Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
SPEAR: A Framework for Autonomous Agents
Learn SPEAR, a framework for autonomous agents that structures AI work into a loop of Scope, Plan, Execute, Assess, and Resolve, preventing common process failures and rework.
I built SPEAR, a management framework for autonomous agents that turns AI work from “prompt → output” into a structured loop: Scope, Plan, Execute, Assess, Resolve.
The core idea is that most AI failures are not model failures; they are process failures. Agents often start strong and finish weak because they skip the same steps humans skip when they are moving too fast: clarifying the scope, making the plan visible, checking the work against a rubric, and narrowing the gap before calling it done.
In the demo, I’ll show how SPEAR works as a reusable workflow for agentic software development and creative production. I’ll walk through the five phases, show how an agent scopes a task, produces a plan, executes the work, assesses the result against a MECE rubric, and loops back until the output passes. I’ll also show the logs, the assessment rubric, and examples of how the same loop can run for seconds, minutes, hours, or a full workstream. The goal is to show a working management system for autonomous agents, not just another prompting trick.