A conference talk transcript from a Microsoft developer on the OS internals team covering how to use AI coding tools more effectively, particularly for C++ development. The speaker breaks down the AI workflow into three components: models (the LLM brain), context (mutable per-request vs. immutable project-level), and frameworks (IDE extensions, CLI tools). Practical demos show how enriching context via memory banks and MCP servers dramatically improves AI output quality — using a C++ logging task as the running example. The talk also touches on AI non-determinism vs. language features like C++26 reflection, and offers tips for iterative improvement of AI instructions.
•54m watch time
3 Impressions