Day 2 Online Workshops

What does it take to build systems that continue to function when connectivity is unavailable?

Modern applications often assume reliable internet access and cloud infrastructure. This talk follows the speaker’s ongoing exploration of local-first system architectures for disconnected and resource-constrained environments. This session will share experiments, design decisions, and lessons from exploring offline‑first, edge processing, and hardware‑aware systems.

Choosing an agentic AI framework is one of the most consequential architectural decisions a software team can make today. LangGraph, CrewAI, and AutoGen each promise to orchestrate intelligent agents, but their design philosophies diverge sharply. Most comparisons rely on toy examples and vendor documentation. This talk is different.

This session presents empirical results from building the same production-grade QA intelligence system across all three frameworks, then connecting them through Google’s Agent-to-Agent (A2A) protocol. The system accepts a software feature specification and autonomously produces a risk register, a prioritized test strategy scored by a proprietary VRTQ framework, and executable Playwright test scaffolds. Every implementation is benchmarked across six dimensions: developer experience, architectural expressiveness, tool integration, observability, token efficiency, and failure behavior.

The talk then demonstrates cross-framework interoperability via A2A, wiring all three into a single pipeline. Attendees leave with a practical decision framework, benchmark data, and a working blueprint for building framework-agnostic agentic systems.