Deep dives into ML engineering, model deployment, and building AI products that actually work.
Build not buy has been my motto for a year. A skate sharpening tracker, a failed sentiment app and a PDF compressor made me question what small apps are really worth.
There is a growing myth that customising a large language model requires a research budget. Here is what the numbers actually say.
Most operators can detect faults and deflect calls with chatbots. The real opportunity, new revenue, platform economics and sovereign infrastructure, is still largely untouched.