AI Codecon: Software Craftsmanship in the Age of AI

The tools have arrived. The hype has settled. Now comes the hard part: building software that actually works. The best developers are discovering that building with agents doesn't replace the need for craftsmanship; it demands new forms of it. Agents can autonomously debug code, coordinate across multiple tools, and handle complex workflows that would have previously required careful orchestration.

Teams shipping these systems are developing new disciplines including context engineering and agent engineering. These new approaches combine the rigor of traditional software development with new approaches to observability, iteration, and reliability. When your system can reason and plan, you need different methods to make it dependable. And you need to be able to communicate context efficiently. The best systems fuse the strengths of LLMs with the strengths of traditional programming—for example, by using MCP or Agent Skills to integrate existing code into workflows via tool calls rather than code generation.

Agent quality comes from shipping early, observing what actually breaks in real usage, and refining prompts, tools, and architectures based on what production teaches you. The teams shipping production agents successfully aren't just writing better prompts or chaining more tools together. They're combining engineering rigor, product thinking, and data science intuition to shape agent behavior through continuous iteration.

The future of software development won't be written by AI alone. It will be crafted by developers who know how to build AI into systems that are legible, modifiable, composable, and shareable, the core properties that made open source transformative.

Please note that slides or supplemental materials are not available for download from this recording. Resources are only provided at the time of the live event.