Our AI journey didn’t begin with ChatGPT
Long before generative AI entered everyday conversation, we were building neural networks and our own models. Here is how that history shapes the way we engineer today.
When ChatGPT arrived, it felt to many people as if artificial intelligence had appeared overnight. For us, it was a turning point in a journey that had already been under way for some time.
Before the boom
Our founders were experimenting with machine intelligence and building neural networks well before generative AI became part of everyday conversation. That curiosity turned into a product.
HatchFast’s first product was a test-automation tool. Engineers arranged visual blocks on a canvas, or recorded the way a user moved through an application, and the tool generated working Selenium and Playwright automation code from it. It was practical, unglamorous work, and it taught us a great deal about building software that has to keep working when the applications around it change.
Teaching the product to read
The product then evolved so that test cases could be written in plain language. To make that possible we developed our own models, and adapted and fine-tuned open models, including Llama, for the job.
This was the period that shaped our instincts about AI. Models are only as useful as the context they are given. Reliability is designed, not assumed. And the interesting engineering is rarely the model itself: it is everything around it.
What changed, and what didn’t
ChatGPT didn’t start our journey. It changed what was possible, and it accelerated what could be built. Capabilities that once demanded custom models and months of work became available to anyone with an API key.
In 2025 we made a deliberate change of our own. We moved from building a single product to technology engineering services. It was an evolution of our business model, not a retreat from building technology. The original product has since been retired, but the experience it gave us came with us: in our AI practice, in how we approach automation, and in the products we now build as ventures.
Why it matters to the people we work with
Because we have built and run AI in our own product, we know where it is genuinely useful and where it is not. We ground systems in an organisation’s own knowledge. We keep people accountable for the decisions that matter. And we plan for production from the start, rather than after the demo.
Our AI journey didn’t begin with ChatGPT, and it won’t end with whatever comes next. We intend to keep engineering what’s next.