AIhero

    Day 6: Human in the Loop Patterns

    Apply taste to your agentic workflow with HITL vs AFK task planning, Kanban-style issues, and cached research phases

    Matt Pocock
    Matt Pocock
    Day 6: Human in the Loop Patterns

    To build great products with AI, you need to know when to step in and when to step back. This workshop teaches you human-in-the-loop patterns that let you apply your taste — your judgment, your design sense, your architectural knowledge — while letting AI handle the grunt work.

    What You Will Learn

    • HITL and AFK Tasks — Identify which tasks need human judgment and which can be fully delegated to AI
    • Kanban Over Plans — Replace rigid multi-phase plans with a flexible Kanban board of independently grabbable issues
    • Using the Kanban Skill — Break a PRD into vertical slice GitHub issues with AFK and human-in-the-loop designations, then run an autonomous loop
    • Research — Cache expensive explore phases into local research documents to speed up every subsequent AI loop
    • Trying Out Research — Collaborate with AI to evaluate approaches, compare services, and produce a focused research asset
    • Prototyping — Build throwaway prototypes to validate ideas, test services, and apply your design taste before committing to implementation
    • Trying Out Prototyping — Turn research into a working proof of concept on a dev-only route using the Do Work skill
    • Designing Codebases AI Loves — Structure your code into deep modules with simple interfaces so AI can navigate, test, and modify it effectively
    • The Improve My Codebase Skill — Walk through your codebase to find shallow modules and refactor them into deep, testable units
    • Adding Module Awareness to PRD Planning — Embed deep module thinking into your PRD skill so new features are architected well from the start

    Day 6: Human in the Loop Patterns

    Matt Pocock
    Matt Pocock