Quoting Jon Udell

2026-06-28 · Source: Simon Willison's Weblog · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Intermediate, quick

Summary

Jon Udell, in a June 28, 2026 post titled "Doctor, it hurts when agents create unreviewable PRs." "Don't do that.", advocates for reframing the common phrase "human in the loop" to "agent in the loop." He argues that "human in the loop" cedes authority to machines, implying humans are secondary. Instead, Udell proposes that processes should remain human-centric, with AI agents invited to join the team as assistants within established human workflows. This perspective emphasizes maintaining human control and agency, particularly when applying AI to tasks like agentic software development. The goal is to integrate agents into existing human-defined loops, ensuring humans are not excluded but rather are the orchestrators of the process.

Key takeaway

For AI Engineers designing agent-assisted systems, reframe your approach from "human in the loop" to "agent in the loop." This means you should architect systems where human processes are primary, and agents are explicitly invited as team members to assist, not dictate. Ensure your development workflows maintain human authority, preventing agents from creating unreviewable outputs and preserving your team's control over the entire process.

Key insights

Humans should maintain authority, inviting AI agents into human-centric processes rather than ceding control.

Principles

• Human processes are primary.
• Agents are invited assistants.
• Maintain human authority over AI.

In practice

• Integrate agents into existing workflows.
• Design agent-assisted software development.

Topics

• Agentic AI
• Human-Agent Collaboration
• Software Development Workflows
• AI System Design
• Human Control

Best for: AI Architect, Software Engineer, AI Engineer, Director of AI/ML

Related on AIssential

• From assistants to agents — Agentic AI autonomously executes goals, shifting human-AI collaboration from reactive assistance to proactive, independent action.
• Loop Engineering — Loop engineering shifts from direct agent prompting to designing autonomous systems that manage agent interactions and tasks.
• Everything We Got Wrong About Research-Plan-Implement - Dexter Horthy — Outsourcing thinking to AI agents degrades coding quality; human ownership and structured approaches are crucial.
• Getting Humans Out of the Way: How to Work with Teams of Agents — Effective coding agent management requires designing autonomous, parallelized, and self-validating systems, not micromanagement.
• Human oversight of agentic systems in practice: Examining the oversight work, challenges, and heuristics of developers using software agents — Effective human oversight of agentic systems requires proactive control and practical heuristics, not just reactive monitoring.
• What's good for human developers is good for AI agents (and vice versa) — Agents can autonomously generate code changes and pull requests by leveraging internal data and collaborative prompting.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.