Blog

Latest insights and articles about AI and automation.

Abstract illustration of a local inference server with automated setupAbstract illustration of a local inference server with automated setup

From Unboxing to private AI brain: Why I Built machine_setup_automation

A Strix Halo or DGX Spark is unpacked in minutes — yet a weekend still passes before the first token is generated. kwisatz reduces local inference server setup to a single YAML file and one command.

August 21, 2026 by Julian Weber

DevOpsLokale KIHomelabllama.cppOpen Source
Abstract illustration of RAG systemsAbstract illustration of RAG systems

RAG Without Frontier Models: Why Good Retrieval Matters More Than the Biggest LLM

How our advanced RAG setup uses hybrid search, reranking, and local open-weight models—and why it does not need a frontier model.

June 24, 2026 by Manuel Spörer

KIRAGLLMOpen WeightWeaviate
Abstract illustration of LLMs, training processes, and quantization formatsAbstract illustration of LLMs, training processes, and quantization formats

LLMs Explained Clearly: Training, Model Size, Quantization, and K-Quants

What LLMs really are, how they are trained, which architectural differences matter, and why quantization and K-Quants determine whether local use is even practical.

May 18, 2026 by Manuel Spörer

KILLMTrainingQuantisierungOpen Source
Abstract illustration of a neural networkAbstract illustration of a neural network

Neural Networks Explained: How Machines Learn From Examples

Neural networks are neither magic nor digital brains. They are systems that learn from examples, measure errors, and gradually adjust internal weights. A clear introduction to the basic idea behind modern AI.

April 26, 2026 by Manuel Spörer

KINeuronale NetzeMachine LearningDeep LearningLLM
Abstract illustration of CUDA, GPU architectures, and toolchain layersAbstract illustration of CUDA, GPU architectures, and toolchain layers

CUDA, Compute Capability, and Drivers: Why New NVIDIA GPUs Suddenly Stop Working

CUDA Toolkit, Compute Capability, and drivers look like the same thing on the surface — but they aren't. Why new NVIDIA GPUs still run (or don't), and how to make sense of the usual errors.

January 20, 2026 by Manuel Spörer

KICUDAGPUNVIDIAPyTorch
Abstract illustration of DGX Spark and RTX BlackwellAbstract illustration of DGX Spark and RTX Blackwell

DGX Spark vs. RTX 5090 vs. RTX PRO 6000 Blackwell: Which NVIDIA Platform Fits Which AI Workloads?

Three Blackwell systems, three completely different use cases. An overview of the DGX Spark, RTX 5090, and RTX PRO 6000 — and what each is really built for.

December 15, 2025 by Manuel Spörer

KIHardwareGPUNVIDIABlackwell