, C, C++, Go, COBOL, PHP, Java). Every language path verified.

Level 7 — Adversarial tests: Fuzzer-generated edge cases, malformed inputs, deliberately pathological programs designed to crash the compiler. None succeeded.

Compilation Targets: 14

Rust, JavaScript, TypeScript, Python, C, C++, Go, COBOL, PHP, Java, Swift, WASM, native x86-64, BIR bytecode. One PCD source. Fourteen verified outputs.

Every target produces identical outputs for identical inputs. 100% cross-target consistency, verified by over 50,000 tests. Write once, compile everywhere, get the same answer everywhere.

Formal Proofs: 207 Files, 0 Unproven Assumptions

Every core monomer has a machine-checked mathematical proof. 207 proof files. Zero unproven assumptions — every single property is formally verified. The proofs cover: signature correctness, domain preservation, composition laws, and termination guarantees. Not tested. Proven.

Input Languages: 10

The Lifter reverse-compiles from JavaScript, TypeScript (including TSX/JSX), Python, Rust, C, C++, Go, COBOL, PHP, and Java into PCD. 515 test cases across all languages. Your existing code, in whatever language you wrote it, can be lifted and certified.

Self-Compilation

The compiler (brikc) is written in PCD and compiles itself. The output is a native x86-64 ELF binary. Gen0 (Rust bootstrap) compiles brikc.pcd to produce Gen1 (native). Gen1 compiles brikc.pcd to produce Gen2. Gen1 === Gen2, byte for byte. True fixpoint. This is the ultimate test of a compiler: it trusts itself enough to build itself, and the output is identical. That is not a test. That is a mathematical proof of consistency.