An AI that knows who it works for.

Every build starts in the same room. We write down who you are: what you value, how you decide, what you sound like, where the hard lines sit. That document becomes the constitution your AI runs on, and everything else obeys it. On top of it comes the product blueprint, the memory design, and an architecture brief your own engineers can build against, because the plumbing was never the point. The perspective layer is the point. That's the product: an AI that answers the way you would, holds your standards when you're not in the room, and gets more useful the longer it lives with you.

The reference implementation is mine. I've run my own OS daily for over a year: local inference, constitutional rules, sixty-odd agents acting on my behalf, and it is how this site got built.The pattern has transferred twice so far: first to a behavioral scientist who runs her practice on it, then to a B2B marketing firm in the US, where it became the firm's perspective layer sitting on top of client brains their own developers build.Each build looked different on the surface and identical underneath, which is what a pattern is.

Your team keeps the infrastructure: pipelines, vector stores, servers, keys. I bring the product mind: what the system should know, how it should decide, what it must refuse, and how you'll know it's working. If what you're after is coaching on how to work alongside AI rather than a system of your own, that practice lives at werkanders.com.

If you want one of these, for yourself or for your firm, write to me. Tell me who the AI would work for. That's the whole brief.