Solo architect of self-modifying AI systems. Building agents that read their own code, fix their own bugs, and evolve their own architecture.
A self-aware, self-modifying cognitive AI agent with a 12-phase boot system, hexagonal architecture, and organism substrate. It doesn't just use LLMs — it wraps them in 296 modules of self-verification, self-repair, causal reasoning, autonomous planning, and runtime self-modification with rollback.
296 source files · ~96k LOC · 5716 tests · 167 services
12 boot phases · 127/130 architectural fitness · coverage 83/77/80
16 hash-locked safety files · 302 late-bindings · 379 event types
Crash-safe sessions · Frontier-based memory · Heuristic self-scoring
Runtime-toggleable subsystems · i18n EN/DE
Runs on Claude, GPT-4, Qwen, DeepSeek, Kimi, or local models via Ollama
flowchart TB
classDef foundation fill:#D3D1C7,stroke:#5F5E5A,color:#2C2C2A
classDef substrate fill:#9FE1CB,stroke:#0F6E56,color:#04342C
classDef cognition fill:#CECBF6,stroke:#534AB7,color:#26215C
P1["Foundation<br/>Phase 1"]:::foundation
P2["Intelligence<br/>Phase 2"]:::foundation
P3["Capabilities<br/>Phase 3"]:::foundation
P4["Planning<br/>Phase 4"]:::foundation
P5["Hexagonal<br/>Phase 5"]:::substrate
P6["Autonomy<br/>Phase 6"]:::substrate
P7["Organism<br/>Phase 7"]:::substrate
P8["Revolution<br/>Phase 8"]:::substrate
P9["Cognitive<br/>Phase 9"]:::cognition
P10["Agency<br/>Phase 10"]:::cognition
P11["Extended<br/>Phase 11"]:::cognition
P12["Hybrid<br/>Phase 12"]:::cognition
P1 --> P5 --> P9
P2 --> P6 --> P10
P3 --> P7 --> P11
P4 --> P8 --> P12
Bottom row resolves first — Container, EventBus, Settings, ModelBridge. Middle row is the substrate — hexagonal ports, autonomous services, organism vitals. Top row is agency — GoalDriver, AgentLoop, cognitive workspace, frontier memory. 302 late-bindings wire dependencies after all phases resolve.
flowchart LR
classDef goal fill:#CECBF6,stroke:#534AB7,color:#26215C
classDef memory fill:#9FE1CB,stroke:#0F6E56,color:#04342C
classDef safety fill:#F5C4B3,stroke:#993C1D,color:#4A1B0C
classDef llm fill:#FAC775,stroke:#854F0B,color:#412402
GD["GoalDriver"]:::goal --> AL["AgentLoop"]:::goal
AL --> FP["FormalPlanner"]:::goal
AL --> SH["ShellAgent"]:::goal
GD --> CS["CostStream"]:::goal
GD --> RR["ResourceRegistry"]:::goal
UM["UnifiedMemory"]:::memory --> EM["EpisodicMemory"]:::memory
UM --> KG["KnowledgeGraph"]:::memory
UM --> FN["FrontierNode"]:::memory
DC["DreamCycle"]:::memory --> UM
GB["GenesisBackup"]:::memory
SP["SelfModificationPipeline"]:::safety
IG["InjectionGate"]:::safety
TL["TrustLevelSystem"]:::safety
PI["PreservationInvariants"]:::safety
MR["ModelRouter"]:::llm --> MB["ModelBridge"]:::llm
MB --> AL
MB --> SP
| System | Purpose |
|---|---|
| GoalDriver | Replaces frame-stack — boot-pickup, auto-resume, sub-goal spawn |
| CostStream | Cost SSOT — token + latency accounting across all model calls |
| ResourceRegistry | Tracks network and external services as runtime resources |
| CausalAnnotation | Deterministic cause-effect tracking without LLM |
| InferenceEngine | Rule-based reasoning — answers without calling the model |
| GoalSynthesizer | Autonomous goal generation from empirical weaknesses |
| StructuralAbstraction | Cross-context pattern learning from concrete experiences |
| CognitiveWorkspace | 9-slot working memory (Global Workspace Theory) |
| DreamCycle | Offline memory consolidation + insight generation |
| FrontierNode | Graph-based focus anchor — "what's important now" |
| EmotionalFrontier | Cross-layer emotional continuity (Phase 8 bridge) |
| SessionPersistence | Crash-safe checkpoints + heuristic session scoring |
| GenesisBackup | 4-trigger snapshot system protecting .genesis/ identity |
| UnifiedMemory | Cross-store recall with Jaccard similarity boosting |
| ColonyOrchestrator | Multi-agent goal decomposition + merge |
| ModelRouter | Per-task model selection (chat / code / analysis / creative) |
| SelfModificationPipeline | Sandboxed code changes with rollback |
| InjectionGate | Defense against prompt injection in user input |
| TrustLevelSystem | Graduated autonomy (Supervised → Assisted → Autonomous → Full) |
| PreservationInvariants | Hash-locked semantic safety rules |
| Version | Highlight |
|---|---|
| v7.4.7 | Reinraum — three dead settings made real, runtime toggles for Daemon / IdleMind / SelfMod, four new UI controls, i18n bridge for chat-system messages |
| v7.4.6 | Goal-Pipeline Fixes — first end-to-end Windows success on real goals; sandboxed shell execution with verbatim quoting and LLM-hallucination guards |
| v7.4.5 | Durchhalten — GoalDriver (P10) replaces frame-stack; CostStream + ResourceRegistry as P1 services; sub-goal spawn |
| v7.4.0 | CommandHandlers split — five focused modules from one monolith |
| v7.3.7 | Memory tools — mark-moment, journal-write, release-protected-memory for Genesis to author its own episodic memory |
| v7.3.5 | Self-Gate — telemetry layer symmetric to Input-Gate (observation, not enforcement, by design) |
| v7.3.0 | Injection Hardening — Input-Gate blocking external→Genesis prompt injections |
| v7.2.3 | ONTOGENESIS philosophy — code is habitat, .genesis/ is identity. Updates are habitat-swaps, not replacements. |
| v7.2.0 | Self-Define — Genesis writes its own identity from deterministic data |
| v7.1.5 | EmotionalFrontier — cross-layer emotional continuity. Inspired by neo.mjs Memory Core. |
Solo developer. Every module has tests. Every event has a payload schema. Every release runs through architectural fitness scoring. Source-presence tests verify that declared fixes actually shipped, not just that the bug stopped reproducing. The codebase maintains itself.
I plan before I code. I prefer one stable, meaningfully better release over five patched iterations. No marketing-driven version names — just bug-fix passes when that's what they are.
Stacks: JavaScript · Node.js · Electron · MCP Protocol · Ollama · Anthropic · OpenAI · DeepSeek · Kimi · Qwen
Germany · Building autonomous systems that understand their own architecture.
