STB-Core v0.2

Reference implementation of Structural Generative Transmission (STB) — a bio-inspired framework for transmitting generative instructions instead of raw data.

⚠️ Experimental. Not independently audited. Do not use in production.

Preprint

Matos Júnior, D. (2026). Structural Generative Transmission (STB). Zenodo. https://doi.org/10.5281/zenodo.20528475

What this implements

Layer	Component	Status
C1	Purification (Ψ scoring)	✅
C2	Gene library + fitness metric	✅
C3	Metalanguage AST + bytecode compiler	✅
C4	Digital capsid (SHA256 checksum)	✅
L5	STB Virtual Machine	✅
—	Residual compression (zstd)	✅ optional

What is not implemented: automatic gene discovery, domain-specific calibration of Ψ coefficients, empirical benchmarking.

Build

cargo build
cargo test

With zstd residual compression:

cargo build --features zstd

Quick example

use stb_core::{Instruction, Compiler, STBVM, GeneLibrary};

let ast = Instruction::Seq(vec![
    Instruction::Def {
        id: "msg".to_string(),
        data: b"Hello STB!".to_vec(),
    },
    Instruction::Rpt {
        count: 3,
        body: Box::new(Instruction::Ref { id: "msg".to_string() }),
    },
    Instruction::Stp,
]);

let bytecode = Compiler::compile(&ast).unwrap();
let mut vm = STBVM::new(GeneLibrary::new());
let output = vm.execute_bytecode(&bytecode).unwrap();

Security limits

Limit	Value
Max output size	64 MB
Max loop iterations	1,000,000
Gene signature validation	SHA256
Bytecode integrity	SHA256 (capsid)

Development note

STB-Core was developed collaboratively using AI-assisted code generation as part of the research process described in the preprint above.

License

MIT