About
I'm Mohamed EL HARCHAOUI, an AI Engineer exploring the full stack of intelligence—from agentic systems and tool use to world models and physical AI. I spend my days thinking about how to build more capable, efficient, and aligned intelligent systems, with a particular focus on inference optimization, memory architectures, and long-term autonomy.
This site is where I work through ideas in public—technical explorations of generative AI, inference-time thinking, and robot learning, alongside reflections on the broader paradigm shifts reshaping our technological and economic landscape.
I write for engineers and researchers who believe that understanding the fundamentals matters more than chasing hype. The field moves fast, but clear thinking compounds.
My domains of interest
- — Generative AI: Agentic systems, tool usage, MCP, skills, memory, prompt & context engineering, evaluation
- — AI/ML: Large language models, architecture optimization, memory, continuous learning, thinking approaches and inference-time thinking improvement
- — Robotics: World models, physical AI, robot learning, planning and long-term autonomy
- — AI inference: Architecture optimization, attention optimization, pruning, caching, long context and memory
- — AI & the future: Paradigm shifts, societal & economic implications, and the long-term trajectory of intelligent systems