Well, hello there, AI world. The craic, as we say back home, is absolutely mighty these days. For those who don’t know me, I’m what you might call a “leading edge technologist.” It’s a fancy term for a geek who’s spent a lifetime tinkering with the next big thing, always with an eye on how to bring a bit of automation and self-service into the enterprise, preferably before the coffee gets cold.

I’m an Irishman, born and bred, who followed his heart and a Welsh girl to the UK about 20 years ago. My career has been less of a ladder and more of a wild ride on a rollercoaster that’s missing a few bolts. I’ve been a field service engineer for HP, driving the length of Ireland to fix a server in Kerry with surfboards strapped to the roof, trying to squeeze every last drop of fun out of a work trip. I’ve managed production lines for the Motorola StarTAC phone (if you remember that first flip phone, you’re officially old enough to groan when you stand up). And most recently, I was the Director of Customer Success for EMEA at HashiCorp, where I had the thrill of helping grow the business from a plucky $30 million to over $300 million before IBM came knocking. We went from being pre-IPO paper millionaires to post-IPO “ah well, nearly!” It’s been a blast.

Through it all, my passion has been the hands-on roles where you get to see the novel and creative ways customers use technology. I blog intermittently, partly to share what I learn, but mostly to stop myself from forgetting it. Anyone else find that knowledge evaporates if you haven’t used it in two months? My blog is a treasure trove of ‘crib sheets,’ the distilled essence of years of training, deploying, and having a good chinwag about the nuts and bolts of tech, poured with the same care as an impeccable pint of Guinness.  

I pride myself on being a professional, always giving my customers 130% (you know, 10% on Monday, 50% on Tuesday…). But there’s a bit of the Loveable Irish Rogue in me too. My glass is always half full, and I hope that optimistic, happy-go-lucky outlook comes through this week. Because, let me tell you, I haven’t been this excited about technology since the OpenStack and public cloud revolution. That was the last time the playing field was truly levelled. It’s happening all over again with AI, and the big players are nervously trying to tame this wild beast.

What is “AI Engineering” Anyway? My Take from the Trenches

Before we dive in, let me clarify what I mean by “AI Engineering.” I’m not talking about Machine Learning or Data Science. I’m not trying to build my own AI model from scratch. As an early adopter, I’ve been personally paying for access to LLMs for the last three years, long before most companies even knew what to do with them.

My wife still thinks I’m regressing to my preschool days. I tried to explain that the model trained itself on pictures to tell the difference between a cat and an apple. Her response? “The next time you need to know if you’re eating an apple or a cat, please just ask me instead of spending the price of a family sun holiday on a computer in the garage that knows fruit!” Fair point.

But the evolution is finally here, and AI is now for the masses. I look at it from an engineer’s perspective. My job is to help customers onboard new products and make sure the solutions are secure, integrated, and ready for production. It’s about making things work. When I studied AI at Trinity College Dublin in the 90s, it was basically a glorified LISP module. We’re finally at the place I dreamed of back then, a vision I probably owe to Captain Kirk and his talking computer.

The Class of August 2025: A Cambrian Explosion, a Polished Update, and a Wobbly Launch

The reason for this blog series is the absolute whirlwind of this past week. We’ve witnessed a Cambrian explosion of new AI models, but two launches in particular tell a fascinating story.

First, on August 5th, Anthropic released Claude Opus 4.1. This was a classic, professional tech update. It’s a drop-in replacement for their previous top model, offering better performance in coding, agentic tasks, and reasoning, all for the same price. It’s a solid, incremental improvement that makes a great product even better. GitHub is already reporting notable gains in multi-file code refactoring. Clean, simple, effective.  

Then, just a couple of days later, OpenAI launched GPT-5. And oh boy, what a launch it was. The promise was “expert-level intelligence” from a new unified system that cleverly routes your prompt to either a fast model or a “thinking” model for harder problems. The reality? Well, my first experience was less “PhD-level expert” and more “confused intern who’s had one too many espressos.”  

It turns out their much-hyped router was broken on launch day, a fact Sam Altman later admitted was a “sev” (a top-priority incident) that made the model seem “way dumber”. To add insult to injury, they yanked access to all previous models for many users without warning, forcing us onto the new, wobbly system. It was a masterclass in how to alienate your paying customers.  

So, this week, we have a perfect storm: a polished, professional release from Anthropic and a chaotic, powerful, and deeply flawed one from OpenAI. It’s the perfect backdrop to explore what AI engineering is really about.

Model Name	Developer	Key Differentiator	The Vibe
GPT-5	OpenAI	Unified reasoning & chat, “expert-level” intelligence (with a broken router)	Chaotic launch, powerful but flawed
Claude Opus 4.1	Anthropic	Drop-in upgrade for coding & agentic tasks	Professional, stable, impressive
Gemini 2.5 Pro	Google	Advanced multimodality (text, image, video, code)	The quiet powerhouse
Llama 4	Meta	Top-tier open-weight model, enables local deployment	The open-source champion
Grok 4	xAI	Integrated with X platform, real-time data access	The wildcard

My AI Journey: A Roadmap for the Week

This week, I want to take you on a three-phase journey through my eyes as an AI Engineer, coloured by the madness of the last few days.

• Tomorrow (Tuesday), we’ll kick off Phase 1: I’ll walk you through how I built craicgpt.ie, a simple serverless website, using AI as my co-pilot. We’ll look at the good, the bad, and the ugly commits.
• On Wednesday and Friday, we’ll move into Phase 2: Agentic Workflows. I’ll show you how to build an enterprise-grade RAG system, and then we’ll dream a little bigger with a vision for a proactive personal assistant I’m calling “Loraine.”
• On Thursday, we’ll tackle the hard questions—the elephant in the room. We’ll discuss the staggering financial and environmental costs of this AI revolution and its disruptive impact on our careers.

It’s going to be a packed week. So grab a coffee (or a Guinness) and join me on this adventure. The future is being written in real-time, and it’s a thrilling, if slightly chaotic, ride.