The Great AI Arbitrage
Technology, not viability, is the reason for disillusionment with generative AI.
Est. 2022 · Munich, DE
Essays on AI product development, ML systems, the gap between research and enterprise, and many other things.
Recent essays
The Rift Between Open-source and Commercial Software Is Not Irreconcilable
The notable difference between commercial software and FOSS isn't quality — it's the culture of being product-focused versus one of sharing.
Data Quality Metrics for Everybody
Training an ML model involves a lot of reasonable assumptions; diagnosing the inevitable reality-check starts with a basic set of data-quality metrics.
The AI of Your ML Products Can Also be an Attack Surface
Security in ML systems spans training data, trained models, and inference endpoints — and the field moves faster than researchers from adjacent disciplines can track.
Who are our Review Meetings for?
An ode to pragmatic scrum — and a question about whom our review meetings actually serve.
Most Data Fabrics are Made Of Lead, not Linen
Most businesses still need a semantic engine on top of their original sources before any data-fabric promise makes practical sense.
Why Aren't Data Behemoths Crushing It Harder?
It isn't enough that historical data is digitalized and archived — what matters is whether the data is activated.
Metaphors in Technology
For a discipline whose fundamental tools are languages, software engineers and statisticians are notoriously bad at putting metaphors to work.
Innovations and Innovations
A paraphrased quote about types of innovation — why some move global productivity and others just replace window units with central AC.
Subscribe
Get new essays by email.
Powered by Buttondown.