Notes from inside the build. What the engineering actually costs, where the received wisdom fails, and what I would do differently. Every example comes from a system that exists.
New posts on AI systems, engineering craft, and lessons from building in production. No spam. Unsubscribe anytime.
Loops, subagents, orchestration and routing, explained without the hype, plus the economics nobody mentions. Part six of six: what this site actually runs, what it deliberately does not, and why more agents is often the wrong answer.
Generation gives you code. Feedback loops turn it into engineering. Part five of six: the failures on this site where a green build was exactly the wrong goal, and how a bug becomes a permanent check nobody has to remember.
If your agent can use tools, what stops it doing something stupid? Part four of six: why prompt instructions are not a security control, the database rules that make bad outcomes impossible on this site, and the gaps I have not closed.
The assistant on this site answers from my actual writing. Part three of six: RAG explained without the jargon, the chunking and search settings I actually use, and three retrieval bugs where the model was fine and the information was not.
An agent that can only edit files makes you the courier between it and everything else. Tools and MCP fix that. Part two of six: what a tool actually is, what I connected, what it fixed in thirty seconds, and the permissions file I am not proud of.
Most people use AI like a vending machine. I used it like a senior engineer sitting next to me, and built this whole site that way without knowing Next.js. Part one of six: what actually separates a chat window from an agent, and how to start.
The job hasn't disappeared. The mental model has changed. Here's what the roadmap looks like for data scientists and engineers entering (or navigating) the AI agent era, with a South African lens.