The AI ecosystem can be really confusing: Agents, MCPs, skills, plugins, everything is an overloaded term and a fuzzy concept. This is an explanation of how these constructs work in plain terms, useful if you're an engineer wanting to go from passively consuming them to genuine understanding.

First-draft AI skills are usually broken in ways the author can't see. A walk through how I built our daily AI spend report skill, with iteration discipline borrowed from Atul Gawande's Checklist Manifesto.

The year AI SRE arrived. 1,152 PRs, a team that grew from 3 to 18, and the hardest I've ever worked. From betting on tooling in January to shipping a product that makes customers say "how did you figure that out?" by December.

What does switching to AI Engineering actually mean for your career? Drawing from my experience pivoting to SRE, I offer an honest assessment of the opportunities and challenges for software engineers considering this move—from high-impact, high-pressure work to the realities of slower progress and less visible product surface.

Working in AI today, I'm seeing the innovator's dilemma play out in real time. While larger companies carefully plan deployments that work for their entire customer base, smaller teams like ours can ship, learn, and improve our AI products through actual usage. This isn't just about moving faster—it's about fundamental advantages in how AI products develop that favor startups, regardless of the resources incumbents can deploy. The dynamics surprised me, and they might surprise you too.

I've met teams who switched to Python just to build AI features, abandoning their normal stack for the ecosystem. But it's really not worth it! At incident.io we stuck with Go and it's been great - turns out static typing and proper concurrency are exactly what you want when building AI systems, provided you build some nice abstractions to go with it.

The gap between demo-ready AI products and production-grade systems is much larger than most realise. This post explains the four stages of AI product maturity, what tooling you actually need to build reliable AI systems, and how to recognise if you're stuck in the MVP trap.

From reliability engineering to wrestling with LLMs, my fourth year at incident.io pushed me harder than I'd expected. We launched On-call, weathered some tough times as a team, and I ended the year diving fully into AI.

A story about how incident response training went wrong, with valuable lessons about pod priorities, isolation, and the importance of a healthy incident response culture.