The subliminal reason for what I have been doing

He didn't solve it in 1979. The tools didn't exist.

He didn't solve it in 1999, when he was writing about neural networks applied to financial data in an MBA paper. The tools were closer, but still not there.

He didn't solve it in 2010, when he was in Victoria evaluating Oracle Policy Automation for British Columbia's Ministry of Children and Family Development. He was looking directly at the problem — eligibility assessment, encoded as rules — but the substrate could only express one program at a time, in one jurisdiction, with no path to citation, no path to audit, no path to the rest of the world's case officers.

He stopped coding in 2010. For fifteen years, he carried the question without the means to answer it.

In July 2022, at a Canadian federal department, he was asked to be the champion of an enterprise chatbot. Mid-flight, ChatGPT launched. He started demonstrating the art of the possible to his internal clients.

In July 2025, in a meeting with a vendor, someone asked how to deliver AI at scale. The answer came back: if it was easy, everybody would be doing it already.

On November 3, 2025, he started a GitHub Copilot Pro+ subscription. It was the first time he had written code in fifteen years.

On November 20, 2025, he read a vision paper out of Berlin called The Agentic State. The first click landed:

I know how to deliver governed AI at scale.

Shortly after, while refactoring Microsoft's open-source Information Assistant into what would become AIA, he encountered SPRIND's Law as Code framework, from Germany's Federal Agency for Disruptive Innovation. The second click landed:

I know how to deliver eligibility at scale.

Forty-seven years of pattern recognition met six months of AI-assisted building. That is the only window in history in which one person could have built this particular thing alone.

AIA — the Agentic Information Assistant is the assistant his father did not have. A governed retrieval system that draws on trusted sources and answers with citations, so that the next case officer staring at a file does not have to hold the entire corpus of law, regulation, and jurisprudence in their head.

GovOps — Law as Code is the engine that would have let the room of fifty officers spend their time on the cases that actually needed human judgment. Deterministic eligibility evaluation. Every dollar traceable to the statute that authorized it. Every rule change dated, cited, approved. Configure without deploy. Two reference programs: Old Age Pension and Employment Insurance — the same shapes his father adjudicated by hand.

AIA on AWS is the same assistant, on a different cloud, for the ministries whose substrate is not Microsoft's. Because a governance pattern that only works on one cloud is not yet a governance pattern. It is a vendor preference.

All three are open source — Apache-2.0 or MIT. All three are jurisdiction-agnostic. All three are federation-ready. None of them are any government's intellectual property. They were built independently, so they could be given away — to any ministry, anywhere — in Canada, Brazil, Germany, Spain, France, Ukraine, Japan — that runs a benefits program and staffs it with case officers.

His father died on May 24, 2026.

This work is for all government case officers, everywhere in the world, in homage to him.