Last week, thanks to David Morgan-Brown and Patchwork Health, I had the opportunity to do something I’ve genuinely missed: presenting in front of a live audience. The subject was orchestrators, the latest buzzword dominating discussions about AI-assisted development.
A prominent example is Gas Town. Since its release last Christmas, its popularity has (predictably) ‘hockey-sticked’ and it has become a reference point for the tools that followed.
I have attached the presentation slides below. The first part provides a recap of the architectural steps that moved us here. We trace the path from small agentic loops, through Ralph (July 2025, which feels like an eternity ago!), to the Orchestrator itself.
The second part focuses on the business impact. I don’t expect orchestrators to remain optional for long. As tools such as Claude Code add more of the necessary building blocks, I suspect this way of working will find its way into many software companies. Much like early LLM adoption, if you believe nobody on your team is experimenting with it, you may be mistaken.
We are venturing into new territory. I don’t believe we have established best practices for these advances yet. I’ve listed some emerging patterns in the slides, but the balance has shifted. There is much more code and, consequently, more risk. I doubt some traditional best practices will survive this change, and I expect new ways of working to be challenging in some environments. Not to mention those still trying to implement Agile practices from 2005…
New practices will be a major theme for us to tackle in 2026. I’m very curious about the outcomes.