Skip to content
back to talks
podcast discussion · AI Dev Podcast · TechTrain

AI Dev Podcast #8: Governed Agent Autonomy

July 27, 2026 · AI Dev Podcast · TechTrain

An AI Dev Podcast episode with Vladimir Yatulchik and Alexander Polomodov on moving from vibe coding to a governed AIDLC process: decomposing development into agent skills, keeping engineering context next to the code, and constraining autonomy through ADRs, traceability, security reviews, and process health checks.

TechTrain channelWritten summary
Share
TelegramLinkedIn

Participants

Episode focus

  1. 01Why do autonomous agents change unrelated code, and how can experiments become a repeatable process?
  2. 02How can teams adapt AIDLC to internal infrastructure and express the SDLC as a set of agent skills?
  3. 03Why move user stories, architecture decisions, and working documentation from Confluence into Git?
  4. 04How do ADRs, traceability, security reviews, and health checks keep an agent aligned with intent?
  5. 05Which responsibilities remain with product managers, analysts, and engineers when agents generate code and infrastructure?
  6. 06Which results have small projects achieved, and what blocks the same approach on a large brownfield system?