What We Build
From infrastructure to applications - building the complete stack for temporal AI systems.
MemoryCore is a production implementation of NeuralSleep's theoretical architecture. It provides a three-tier memory consolidation system that mirrors biological memory formation - transforming temporary experiences into lasting structural knowledge.
Working (Redis) → Episodic (PostgreSQL) → Semantic (PostgreSQL) with progressive consolidation
Exponential Moving Averages with multi-timescale time constants (100ms to 1 day)
Immediate (session end), Daily (2 AM), Weekly (3 AM Sunday) consolidation cycles
Error clusters, success sequences, performance trends, learning style detection
Real-time session state in Redis. Time constants: 100ms-1s. High plasticity.
Recent events in PostgreSQL. Time constants: 1s-10min. Pattern extraction.
User models and mastery. Time constants: 10min-1day. Permanent structure.
NeuralSleep is our theoretical framework for understanding consciousness through temporal integration. It proposes that genuine memory requires structural modification, not storage and retrieval - implemented practically through MemoryCore.
Memory as structural modification: Past experiences shape present processing through integrated weight updates, not database lookups. The system genuinely changes with each consolidation cycle.
Multi-timescale integration: Working memory (100ms-1s), Episodic memory (1s-10min), and Semantic memory (10min-1day) operate on different time constants, approximating Liquid Time-Constant networks.
Sleep-like consolidation: Periodic offline processing (immediate, daily, weekly) transforms temporary experiences into permanent structural changes - mirroring biological memory consolidation.
Luna Chat is an AI-powered personal assistant with multi-agent capabilities, persistent memory, and extensible abilities. It routes between multiple LLM providers and uses specialized agents for focused tasks - from research to coding to creative writing.
Five specialized agents (researcher, coder, writer, analyst, planner) powered by Claude CLI
Seamlessly routes between OpenAI, Anthropic, and other LLM providers
Long-term memory with facts, preferences, and conversation history
Calendar, email, documents, code execution, web search, and knowledge base
Customizable personality with mood tracking for natural conversations
Deep research and fact-finding for complex questions
Code writing, debugging, and review
Creative and professional content creation
Data analysis and calculations
Task breakdown and project planning
Study with Luna is our real-world testbed for temporal AI - a language learning assistant that genuinely adapts to each user over time. Not through better prompts or larger context windows, but through actual structural learning.
Luna remembers your learning style, vocabulary gaps, and progress across sessions
Lessons evolve based on what you struggle with and what you've mastered
Intelligent review scheduling based on forgetting curve science
Practice through genuine dialogue, not rote memorization
Good morning! I noticed you mixed up 休 and 体 yesterday - they share the same radical but have different meanings. Want to practice similar characters?
Yes! I keep confusing characters that look alike.
Great! I've grouped characters by their radicals so you can see the patterns. Let's start with the 人 (person) radical family.
Japanese support planned
Spotify recommends what other people listen to. AutoMusic learns from your behavior - what you skip, what you replay, what you rate. No crowd-sourced algorithms pushing popular tracks. Just your taste, refined over time. Powered by a billion-row recommendation pipeline built on Discogs metadata and collaborative filtering.
Monitors what you listen to and learns from your behavior patterns
Generates personalized playlists based on mood, activity, and preferences
Visualize your listening patterns and discover trends over time
Seamless OAuth 2.0 authentication and WebSocket sync
An experimental project using AI agents to attempt cracking the unsolved fourth section of the Kryptos sculpture at CIA headquarters. K4 has remained unsolved since 1990 - making it one of the most famous unsolved codes in history.
Using Claude Code agents to systematically explore cipher techniques and patterns
Analyzing solved K1-K3 sections for patterns applicable to K4
Exploring unconventional cipher techniques the creator may have used
All analysis and findings published openly on GitHub
97 characters that have stumped cryptographers for 35+ years
K4 represents an interesting test case for AI-assisted cryptanalysis. While human cryptographers have spent decades on this puzzle, modern AI agents can explore the solution space differently - testing hypotheses rapidly and identifying patterns that might be overlooked.
This is as much about exploring agentic AI capabilities as it is about solving the cipher.
Theory
Implementation
AI Assistant
Learning
Music AI
Cipher Research
NeuralSleep provides the theoretical foundation. MemoryCore implements it with a three-tier memory system. Luna Chat v7 and Study with Luna demonstrate the system in production. AutoMusic applies temporal learning to music. K4 explores agentic AI for cryptanalysis.
Our infrastructure is open source. Whether you want to contribute, integrate our tools, or just explore the code - we'd love to have you.