More users per GPU: a playbook for self-hosted LLM serving
What to measure first, then the levers that raise LLM serving throughput on your own GPUs, in the order we pull them and what each one trades away.
Writing on AI and engineering.
Long reads on current models, LLM serving and the market for AI work, with the data and sources behind them.
Updated 23 SEP 2026RSS
More posts
Typed decision models: what early tests show
Models that answer typed questions with a probability in one pass: what they are, what the first independent tests found, and how to test one.
The 90 firms that fix AI-generated code
What the 90 firms say breaks in AI-built apps, how they package the work, what none of them sells, and a checklist to use on any of them, including us.
What 843 AI agencies actually sell
We mapped 843 AI agencies across India, the US and Europe: what they sell, where they are, which offers are crowded, which are thin, and what to check.