AI
What Large Language Models Are Good At, and Where They Fail
Useful adoption starts with an accurate model of the tool. LLMs are excellent at some things and confidently wrong at others.
Writing
Blog
Guides, updates and the occasional opinion.
AI
Adding AI to a Product Without Wasting a Quarter
Most failed AI features fail for product reasons, not technical ones. Start where the value is obvious.
AI
Retrieval-Augmented Generation, Explained Plainly
RAG is a technique for answering questions from your own documents. The idea is simpler than the acronym.
AI
Running Smaller Models on Your Own Infrastructure
Self-hosting a model is more achievable than it was, and worth it for specific workloads.
AI
Prompt Engineering Is Mostly Specification Writing
The techniques that reliably improve output look far more like requirements than tricks.
AI
Using AI for Customer Support Without Annoying Customers
Automated support succeeds or fails on one design decision: how quickly a person can take over.