In early 2025, in the founder community, you couldn't swing a cat without hearing about the billion-dollar solopreneur. Using the power of AI and a great idea, they could build and service a SaaS company on their own.
Loads tried. And basically we ended up with a bunch of AI slop: products that looked amazing, but when you scratched at the surface the gold came off and all the security holes and poor architecture decisions were laid bare.
By the summer the scene went quiet. Most people realised that you needed expertise in HCI and software engineering to hammer these products into something that could actually work.
By the autumn all the fanfare had died down and the vibe coding hype moved on to the next big thing.
That's not the full story though, because working behind the scenes, away from the content creator hype that surrounded vibe coding, some pioneering engineers and designers were pushing boundaries. They were using their domain expertise to collaborate with AI to build great products, fast.
For most of 2025 this was frustrating work. The models weren't good enough, the output was poor, and it mostly felt as if it was quicker to do the work yourself.
Then, and I've written about this before, November 2025 hit and everything changed. Opus 4.5 and Codex 5.2 came out and made October's tech seem like relics of a bygone era.
These pioneering engineers started collaborating on creating specs and then having the AI write the full end-to-end code. And what was produced wasn't rubbish. It was actually pretty bloody good.
They recognised the shift and have started to build the professional practice of AI Engineering: domain experts working with AI to deliver new products.
So is this still vibe coding?
Yes and no.
Vibe coding was characterised by inexperienced people relying on AI to build products for them. The inputs were vague prompts and good intentions, and the outputs were fragile products that couldn't survive contact with real users.
AI engineering is different. It still uses AI to write code and design screens, but it relies on domain experts, solution architects, and UX professionals to drive the product strategy. The inputs are proper specifications, informed by years of experience. The AI handles the execution; the humans handle the thinking.
Why this matters
This distinction matters because it means the speed gains from AI aren't going away. They're just moving into the hands of people who know what to do with them.
For businesses, this is significant. It means you can move from concept to working product in a fraction of the time, without sacrificing the quality and robustness that your users expect. It means smaller, more senior teams can deliver what used to require large departments. And it means the competitive advantage isn't access to AI (everyone has that) but the depth of expertise you bring to the table.
The era of vibe coding was a false start. AI engineering is the real thing.
If you're looking to accelerate any of your programmes, come speak to us about how our domain experts can work with yours to build products that your clients will love, quickly and reliably.
