AI augmented development makes it easier than ever to generate code, but that does not mean it has become free. Every repeated prompt consumes context, and is non deterministic. Every generated variant needs review, and every inconsistency becomes future maintenance. For repeatable patterns, C# source generators offer a different approach: codify once and let the compiler produce the code for you.
In this talk, we’ll explore how to build practical source generators that solve real world problems, not just hello world examples. We’ll look at where source generators fit, how they work, and how they can reduce boilerplate for tasks such as mapping, registration, serialization support, strongly typed accessors, or repetitive glue code.
We’ll focus on the trade-offs that matter in production: keeping generated code understandable, designing a useful developer experience, debugging generator output, avoiding over-engineering, and deciding when a generator is worth the investment. Along the way, we’ll connect this to AI-augmented development: when should you ask an agent to generate code? And when should you turn the pattern into a tool? Of course we'll use AI augmented development to manufacture the actual generators, it's 2026, duh!