A

ai-reasoning

ThreadGraph is an open-source, AGPL-3.0+ modular agent swarm runtime that transforms complex tasks into executable, graph-structured workflows. It orchestrates multiple specialized agents, tools, and memory systems through a dynamic execution graph, enabling parallel reasoning, structured decomposition, and fully observable workflows. Designed for extensibility and composability, ThreadGraph turns agent-based systems into scalable, reproducible infrastructure rather than linear prompt chains.

Project Solstice is a layered reasoning system stack designed to transform computation into structured, verifiable cognition. It combines neural inference, symbolic reasoning, adversarial validation, and autonomous orchestration into a unified architecture that emphasizes traceability, modularity, and correctness. Rather than producing isolated outputs, it builds full reasoning pipelines that can be decomposed, verified, versioned, and improved over time.

Most LLM reasoning benchmarks come from Western math, English logic, or code. Bazi-Bench tests multi-step rule-following inference in a different formal system: traditional Chinese Ba Zi. Frozen tables, Python reference impl, gold CoT cases, all mechanically verifiable.

An interactive React application to explore and visualize Gemini chat conversations with advanced rendering, sidebar navigation, and local JSON file loading. (This is a mirror project)

Logical concept analysis MCP server - break through confused thinking with structured decomposition. Uses Wittgenstein's Tractatus method to show WHAT something is (not HOW to do it) by revealing hidden dependencies and multiplicative requirements that sequential thinking misses.