subtopics
Hyperon Experimental
Hyperon-Experimental is the original reference implementation of MeTTa, serving as the “master blueprint” for the language and the primary engine for R&D. Built in Rust, it is designed for maximum extensibility.
A notable characteristic is its deep Python integration, which enables a hybrid development model where MeTTa and Python code can interoperate seamlessly within the same application. This provides the leverage of the entire Python ecosystem, including its vast AI, data science, and machine learning libraries, directly within MeTTa’s symbolic reasoning framework.
Furthermore, Hyperon-Experimental is engineered as an extensible library with a C API, allowing it to be integrated with programs written in other languages like C or C++. While this architecture is robust and forward-looking, it intentionally prioritizes flexibility and semantic correctness over raw execution speed. As a result, it has merit for conducting small experiments but does not, at present, provide production-grade performance.
Repositories
Papers & Publications
- Potapov A., Bogdanov V. Univalent foundations of AGI are (not) all you need // Springer: LNCS, V.13154 (proc. AGI'21). 2022. P. 184–195.
- Warrell J., Potapov A., Vandervorst A., Goertzel B. A Meta-Probabilistic-Programming Language for Bisimulation of Probabilistic and Non-Well-Founded Type Systems // Springer: LNCS, V.13539 (proc. AGI'22). 2023. P. 434–451.
- Add capability to integrate various expression evaluation mechanisms
- Integration of Prolog VM-based modules
- Release Python packages for Windows
- Address inefficient variable bindings for performance improvement
Continue Reading
PeTTa
PeTTa is a high-performance compiler and runtime for the MeTTa language, designed to execute complex symbolic AI code at speeds required for real-time applications like robotics and large-scale reasoning. It achieves this by translating MeTTa source code directly into highly optimized Prolog.
Its core innovation is a “Smart Dispatch” compiler, which intelligently solves the key challenge of deciding whether a piece of MeTTa code is a function to be executed or a piece of data to be structured. By eliminating the slow, “check-at-runtime” (dynamic dispatch) methods used by typical interpreters, PeTTa generates code that achieves execution speeds comparable to handwritten, idiomatic Prolog.
Crucially, it fully adheres to the Hyperon-Experimental (HE) semantics, ensuring it is a correct and compatible implementation while providing a massive performance boost. It is also fully interoperable with high-performance backends, capable of manipulating MORK spaces and executing MM2 expressions directly from MeTTa code.
This makes PeTTa an essential component for running computationally intensive symbolic architectures—like MeTTa-NARS and PLN—in production, bridging the gap from research-grade interpretation to real-world, high-speed deployment.
Repositories
Documentation
Continue Reading
MeTTaLog (Legacy)
MeTTaLog, developed by Douglas Miles and LogicMoo, served as a vital intermediate engine grounded in the Warren Abstract Machine (WAM)—the long-standing standard for logic programming. It successfully validated the potential for concurrent symbolic execution, yet the unique demands of neural-symbolic interoperability eventually exposed the rigidities of a pure WAM approach when scaling for AGI.
This inspired a strategic pivot to PeTTa, which derives its architecture from the ZIP Virtual Machine used by SWI-Prolog. The ZIP model offers superior handling of the dynamic, non-deterministic workloads essential to the Hyperon ecosystem. Consequently, PeTTa’s ZIP-inspired foundation was selected as the optimal path to secure the high-performance integration required for future AGI capabilities.
Papers & Publications
Extends the MeTTa language with concurrent operations and runtime answer pruning, enabling multi-threaded evaluation, dynamic clause filtering, and adaptive control over nondeterministic result sets during inference:
Continue Reading
JeTTa
JeTTa is an experimental MeTTa compiler developed for the Java Virtual Machine (JVM) using Kotlin. It leverages the JVM’s mature architecture to support high-performance multithreading and custom Space implementations. By utilizing extensive program transformations for optimization, JeTTa offers a pathway to deploy Hyperon within scalable, multiprocessor enterprise environments.
Continue Reading
MeTTa-Morph
MeTTa-Morph is a macro-based translator from MeTTa to (Chicken) Scheme, designed to significantly accelerate execution of a practical subset of the MeTTa language. Rather than supporting full language coverage, it focuses on compiling a performant core of MeTTa constructs into an efficient Scheme library, achieving speedups on the order of 100x.
MeTTa-Morph transforms MeTTa code into Scheme using macros corresponding to core MeTTa constructs, including function definitions, variable binding, query execution, pattern matching, type-related constructs, and space operations such as collapse and superposition. The generated Scheme code can be compiled to C and ultimately to native machine code.
Continue Reading