Dylan Ratcliffe

And that was the first feature that gave us sort of commercial traction.

And it was still like the original product is still the underlying thing, but it's now pretty much entirely in the back end.

And it serves the LLM and the LLM serves everything else.

At this point, we were now basically an AI native company, like none of the things that we did we could do without AI.

It meant that you can't build a two year roadmap, especially when GPT 3.5 was just coming out.

Like every six weeks, the costs would halve and the performance would double.

And so we continued to do that in-person event.

and then work for three months and then in-person event and work for three months.

And our roadmap was very driven by those in-person events.

We would do a whole bunch of engineering proof of concepts to work out, okay, what is now possible?

And invariably what we achieved each time would have been impossible to achieve the previous time because it was just coming forward so quickly.

And so you can really trace the product and the company and the roadmap to those in-person sessions where we would all make a decision on, okay, we've got a proof of concept.

We know it works.

In the next three months, we're going to turn it into a product.

We're going to ship it.

And then we're going to meet up again and make that same decision again.

Not to say that we don't have an overarching goal.

direction of where we want to go.

I want to remove the fear of pressing that deploy button.

I want us to be better than a human at making those decisions.