Insights from our first AI Club 🪄

2026-05-27 · Source: Refactoring · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, quick

Summary

The first Refactoring AI Club session, attended by over 30 people, featured five members sharing practical applications of AI at work. Key insights included Matej Vitasek's "combine harvester" pattern, where AI agents maintain task state, instructions, and progress in a markdown file for multi-hour tasks. This low-tech approach allows agents to pause, reflect, and resume without losing context, making long tasks manageable. Additionally, Luca shared his "overnight procedures" or "guards," which are daily AI-powered checks designed to catch judgment-based issues that real-time rules might miss. These include identifying missed architecture decision records, refactoring opportunities, or performance drifts. The AI Club is a monthly event, facilitated by professional coaches, and is exclusive to paid Refactoring members, with the next session scheduled for June 5th.

Key takeaway

For AI Engineers managing complex, long-running agent tasks, consider implementing the "combine harvester" pattern. By having your agents maintain their state, instructions, and progress in a simple markdown file, you can enable task pausing, reflection, and seamless resumption, preventing context loss. Additionally, MLOps Engineers should integrate daily AI-powered "guard" procedures to retrospectively identify judgment-based issues like missed architecture decisions or performance drifts that static rules often overlook, enhancing overall system quality.

Key insights

Practical AI applications benefit from state persistence for long tasks and retrospective "guard" procedures for judgment-based issues.

Principles

• AI agents benefit from external state persistence.
• Retrospective AI "guards" catch nuanced issues.
• Simple, portable patterns enhance AI task management.

Method

The "combine harvester" method involves AI agents updating a markdown file with state, instructions, and progress, allowing tasks to pause, reflect, and resume. "Guards" are daily AI procedures that retrospectively identify judgment-based issues.

In practice

• Store AI agent state in markdown for long tasks.
• Run daily AI "guards" for judgment-based code issues.
• Automate detection of missed refactoring opportunities.

Topics

• AI Agents
• State Management
• Long-running Tasks
• Automated Code Review
• Technical Debt Detection
• Performance Monitoring

Best for: AI Engineer, Software Engineer, MLOps Engineer

Related on AIssential

• Build Long-running AI agents that pause, resume, and never lose context with ADK — Long-running AI agents require durable state machines, event-driven resumption, and multi-agent delegation to manage complex, multi-day workflows.
• AI REWIND 2025 - MLOps Reading Group Year-end Special — AI's rapid evolution demands robust production systems, precise context management, standardized protocols, continuous evaluation, and advanced post-training techniques.
• Debugging Multi Agent Memory Loss in Long Running Pipelines — Agentic amnesia in long-running AI pipelines is a systems architecture flaw, not an LLM defect, requiring structured memory management.
• New Relic and Agentic DevOps with Nic Benders — Observability is evolving from passive monitoring to AI-driven active intelligence, automating problem detection and resolution.
• Designing a Multi-Agent System for Engineering Support at Scale: A Case Study From Grab — Grab's multi-agent AI system automates engineering support, freeing up engineers for higher-value platform development by streamlining diagnostic and enhancement workflows.
• Build long-running MCP servers on Amazon Bedrock AgentCore with Strands Agents integration — Persistent state management enables AI agents to execute long-running tasks reliably across disconnected sessions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Refactoring.