Architecting Agentic Development Platforms
summary
Three-layer architecture converts an IDP into an Agentic Development Platform: the tooling layer (GitHub/GitLab, Jenkins/Argo, Terraform/Kubernetes, Datadog), a path-definitions layer with probabilistic, deterministic, hybrid paths, and an agent infrastructure with a harness plus governance (identity, secrets, observability). Demo shows an open-source ADP starter kit, a probabilistic PR review that posts comments only, and a hybrid validate-change loop where an agent reads CI failures, retries under a bounded budget, and claims near-instant results versus human-hours.
AI-written summary · about
chapters (12)
- 0:00Agentic development platforms intro
- 3:19AI alone isn’t enough
- 5:04IDP to agentic platforms
- 6:08Maturity levels of agentic software
- 8:36Three layers of agentic platform
- 13:12Agentic paths explained
- 14:37Eight agentic development paths
- 17:07Building agent infrastructure
- 18:51Governance for agent safety
- 20:35Demo: PR review paths
- 26:27Key takeaways and plan
- 28:32Final insights and books
related talks
26mArchitecting Agentic Development Platforms
30mCreating an IDP for AI Engineering - Max Körbächer
30mAgentic System is the New Full Stack for Developers
17mFrom Silos to Synergy: How Platform Engineering Bridges Teams with AI
44mAgentic Delivery, Guardrailed: AI in CI/CD & Platform Engineering (Open Source Patterns)
26mAgentic DevOps in the Autonomous Cloud
26mLet the Platform Build Itself: Using AI To Construct an Internal Developer Platform With CNCF Tools
06mThe End of Localhost: A Platform Team's Guide to Securing Agenti...