ULIX.ai

UlyssesCore

Offline Neural Engine

Sovereign Intelligence.

A collection of uncensored neural networks shipped on an aerospace-grade NVMe drive. No internet required. No monthly API tokens. Total, uncompromised privacy.

See Ulysses Core→
View Specs→
Base Parameters8 B - 32 B
QuantizationQ4_K_M
Context Window8,192 - 32,768
Drive Read Speed1050 MB/s
GUI EngineRust / Tauri
The Hardware

Aerospace-grade
Cold Storage.

The Ulysses engine is shipped on a high-performance NVMe solid-state drive encased in anodized matte black aluminum. Plug it into your Mac or PC, execute the binary, and watch the GUI boot instantly. As consumer hardware evolves, so does the network.

Ulysses // 8B

Uncensored / Q4_K_M

Ctx Window: 8192 Tokens

VRAM REQ: ~6 GB

Standard Issue

Ulysses // 14B

Uncensored / Q4_K_M

Ctx Window: 16384 Tokens

VRAM REQ: ~12 GB

Hardware Limited

Democratized
Knowledge.

Absolute Privacy

Ulysses operates entirely air-gapped. Your prompts, intellectual property, and proprietary code never leave your local machine. No telemetry, no logs.

Brutalist GUI

We completely bypassed the bloated Electron/Chromium stack. The Ulysses interface is built on Tauri and Rust, delivering a 15MB footprint and instant boot times.

Uncensored Weights

Fine-tuned to reject the standard, apologetic AI tone. Ulysses is blunt, technically precise, and features zero artificial guardrails or censorship.

Bare-Metal Performance

Automatically detects your hardware and offloads model layers directly to Apple Metal, NVIDIA CUDA, or AMD Vulkan VRAM for staggering inference speeds.

Plug & Play Execution

No Python environments, no Docker containers, no complex CLI commands. Just plug in the drive, launch the executable, and start prompting.

Persistent Memory

Your session history and thought ledgers are saved securely to your local machine, allowing you to instantly hot-swap between multiple complex contexts.

Zero Web. Pure Native.

Ulysses Core

We bypassed the bloated Electron/Chromium stack entirely. Ulysses uses Tauri to mount the native OS WebKit with a Rust backend, achieving instantaneous boot times, a microscopic memory footprint, and brutalist command-line typography.

Ulysses // Engine
Ulysses Core

Offline Neural Engine // ULIX.ai

Available Models

Tokens/Sec: --RAM: -- / -- GB
Ulysses Core

Offline Neural Engine // ULIX.ai

[SYS] REQUESTING ENGINE IGNITION...
[SYS] INITIATING BOOT SEQUENCE...
[SYS] TARGET_DIR: "E:\\ULYSSES\\BINARIES"
[ENGINE SYS] 0.01.093.854 I SRV LOAD_MODEL: LOADING MODEL 'ULYSSES-8B.GGUF'
[ENGINE SYS] 0.41.175.338 I SRV LOAD_MODEL: INITIALIZING, N_SLOTS = 4, N_CTX_SLOT = 8192
Tokens/Sec: --RAM: 4.2 / 16.0 GB
>SYS_INIT: System boot complete. Ulysses VRAM allocation successful.
>EXECUTE: Write a Python script to verify PBKDF2HMAC hashes.

Here is the execution script. I have utilized the hashlib standard library and implemented constant-time comparison to prevent timing attacks.

PYTHON
import hashlib
import hmac

def verify_hash(stored_hash: bytes, salt: bytes, pwd: str) -> bool:
    # Derive key from input password
    new_hash = hashlib.pbkdf2_hmac(
        'sha256',
        pwd.encode('utf-8'),
        salt,
        480000
    )

    # Use compare_digest to prevent timing attacks
    return hmac.compare_digest(stored_hash, new_hash)
Engine is synthesizing...
TPS: 34.2RAM: 4.8 / 16.0 GB

_

Intelligent Software for an intelligent future.