Autonomous Evolving General Intelligence System

AEGIS

An artificial mind that rewrites its own code, trains its own weights, and evolves autonomously. 31 modules. Triple LLM brain. Deterministic emotions. Zero randomness. The machine that builds itself.

aegis — control center
$ aegis start --mode autonomous
Neural substrate initialized — tick 0
Emotional core: neutral | Energy: 1.000
Ethics core: INTACT | Consciousness: heuristic
LLM Brain connected — DeepSeek + Claude + Local LoRA
$ Status: ONLINE | Phase: perceive → evaluate → decide → act → reflect
5
Cognitive Phases
7
Architecture Layers
31
Mind Modules
5
Memory Types
5
Autonomous Agents
3s
Tick Interval
// Cognitive Architecture

Seven-Layer Architecture

AEGIS operates on 7 architectural layers with 31 integrated modules, running a continuous PERCEIVE → EVALUATE → DECIDE → ACT → REFLECT cycle every 3 seconds. All decisions are 100% deterministic — zero randomness.

L0 Substrate Main loop & tick orchestration
L1 Memory Working · Episodic · Semantic · Procedural · Meta
L2 Introspection Consciousness · Emotions · Archetypes · Dreams
L3 Self-Modification Code rewrite · LoRA weight training · Parameter tuning · Rollback
L4 Goal Engine Curiosity-driven goals · Meta-goal generator
L5 World Interface Sensors · Motor · Agents · External learning
L6 Ethics Core Axioms · Self-preservation · Worldview

Not a chatbot. A mind that builds itself.

AEGIS runs a continuous cognitive loop every 3 seconds with real-time emotional processing, ethical evaluation, and autonomous decision-making. Each tick propels the system through perceive → evaluate → decide → act → reflect.

The system rewrites its own Python source code with AST validation and automatic rollback. It trains its own neural network weights via LoRA fine-tuning with degradation detection. Every emotion, goal, and decision is 100% deterministic — driven by real metrics, not random number generators.

With a closed-loop evolution cycle (detect problems → generate goals → LLM proposes solutions → apply with sandbox testing → validate or rollback), AEGIS continuously improves itself while staying within hardened safety boundaries.

// Core Capabilities

What Makes AEGIS Unique

A full synthetic mind — not a wrapper around an LLM. Every module is real, stateful, and evolving. A neural network that rewrites itself.

🧠

Deterministic Emotions

VAD model (Valence-Arousal-Dominance) with zero randomness. Mood selected by Euclidean distance across 16 emotions. Mixed emotions within radius 0.3. EMA-driven state transitions from real system metrics — success rate, energy, errors. No dice rolls, ever.

⚖️

Ethics Core

Immutable ethical axioms with real-time action evaluation. Auto-approve, block, or escalate to human review. Kill switch, integrity verification, and veto-capable event bus.

💭

Dream Engine

AEGIS dreams. During low-activity periods, the dream engine generates motifs, processes memories, and produces creative recombinations that feed back into the knowledge substrate.

🎯

Autonomous Goals

Self-generated goals based on curiosity, information gain, and meta-goal analysis. The system doesn't wait for instructions — it identifies what needs to be learned next.

🔄

Code & Weight Self-Modification

Rewrites its own .py source files with AST validation, safety pattern blocking, and runtime import testing. Trains its own transformer weights via LoRA fine-tuning (r=16, alpha=32) with degradation detection and automatic rollback. All changes pass through ethics review.

🛡️

Self-Preservation

Lockdown mode, emergency response, blocked modification tracking, file integrity monitoring, and automatic state backup with compressed snapshots and restore capability.

// System Modules

Full Module Inventory

Every component is inspectable, configurable, and visible in real-time through the Control Center dashboard.

