Applied AI in 2026: everyone’s using it, almost nobody can prove it works
Models got dramatically better in 2026. Business results mostly didn't. A practitioner's map of the tools, frameworks, and open questions in that gap.
A practitioner’s blog covering his experience building the “intelligent” enterprise — From AIOps pipelines to AI-driven supply chains and service life-cycles, where technology meets the shop floor.
Fresh tutorials and insights, published regularly.
Models got dramatically better in 2026. Business results mostly didn't. A practitioner's map of the tools, frameworks, and open questions in that gap.
AI hasn't made knowledge obsolete — it's made shallow knowledge worthless and deep domain expertise priceless. Here's why business analysts are the most valuable people in the room right now.
Last time I made the case for ML over rules in warranty fraud detection. This time, I'm exploring what happens when you point Karpathy's autoresearch at the fraud model itself — and let an LLM do the tuning overnight.
ML-driven fraud analytics is reshaping how aftermarket businesses detect warranty fraud—quietly, continuously, and far more effectively than any rulebook ever could.