A research scaffold for a 7-layer cognitive loop grounded in the G^16 family of mathematical operators.
A non-destructive scaffold implementing a mathematically grounded core for a 7-layer cognitive loop using the G^16 family set. Part of the MEGAMIND distributed AGI research program at ThatDeveloperGuy.
- C++20 core (
src/common) — minimal, testable state update of the formΨ(t+1) = F(Ψ(t), inputs). - G^16 family mapping — clear association of each family (continuous, discrete, optimization, Bayesian, mutual information, …) to a code path.
- Standalone app binaries in
apps/:agi_core— single step or short rolloutagi_loop— full cognitive loopagi_opt— optimization variantagi_profile— performance profilingagi_step— step-by-step inspectionagi_bridge_v2— streaming bridge with the G^16 v2 core
- Memory indexer in Go (
tools/memory_indexer) — lightweight knowledge index from opt-in directory paths. No Python dependency. - Polyglot bindings — C ABI plus stubs for C#, Java, Julia, Node, Ruby, Rust, and Swift, each in
bindings/<lang>. - Vite + React dashboard in
dashboard/— real-time state visualization.
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j$(nproc 2>/dev/null || sysctl -n hw.ncpu)
./build/apps/agi_core --helpSee docs/ARCHITECTURE.md for the full architectural overview. Key documents:
docs/ARCHITECTURE.md— overall system architecturedocs/BLUEPRINT_EXECUTION.md— execution sequence walkthroughdocs/BUILD_AND_RUN.md— detailed build + run instructionsAGENTS.md— orchestrator agent specification
This codebase only operates on directories it owns or directories you explicitly list in config/roots.txt. The memory indexer is read-only on external paths.
Edit config/roots.txt to list directories you want the memory indexer to read. Each path becomes an opt-in source of knowledge for the index. The file ships with placeholder examples — replace with your own absolute paths before running the indexer.
Each language binding lives under bindings/<lang>/ with its own README. The Rust and Node bindings have working stub implementations; the others are protocol-level stubs ready to be filled in.
SEFERIM AGI G16 is a sub-project of the broader MEGAMIND federated AGI research program. It can be used standalone (the build above is fully self-contained) or as a component of a larger federation of cognitive nodes.
- Studio: ThatDevPro
- Parent agency: ThatDeveloperGuy
- Research hub: feedthejoe.com
- Lead researcher: Joseph W. Anady (ORCID 0009-0008-8625-949X, Wikidata Q139901957)
MIT © 2026 Joseph W. Anady. See LICENSE.