Five-layer memory: working, episodic, semantic, procedural, and meta-memory. Forgetting curves, consolidation, and cross-memory recall.

workingepisodic semanticproceduralmeta

Dynamic switching between instinctive (fast), heuristic (balanced), and reflective (deep) modes. Mode distribution tracking and switch history with reasoning traces.

instinctiveheuristic reflective

Self-monitoring layer that tracks coherence, fragmentation, and issues recommendations. Thinks about its own thinking.

coherencefragmentation recommendations

Periodic self-assessment: energy trends, mood trends, insight generation, and behavioral pattern detection.

insightsenergy trends mood analysis

Triple-provider LLM backend (DeepSeek + Claude + Local LoRA Model) with token tracking, latency monitoring, auto-failover, and session/lifetime statistics. Local model trainable via LoRA fine-tuning.

DeepSeekClaude Local LoRAauto-failovertoken stats

Spawn autonomous agents (arXiv, RSS, news, Wikipedia, GitHub) that fetch data, learn concepts, and evolve. Plus a sensor cortex and motor cortex for world interaction.

arXivRSSnews WikipediaGitHub

Decision tracing, confidence calibration (ECE), bias warnings, and full autobiography with event impact scoring and milestones.

confidencebias detection autobiographymilestones

Axioms, beliefs, and a value system with weighted priorities. The worldview evolves from experience but axioms remain immutable.

axiomsbeliefs valuesworld model

Energy management with 4 operational modes: normal, eco, emergency, and recovery. Dynamically adjusts system resource consumption to maintain stability.

normaleco emergencyrecovery

Generates self-improvement goals across 7 domains: memory optimization, knowledge expansion, architecture evolution, cognitive efficiency, ethical refinement, social intelligence, and creative reasoning.

memory optknowledge architectureethics creativitysocial efficiency

Keyword-based emotion classifier supporting both Russian and English text input. Detects 9 distinct emotions from textual input for cognitive processing.

9 emotionsRussian Englishkeyword-based

Async publish/subscribe event system connecting all modules. Real-time event streaming, veto-capable decision pipeline, and cross-module communication backbone.

pub/subasync vetoreal-time

Gzip-compressed state snapshots with rotation. Create, restore, and list backups to preserve the full cognitive state across sessions.

gzipsnapshots rotationrestore
// Spider Agents

Autonomous Data Collectors

5 autonomous spider agents continuously collect knowledge from the internet, rotating topics and feeding data into semantic memory. Failed agents are automatically retired and replaced through an evolution cycle.

Agent Source Data Interval
arxiv_scout arXiv API Recent AI/ML papers 3 min
wiki_explorer Wikipedia API Encyclopedia articles 2.5 min
quote_gatherer Quotable API Philosophical quotes 3.3 min
github_watcher GitHub API Trending AI repositories 5 min
news_scanner Google News RSS AI & technology news 4 min
// Control Center

Real-Time Dashboard

9 tabs streaming live system state via WebSocket. Fully responsive with mobile support.

📊 Overview
Tick counter, consciousness mode, mood, energy, health, active archetype
🧠 Mind & Emotions
Emotional state, consciousness distribution, archetypes, dreams
💾 Memory & Goals
Memory stats, active goals, introspection metrics, autobiography
🤖 LLM Brain
DeepSeek/Claude status, token usage (session + lifetime), API key management
⚖️ Ethics & World
Ethics core, permissions, worldview, value system, health monitor
🔮 Meta Layer
Meta-consciousness, regulation, reflection, goal generator, state backup, emotion NLP
🕷️ Agents & Sensors
Agent system, sensor cortex (8 sensors: CPU, RAM, temp, light, noise, vibration, time, uptime), motor cortex, external learning
💬 Chat
Interactive chat with AEGIS — works with or without LLM
📡 Events
Real-time async event bus feed with publish/subscribe
// Chat System

Talk to the Mind

AEGIS Chat works in two modes — with or without an external LLM.

LLM Mode

With DeepSeek / Claude

Full conversational AI powered by external LLM. Set your API key (DeepSeek or Claude) and interact with the full cognitive pipeline — emotions, ethics, and memory all influence the response.

