Writing
Mostly about production ML failure modes and what the measurements actually showed. Published in edited technical publications.
Towards AI · DataDrivenInvestor · NextGenAI · Python in Plain English
Selected
Every AI Tool Shows You Sources Now. I Checked Whether They Actually Say What the AI Claims
Towards AI
The AI Feature That Works in the Demo and Breaks in Production
Towards AI
I Built a Recommender That Learns Per Click. Then It Lost the Benchmark
Towards AI
I Built a Causal Inference Engine. The Best Thing it Does is Refuse to Answer
Towards AI
I Built a System that Automatically Rolls Back ML Models Before They Ruin Your Production Data
Towards AI
I Tested 5 RAG Search Strategies on the Same Dataset. Here Are the Real Latency Numbers
NextGenAI
More, live from Medium
- Ajo Has Run on Trust for Generations. Tokenization Could Let It Run on Proof Instead.Medium · Sep 26, 2026
- There’s A Name Now For The Bug I Kept Finding In My Own Benchmarks: Metric OverfittingMedium · Sep 19, 2026
- AI agents need their own identity system. I built one.AI in Plain English · Sep 16, 2026
- I built a database migration linter.DataDrivenInvestor · Sep 14, 2026
- I Built a Harness Across Six Tools Before The Industry Had a Name For itTowards AI · Sep 9, 2026
- Every AI benchmark I’ve published had a bug that would have flipped the headline.Medium · Sep 3, 2026
This list updates itself as new articles are published. Everything on Medium.