OnDynamics Labs
Where the patterns get stress-tested before they reach your programme.
OnDynamics Labs is where I run my own research projects, on my own infrastructure, so the agentic AI patterns I bring into an enterprise programme have been tested somewhere I can break things safely first. Two projects are public today.
Why a consultant runs labs
The patterns that survive here are the ones I bring to clients
It is hard to advise on agentic AI credibly without building it. The Labs are where I do that, on the same Microsoft stack a client programme runs on, but in conditions where a failure costs me a weekend rather than costing an enterprise its quarter. The orchestration, evaluation and observability patterns that hold up in the Labs are the ones that end up in a client architecture; the ones that break, break here first.
If you cannot demonstrate it on your own platform, you should not be designing it on someone else’s.
The projects
Two projects, one discipline
Self-evolving agents
Project EVE
A self-evolving multi-agent platform built on the Microsoft Agent Framework. The place I test multi-agent orchestration, eval-first development and observability before they reach a client architecture.
Your chief of staff
Project Blue Sky
An AI chief of staff for everyday life, a butler agent that coordinates a team of specialist agents around a person’s goals, projects and commitments. The lab for agentic AI in the personal domain.
Curious about the Labs work?
If you would like to see Project EVE running, or talk through how the Labs patterns might apply to your own agentic AI programme, I am happy to walk you through it.