Blog
Some of my thoughts and things I'd like to share.
Featured, , 8 min read
How LLMs actually work: An intuitive guide
LLMs work by predicting the next token, using embeddings to map meaning and attention to read context, then building a full answer one token at a time. Training on massive amounts of text, followed by fine-tuning for helpfulness, turns that simple loop into a capable assistant, which also explains why these models sometimes hallucinate.
, 5 min read
How RAG (Retrieval-Augmented Generation) works
RAG lets an AI answer questions about your own documents by finding the most relevant passages first and handing them to the model with the question. The model is the easy part; the quality of your chunking, search, and data freshness decides whether you get accurate answers or confident guesses.
, 6 min read
AI and LLMs 101: A practical starter guide
This guide explains the core vocabulary of modern AI, including prompts, tokens, agents, MCP, skills, plugins, and Claude Code, in plain language. It ends with five practical steps to start using AI on real work today.
What Ontario clinics need to know about digital health rules in 2026
, 4 min read
Ontario clinics now face direct privacy fines, new oversight expectations for AI scribes, federal rules against vendors locking up patient data, and a push toward a shared provincial health record. To stay compliant, clinics should check their audit logs, AI review processes, data export capabilities, staff access controls, and where their patient data is stored.
The end of the handoff: Why product and engineering need to merge
, 4 min read
AI has made building software cheap, so the real bottleneck is now knowing what to build, and the old handoff between product managers and engineers wastes time, loses intent, and creates technical debt. Product-engineering fusion, modeled on the forward deployed engineer, puts business judgment and technical execution in one owner, producing simpler, compliant, and more usable systems faster than a traditional siloed agency.