(The event URL will be provided later to approved RSVP attendees.}

In my experience using coding agents for more than a year, and Claude Code for 10 months, payoffs come from small, deliberate, incremental steps. On this journey, part of my challenge has been dealing with the low maturity of coding assistants—until very recently. Even with incredible advances in LLM capabilities, generating the domain model within a Domain-Driven Design context boundary is still quite disappointing. It's rarely better than unacceptable. Even when I provide significant guidance, the agent quickly and autonomously falls back to CRUD. In case you are unaware, CRUD is the opposite of a rich, behavioral, domain-driven software model.

It all comes down to the lack of well-designed domain models openly available for use as training data.

So, what's a poor software modeler to do? I'll share how I've overcome frustrating LLM limitations and constant rework using practical agent guidance that makes me smile.