Coding agents have moved past the demo stage. Teams now face a harder question. How do you govern, scale, and maintain software when an agent writes much of the code?
This XConf 2026 playlist on software craftsmanship in the AI era brings together four talks that look past unconstrained code generation. They show how teams govern AI assisted delivery with harnesses, preserve context across maintenance and delivery, and evaluate non-deterministic systems once they hit production.
Watch the playlist below to scale software craftsmanship in the AI era.
Inside this playlist
Code is cheap. Now what?
Coding agents are changing individual developer workflows and reshaping how engineering organizations make decisions. This session covers how teams are rethinking legacy code treatment, architecture design and build versus buy choices now that generating code is no longer the bottleneck.
Engineering the harness
Smart models still ignore team conventions when left unguided. This session shows how to build guides and sensors around AI coding agents, drawing on both brownfield multi repo environments and greenfield agentic workflows. You walk away with a checklist to audit your own AI setup.
AI-first software maintenance & delivery
A maintenance team and a delivery team working on the same core service, with multiple downstream consumers, cannot afford to operate as separate systems once AI enters the loop. This session breaks down how to connect them into one AI first system. You get shared triage, data correlation across teams, clear cross team communication and clean context handoff from one team to the next.
Evals are the new tests
Treat evaluations as the modern equivalent of software testing. This session walks through a real world deployment at BMW, where an AI defect analysis system uses evaluation driven development for summarization, data extraction and duplicate detection, combining offline benchmarks with live field feedback to deploy with confidence.
Why you should watch
Software economics when code is cheap
Understand how near-zero generation costs change build versus buy trade-offs, legacy refactoring strategies, and where architectural boundaries deliver the highest return.
Steer and control coding agents with harnesses
Deploy pre-execution guides, downstream syntax inspection, and a multi-repo audit checklist to keep agent-written changes aligned with existing production patterns.
Bridge incident triage and delivery
Connect production maintenance directly into active feature workflows so teams can resolve defects faster while preserving institutional system knowledge.
Build evaluation loops for production AI
Discover how BMW tests duplicate detection, extraction, and summarization pipelines using offline benchmarks before deployment and continuous production feedback after release.
Meet the speakers
Let's stay in touch
Get notfied for future events, updates and related information.