ABOUT
Robbie Smith
Technology architect. Systems builder. I like understanding why things work, why they fail, and how to stop useful learning disappearing.
HOW I GOT HERE
I did not start in technology.
I started my working life in Ireland as an apprentice bricklayer. Later I went to university at night to study computing and business while still working. That eventually led to my first job in IT, then consulting in London, and later a move to Australia.
My career became progressively more technical and more architectural. I spent years working with Salesforce, enterprise platforms and energy and utilities, eventually leading architecture across large delivery teams, setting standards and patterns, and running design authority for a shared enterprise platform.
I still think the same way I did when I was learning a trade: understand the structure, know what carries the load, do not hide a bad foundation, and make the next person's job easier.
THE WORK
A lot of architecture happens before code.
The hard part is often understanding what the business is actually asking for, finding the constraints that matter, deciding where a standard is needed, and making the important decisions difficult to lose.
I have spent more than 15 years working with Salesforce and years in energy and utilities. I have worked across architecture, integration, delivery governance and design authority, including leading standards and patterns for teams of around 30 people.
I care about systems that teams can actually build and operate, not architecture that only looks good in a diagram.
WHY AI
AI gave me a new version of an old architecture problem.
LLMs can reason through difficult problems at remarkable speed. They can also forget context, repeat mistakes, sound confident when they are wrong and make poor substitutes for durable state, exact rules and independent proof.
That tension interests me. I became less interested in trying to make the model do everything and more interested in engineering the system around it: better context, deterministic controls, retained learning, verification and evidence.
That is what much of robbie.build documents. Not polished claims. The thinking, building, failures, repairs and lessons behind the work.
HOW I THINK
Explain before impress. Evidence over noise.
My instinct is to understand the system, make the relationships visible, test the claim and retain what we learn. I am suspicious of answers that cannot be checked and of process that exists only because nobody has challenged it yet.
If a script can know the answer, I do not want an AI guessing it again tomorrow. If an agent solves the same problem twice, I want to know why the first lesson was lost.
I would rather show the failure and what changed because of it than pretend everything worked first time.