Reading
This article introduces the PicoCluster Claw project, a localized AI agent device built on Raspberry Pi 5 and NVIDIA Jetson Orin Nano. It runs OpenClaw and Ollama to enable fully private LLM inference, with an overall power consumption of only about 15 watts.
Section 01
This article introduces the PicoCluster Claw project, a localized AI agent device built on Raspberry Pi 5 and NVIDIA Jetson Orin Nano. It runs OpenClaw and Ollama to enable fully private LLM inference, with an overall power consumption of only about 15 watts.
Section 02
In today's era of increasingly centralized cloud computing and AI services, data privacy and long-term subscription costs have become key concerns for many users. The PicoCluster Claw project offers a refreshing solution—a fully localized, low-power, self-hosted AI agent device with a monthly electricity cost of less than $2.
This device consists of two core components:
Together, they form a private, always-on, low-energy personal AI infrastructure.
Section 03
The Raspberry Pi5 acts as the system's "brain" and runs the following services:
Section 04
The Jetson Orin Nano Super serves as the "computing engine" and focuses on AI inference:
Section 05
The power consumption performance of the entire system is impressive:
| Scenario | Power Consumption | Monthly Cost | Annual Cost |
|---|---|---|---|
| Idle | 14W | $1.61 | $19.62 |
| Typical Load | 20W | $2.30 | $28.03 |
| 90% Idle Mix | 15W | $1.73 | $21.02 |
Based on current electricity prices, the annual operating cost is only about $21, which is far lower than the monthly fee of most cloud services.
Section 06
The clustercrush node comes pre-installed with 9 large language models, occupying approximately 27GB of NVMe storage:
| Model | Size | Type |
|---|---|---|
| llama3.2:3b | 2.0 GB | General Dialogue (Main) |
| llama3.1:8b | 4.9 GB | General Dialogue (High Quality) |
| gemma3:4b | 3.3 GB | General (Multilingual) |
| phi3.5:3.8b | 2.2 GB | Reasoning Specialized |
| deepseek-r1:7b | 4.7 GB | Reasoning (Chain of Thought) |
| qwen2.5:3b | 1.9 GB | Code/Structured Output |
| starcoder2:3b | 1.7 GB | Code Generation |
| llava:7b | 4.7 GB | Visual Understanding |
| moondream:1.8b | 1.7 GB | Vision (Lightweight) |
Ollama automatically manages model loading and unloading, dynamically allocating GPU memory based on usage needs.
Section 07
PicoCluster Claw has 5 built-in MCP (Model Context Protocol) servers, providing 28 practical tools for local LLMs:
Section 08
Through the Pimoroni Blinkt! LED strip, it enables:
Keep going with more reads from the same topic.
Splinter is a minimalist, high-performance key-value (KV) and vector storage system enabling zero-latency inter-process communication via shared memory and atomic operations. With only 766 lines of core code, it supports millions of operations per second and 768-dimensional vector storage, offering a new architectural approach for local LLM inference and data-intensive applications.
Folkering OS is the world's first AI-native bare-metal operating system, entirely written in Rust no_std without relying on Linux, POSIX, or libc. It can generate commands from scratch, compile them into WASM, and run them in 10 seconds, achieving true self-evolution.
This project explores how to use Large Language Models (LLMs) to detect logical vulnerabilities in Ethereum smart contracts, especially business logic flaws that are difficult for traditional static analysis tools to capture, such as economic manipulation and execution order errors.
A self-hosted MCP server that provides cross-session persistent coordination, task management, event bus, and observability for AI agents. It supports 35 tools and can be integrated into clients like Claude Code and Cursor with zero code changes.