RAG at work: why document access does not guarantee a good answer
Evaluate a RAG assistant against a concrete task: product revisions, bulletins, permissions, answer completeness and citations in a reproducible teaching example.
Artificial intelligence
Models, data and evaluation: from AI fundamentals to computer vision and product applications.
The model is 95% accurate. Should it make the decision?
Choosing an AI decision threshold around error costs and inspection capacity: a reproducible example, calibration, deferral and acceptance criteria.
YOLO26 in practice: from object detection to system decisions
What does YOLO26 actually return, what changes compared with earlier models, and how do boxes on an image become a useful system? Real deployments, misleading metrics and alternatives worth comparing.
AI vs ML: how do they differ, and where is the boundary?
From Deep Blue to language models: what separates AI from ML (machine learning), what learning actually changes, and why 99% accuracy can conceal a useless detector.
