Recent Blog Posts
Insights on AI Agents, Modern Web, and the Future of Engineering.
Anatomy of an Agentic Loop: The ORAE Cycle
Explore the fundamental Observe-Reason-Act-Evaluate cycle that powers modern autonomous AI agents, state persistence, and termination strategies.
Automated CI/CD Regression Evaluation for ADK Agents
Implement an automated Python quality gate for ADK agents in CI/CD pipelines to prevent trajectory regressions and enforce pass-rate thresholds before deployment.
Build a Custom Autonomous Coding Agent from Scratch
Build a production-ready autonomous coding agent loop in Python implementing the ORAE cycle, tool registries, safety gates, and budget guards.
Building a Dispatcher Agent with Specialized Subagents
Implement dynamic triage and intent-based routing using Google ADK dispatcher agents and specialized subagents for billing, tech support, and beyond.
Building an Eval Test Suite for ADK Agents
Move from vibes-based testing to structured evaluation for ADK agents, tracking tool call accuracy, trajectory efficiency, and LLM-as-a-judge quality scoring.
Building Custom Async Function Tools in Google ADK
Learn how to build non-blocking asynchronous function tools in Google ADK to handle database queries, REST APIs, and external microservices with high performance.
Building an Approval & Refund Graph Workflow with ADK 2.0
Step-by-step tutorial for constructing an auditable, graph-based refund processing workflow using ADK 2.0 and Python conditional routing.
One-Command CI/CD Deployment for ADK with GitHub Actions
Set up an end-to-end GitHub Actions workflow that executes tests, runs ADK evaluation quality gates, builds multi-stage Docker images, and deploys to Cloud Run.
Connecting MCP Tool Servers to ADK Agents
Connect Google ADK agents to Model Context Protocol (MCP) tool servers to seamlessly access pre-built integrations like SQLite, GitHub, Slack, and Google Drive without writing custom wrappers.
Containerizing & Deploying ADK Agents to Cloud Run
Learn how to package your Google ADK agent harnesses into lightweight Docker containers using uv and deploy them as auto-scaling microservices on Google Cloud Run.
Context Window Engineering for Long-Running Loops
Master context drift, context rot, memory compaction, Git worktree isolation, and token budgeting for reliable long-running AI agent loops.
Cross-Language Agent Communication via A2A Protocol
Connect Python ADK agents with remote microservices and workers written in Go or Java using the Agent-to-Agent (A2A) HTTP standard protocol.