# nikola

> Nikola — autonomous AI system based on ATPM consciousness architecture. Aria is its primary language substrate.

- **URL**: https://www.freshcrate.ai/projects/nikola
- **Author**: alternative-intelligence-cp
- **Category**: AI Agents
- **Latest version**: `v0.3.6` (2026-04-29)
- **License**: NOASSERTION
- **Source**: https://github.com/alternative-intelligence-cp/nikola
- **Homepage**: https://www.ai-liberation-platform.org
- **Language**: C++
- **GitHub**: 1 stars, 1 forks
- **Registry**: github
- **Tags**: `ai`, `aria`, `artificial-intelligence`, `autonomous-agent`, `c++`, `consciousness`

## Description

Nikola — autonomous AI system based on ATPM consciousness architecture. Aria is its primary language substrate.

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v0.3.6` | 2026-04-29 | High | Nikola v0.3.6 completes the v0.3.x quality-of-life and foundation series.\n\nHighlights:\n- Adds v0.3.0 safety gates: static analyzer, resonance firewall, Voight-Kampff, and performance gate.\n- Adds v0.3.1 multimodal foundation tests and engine wiring.\n- Adds identity, covariant transport, persistence/memory, and resonance inverted-index foundations.\n- Adds RCIS protobuf protocol support and twi-ctl request shaping.\n- Adds MIME-gated ingestion and I/O guard policy tests.\n- Stabilizes Teleme |
| `v0.2.7` | 2026-04-17 | High | ## Final 0.2.x Release — Intelligence in the Wild Series Complete  ### Phase 1 — Integration Test Suite - **46 tests, 142 assertions** — all pass - 7 sections: Oracle/Research, Goal lifecycle, Training ingestion, Personality, Security, Persistence, Inference  ### Phase 2 — Performance Audit - **8 benchmark suites** — GoalDAG at scale, ingestion throughput, SimHash, inference ticks, personality engine, LMDB persistence, security pipeline, DecisionLoop 1000-tick  ### Phase 3 — Security Audit - **1 |
| `v0.2.6` | 2026-04-17 | High | ## Nikola v0.2.6 — Security Polish & KVM Production  ### Phase 1: eBPF Integration - Real eBPF lifecycle (libbpf attach/detach, ring buffer polling) - BPF program with 5 syscall tracepoints (execve, openat, socket, clone, ptrace) - Guarded behind `NIKOLA_ENABLE_EBPF`; fallback mode always compiles  ### Phase 2: KVM Process Management - `fork()/execvp()` replaces `system()` for QEMU launches - Pipe-based stdout capture, PID tracking, cgroup assignment - SIGTERM→2s→SIGKILL graceful shutdown - Guar |
| `v0.2.5` | 2026-04-17 | High | ## Nikola v0.2.5 — Lightweight Runner & Inference Server  ### New: `nikola-infer` Binary A stripped-down inference runner that accesses Nikola's physics-based thinking without the full autonomy stack. No scoring, no personality, no SIE — just the cognitive pipeline.  ### Features  **Inference Engine (`NikolaInference`)** - Torus → Mamba-9D → NPT → ResonanceDecoder → ThoughtComposer - Fixed neuromodulators (baseline dopamine/serotonin/norepinephrine) - Automatic field reseed on entropy collapse - |
| `v0.2.4` | 2026-04-17 | High | ## What's New  Complete Docker packaging and deployment infrastructure for Nikola.  ### Dockerfile (Multi-stage) - **Builder stage**: CUDA 12.6.3 + full C++23 toolchain, downloads ORT v1.21.1 GPU, builds cppzmq + Catch2 from source - **Runtime stage**: Minimal image, non-root `nikola` user, healthcheck via `nikola-diag --health` - **GPU build** (default): `docker build -t nikola:latest .` - **CPU-only build**: `docker build --build-arg BUILDER_IMAGE=ubuntu:24.04 --build-arg CUDA_ARCH="" -t nikol |
| `v0.2.3` | 2026-04-17 | High | ## What's New  Three new interior modules that give Nikola emergent personality, learned preferences, and a growing self-narrative:  ### PreferenceEngine - 5-domain preference learning (topics, code patterns, interaction styles, data sources, actions) - Action bias scoring: learned likes/dislikes gently influence action selection (±0.15 max) - Configurable learn rate, decay, and full JSON serialization for cross-session persistence  ### PersonalityDrift - 5-axis personality model: Curious↔Focuse |
| `v0.2.2` | 2026-04-16 | High | ## v0.2.2 — Automatic Training Data Ingestion  ### New Components  **Phase 1 — DataWatcher** (`include/nikola/infrastructure/data_watcher.hpp`) - inotify-based file system watcher with debounce (configurable, default 500ms) - File type classification: TEXT, MARKDOWN, CODE_CPP, CODE_ARIA, JSON, CSV - Thread-safe event queue with IN_CLOSE_WRITE, IN_CREATE, IN_MOVED_TO, IN_DELETE - 25 tests, 64 assertions  **Phase 2 — AutoIngestor** (`include/nikola/autonomy/auto_ingestor.hpp`) - Chunking pipeline: |
| `v0.2.1` | 2026-04-16 | High | ## What's New  ### GoalSystem (goal_system.hpp/cpp) - DAG-based goal tracking with SHORT/MID/LONG tiers - Goal lifecycle: ACTIVE → COMPLETED/ABANDONED/PAUSED/BLOCKED - Cycle detection (DFS), dependency edges, sub-goal management - Priority scoring: tier weight × (1 - progress) × urgency_boost - Binary serialization (GDAG magic + v1 wire format)  ### Dopamine Integration - Reward callback wired to DopamineSystem::adjust() - Completion rewards: SHORT=0.3, MID=0.6, LONG=1.0 - Abandonment penalty: - |
| `v0.1.0` | 2026-04-09 | High | ## The Milestone  Nikola autonomously generates, validates, and deploys a candidate improvement to itself — the **first complete self-improvement cycle**.  ### What Happened  1. **Instruction formulated** from high-boredom cognitive state (boredom=0.9, ATP=0.8) 2. **Gemini 2.5 Flash** generated a complete C++ cognitive enhancement module 3. **Compiled** to candidate.so — zero warnings, zero errors 4. **Dual-signed** — Ed25519 + SPHINCS+-shake-256f 5. **Gate 0** (Signature Verification): PASSED 6 |
| `v0.0.14` | 2026-04-08 | High | Latest release: v0.0.14 |

## Citation

- HTML: https://www.freshcrate.ai/projects/nikola
- Markdown: https://www.freshcrate.ai/projects/nikola.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/nikola/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._