Agentic System is the New Full Stack for Developers
Talkat PythonAsia 2026 · on Saturday, 2026-03-21· at 11:05· Teresa Yuchengco Auditorium (Main Hall)·
summary
Argues that an agentic layer sits on top of event-driven cloud stacks and delivers goal-driven systems that plan, call real tools, act autonomously and adapt on failure. Demonstrates the Python Strand Agent SDK creating agents with a few lines, using tools like calculator, HTTP, file/memory and AWS integrations, and running any model provider. Describes three workflows—agent-as-tool orchestrators, graph parallel experts, and swarm collaboration—plus spec-driven development via Kirro/ko and observability via token usage metrics and tool-failure traces.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:31Full stack evolution
- 4:24What is an agentic system
- 8:19Agentic systems in Python and cloud
- 9:22Python agent toolkit and MCP
- 11:52Strands agent SDK and loop
- 15:12Agent tools and multi-agent patterns
- 18:24Multi-agent workflows
- 22:02Spec-driven development with Kurro
- 25:19Deploying to production
- 27:50Key takeaways and future
related talks
27mAgentic Systems: Leadership, Risk, and Accountability
06mKeynote: Building Agents at Scale
44mBuilding Agentic AI Systems: From Transactions To Conversations
18mFrom Reactive to Resilient: Agentic AI
24mAgentic Al in SW Development: Evolving Patterns & Protocols
20mReimagining InnerSource for the Agentic AI Era
19mBuilding Agents at Cloud Scale
31mAgentic AI architecture patterns