kekePower

My journey into AI began with the launch of GPT-3 and quickly turned into a full-time obsession with testing and understanding language models. From local open-source projects to cutting-edge commercial models, I dive deep into how they run, where they break, and how to make them work better-especially on consumer-grade hardware.

With three decades of IT experience and a strong foundation in Linux systems, I bring a practical, no-nonsense lens to a field often clouded by hype. I document hands-on experiments, benchmark models, and share insights from real testing-done on my own machines, with full transparency.

This blog is my way of making AI exploration more open, useful, and grounded-whether you're new to the field or running your own local stack. I focus on clarity over jargon, and results over theory.

Recent highlights include optimizing Qwen3 models for consumer GPUs and testing over 30 different LLMs in real-world conditions.

Expect a mix of technical deep dives, opinion pieces, and the occasional behind-the-scenes reflection. I’m not here to sell hype or push products-just to explore, explain, and share what I learn.