gstack is rigorous once evidence exists. Before code, customer discovery skills and a reliable memory layer are needed to produce and preserve the proof it expects.
AI-assisted software, built with discipline
Building serious software with AI, without outsourcing judgment.
A public learning log about architecture, trust boundaries, reviewability, state, and product judgment in AI-assisted software development.
I come from enterprise procurement and IT governance, and I write about what changes when that judgment meets hands-on software building with AI.
Recent notes from the work
Articles
A Hermes personal shopper exposes a durable architecture boundary: the agent runtime may produce decisions, but canonical product state needs a life outside it.
Agent memory is a scoped service layer, not runtime state owned by an agent instance.
How to test whether a layered agentic workflow preserves context, responsibility, verification, and recovery for a solo developer.
Why an agentic factory is an architecture of responsibility, evidence, and boundaries—not a stack of tools or a more powerful coding agent.