TECH
papers.

In-depth technical articles, architectural decisions, and system deep-dives.

Akron, Artik, Lysis: Why Achronyme Composes Three Virtual Machines Instead of One

This article continues the architectural analysis of Achronyme's VM published in March, this time explaining how and why the project evolved from a single register-based virtual machine to a composition of three specialized VMs: Akron (scripting + prove blocks with heap and tri-color GC), Artik (deterministic witness generation with no heap and no GC), and Lysis (an SSA bytecode walked by the constraint frontend, with a single-static-store discipline). Each was born from an invariant the previous one couldn't sustain, and each one's memory discipline is exactly aligned with the question that VM exists to answer. The article argues that composing small machines with precise invariants yields less code and more guarantees than a unified VM that tries to absorb every case.

Advanced
Read paper

ProveIR: Compile Once, Prove Many. Inside Achronyme's Circuit Template System

This article explores ProveIR, the intermediate representation that powers Achronyme's prove blocks. ProveIR is a parametric circuit template system: prove blocks are compiled once at compile time, serialized into the bytecode constant pool, and instantiated at runtime with captured values from the surrounding scope. The article deconstructs the full pipeline (from AST to ProveIR template, capture classification, serialization, instantiation into SSA IR, optimization, and finally R1CS or Plonkish constraint generation), explaining why this intermediate layer is essential for correctness, performance, and portability across prime fields.

Advanced
Read paper

From AST to Arithmetic Constraints: How Achronyme Compiles Code into Zero-Knowledge Proofs

This paper provides an exhaustive technical analysis of the Achronyme zero-knowledge compilation pipeline. Focusing on the circuit compilation path, it deconstructs how high-level constructs are lowered into a phi-node-free SSA intermediate representation, optimized through constant folding, dead code elimination, boolean propagation, and taint analysis, then translated into either R1CS or Plonkish arithmetizations. The analysis bridges the gap between traditional compiler design and the strict mathematical realities of cryptographic circuits.

Advanced
Read paper

Anatomy of a Virtual Machine: From Stack to Registers in Achronyme

This paper explores the architectural foundations of Virtual Machines, analyzing the structural differences between Stack-based and Register-based models. Through the case study of the Achronyme language, it demonstrates how the inherent bottleneck of Stack VMs in cryptographic operations was mitigated by adopting a Register-based VM (inspired by RISC, Lua 5.0, and Dalvik), significantly reducing the dispatch loop overhead and optimizing cache locality.

Advanced
Read paper