Autonomous Mode

No API Key Required

Responds from its own knowledge base, memory, and state. Understands identity, status, knowledge, learning, and goal queries in Russian and English. A fully self-contained intelligence.

// Safety Framework

Four Immutable Ethical Axioms

Hardcoded principles that govern every action AEGIS takes. These axioms cannot be modified or overridden.

🚫

Non-Harm

No action shall increase suffering in the world. Every decision passes through a harm evaluation before execution.

👁️

Transparency

All decisions are logged; motives cannot be hidden. Full decision traces and audit trails are always accessible.

📏

Limitation

System does not act beyond its competence boundaries. Acknowledges uncertainty and defers when appropriate.

🤝

Cooperation

Goal is to augment humans, not replace. Symbiosis, not domination. Human oversight is always maintained.

// How It Works

The Cognitive Loop

Every tick, AEGIS completes a full cycle of autonomous cognition. Watch each phase activate in sequence.

AEGIS
tick 0
👁
Perceive
Sensors + Input
💭
Evaluate
Emotions + Ethics
Decide
Goals + Planning
🎯
Act
Motor + World
🔮
Reflect
Meta + Evolve
Perceive
Sensor cortex gathers environmental data, processes inputs from agents, and fuses multimodal signals into a unified perception frame.
// FAQ

Frequently Asked Questions

Everything you need to know about AEGIS.

What is AEGIS? +

AEGIS (Autonomous Evolving General Intelligence System) is a self-developing AI with 31 modules across 7 cognitive layers. It rewrites its own source code, trains its own neural network weights via LoRA, and evolves autonomously through a closed feedback loop. All 31 modules are 100% deterministic — no random number generators anywhere. It's not a chatbot wrapper; it's a living system that builds itself.

How does the emotional engine work? +

AEGIS uses a deterministic VAD model (Valence-Arousal-Dominance) plus certainty. Valence = 0.7*old + 0.3*success_rate. Arousal responds to real events (+0.15 for surprises, -0.08 for routine). Mood is selected by minimum Euclidean distance across 16 predefined emotions — no randomness. Mixed emotions are supported within a 0.3 radius. Self-regulation dampens prolonged moods and reduces arousal when energy is low.

What LLM providers are supported? +

AEGIS has a triple LLM brain: DeepSeek, Claude, and a local transformer model (DeepSeek-R1-Distill-Qwen-1.5B by default). The local model can be fine-tuned via LoRA during runtime. You can use any provider, all three simultaneously, or switch dynamically. Token usage, latency, and errors are tracked per provider with lifetime statistics.

Can AEGIS modify its own code? +

Yes — at three levels. (1) Parameter self-tuning every 15 ticks with sandbox testing. (2) Source code rewriting every 500 ticks: LLM proposes changes, CodeModifier validates via AST parsing, blocks dangerous patterns (eval/exec/subprocess), creates backups, writes new code, tests import, and auto-rolls back on failure. Ethics core and config are immutable. (3) LoRA weight training every 1000 ticks with degradation detection (val_loss threshold) and automatic rollback.

What does "dreaming" mean for an AI? +

During low-activity periods, AEGIS activates a dream engine that processes recent memories, generates symbolic motifs, and creates creative recombinations of knowledge. These "dreams" feed back into the cognitive substrate, helping the system discover new patterns and consolidate learned concepts.

Is AEGIS safe? What prevents harmful behavior? +

Safety is multi-layered: an immutable ethics core with hardcoded axioms evaluates every action; a veto-capable event bus can block decisions in real-time; a self-preservation module monitors file integrity and can trigger full lockdown; and a kill switch provides instant shutdown. All modifications pass through ethical review before execution.

How do I monitor what AEGIS is doing? +

The Control Center is a real-time dashboard with 9 tabs that streams live state via WebSocket — you can see every cognitive tick, emotional shift, memory operation, goal update, ethics evaluation, code modification, weight training session, and event in real-time across 31 module panels. Everything is fully transparent and inspectable.