AI as a useful addition to the engineering toolkit — not a replacement for thinking.

I’ve been spending more time with AI tooling — using it for development, research, and exploring solutions. Context engineering and prompt design are things I’ve picked up through regular use: learning what to feed a model, how to structure a conversation with it, and when to ignore its suggestions entirely.

Where I find it most interesting is on the engineering side: wiring models into products and building systems where AI output is actually reliable. That means thinking about token limits, context windows, fallback strategies, and the glue code that sits between an LLM and a real product.

I’m not an AI researcher or a model trainer — my background is software engineering. But I’ve found that experience useful for the practical side of AI integration: making it work in production, knowing its limits, and keeping expectations honest.