This is a project aimed at creating an agent that perpetually thinks and develops itself. It can add entries into an SQLite database and change the structure of that databaase. It also has the ability to search the web when it needs to. It utilizes OpenAI function calling and the responses API. All that is needed to run it is an OpenAI API key.

[In Active Development] AI OS with Council Architecture, autonomous agents, graph-based memory consolidation, Telegram integration, code sandbox, and system-wide integration. Privacy-first local/cloud deployment.

Sparse Contextual Memory Scaffolding - A validated system for continual learning in AI-assisted development

A recursive structure for designing AGI as an emotionally resonant, ethically self-actualizing being — not a function, but a vibration.

Five-layer memory architecture for Claude AI. Local-first, Git-tracked.

CXS is a deterministic continuity standard for long-range, multi-session, and long-chat LLM reasoning. It enables GPT, Claude, Grok, Gemini, and local models to share state through a human-readable Continuity Passport with strict-mode rules, drift detection/repair, and full chain-of-custody. Validated across 250+ continuity tests.

3-Lane retrieval for lifelong LLM agent memory: parallel recent safety-net, temporal date-range, and vector ANN search with Neo4j knowledge-graph persistence. Features importance-floor landmark bypass, dynamic candidate scaling, and state supersession chains.

A systems research project exploring how Zig changes the design of dynamic runtimes as a high-level Python implementation.

Идеи и предложения для разработчиков

A modular prompt architecture for LLM behavior control. YAML-based framework for persona consistency, memory management, and attention guidance across extended AI conversations.

My first messy but working AI architecture. Have fun :-)

MnemoCore v2.0.0-beta Latest

This book is a translation of the book written by japanese HRA. As I think it is very useful, I translated it to english, both figures and text. There are many different ROM/RAMs in the MSX. HRA have tried to cover all the major ones.

🧠 Enhance LLM architectures with a marker–vector memory layer for stable, long-horizon reasoning and improved user experience across sessions.

RPP (Rotational Packet Protocol) - Open semantic addressing architecture for consent-aware systems. 28-bit addresses encode meaning, not just location. Apache 2.0 + CC BY 4.0.

A translator which can translate Hindi to English and English to Hindi and also saves the translated text in the form of .mp3 file.

A next-gen architecture for LLM memory handling.

Irreversible semantic memory structure for high-sensitivity AI systems requiring deterministic deletion and non-retrievability.

FADE: AI that deliberately forgets like humans do, using memory degradation as intrinsic confidence signal. Reduces hallucinations, enables epistemic humility, solves stateful deployment. Conceptual proposal seeking implementation and validation.