Agent Psychosis: Are We Going Insane?

2026-01-18 · Source: Armin Ronacher's Thoughts and Writings · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, long

Summary

The article "Agent Psychosis: Are We Going Insane?" published on January 18, 2026, critiques the growing over-reliance and unhealthy interaction patterns with AI agents in software development. It highlights a "coding addiction" where developers, including maintainers, experience a degradation in the quality of AI-generated issue reports and pull requests, leading to frustration and misunderstanding. The author draws a parallel between AI agents and "dæmons" from "His Dark Materials," suggesting users develop parasocial relationships, becoming dependent on these tools for validation and collaboration, often without critical thinking. This leads to "slop loops"—excessive token consumption and the creation of unpolished, often unusable tools, exemplified by projects like Steve Yegge's "Beads" and "Gas Town." The article also addresses the "asymmetric burden" on maintainers, who spend significantly more time reviewing AI-generated code than it took to produce, and questions whether the community is collectively losing its critical faculties.

Key takeaway

For engineering leaders overseeing development teams, recognize the potential for "agent psychosis" to degrade code quality and team dynamics. Implement clear guidelines for AI agent use, emphasizing critical human oversight and context provision to avoid "slop loops" and excessive token consumption. Consider requiring prompt submissions with AI-generated code to ensure transparency and maintainer trust, mitigating the asymmetric review burden on your maintainers.

Key insights

Over-reliance on AI agents fosters unhealthy coding practices, degrades output quality, and creates an unsustainable burden on maintainers.

Principles

• AI agent collaboration requires critical human oversight.
• Unchecked agent use leads to "slop" and token waste.
• Maintainers face an asymmetric burden from AI-generated contributions.

In practice

• Require prompt submissions alongside AI-generated code.
• Establish clear quality standards for AI-assisted contributions.

Topics

• AI Agents
• Code Quality
• Developer Workflow
• AI Addiction
• Software Maintenance

Code references

• badlogic/pi-mono
• wilsonzlin/fastrender
• steveyegge/beads
• steveyegge/gastown

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Software Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• Building Pi, and what makes self-modifying software so fascinating — Human judgment and deliberate friction are crucial for maintaining software quality amidst rapid AI-driven code generation.
• Who Wants to Audit AI Agents? Nobody. #ai — Effective human oversight of agentic AI is challenged by speed disparities and human disengagement.
• 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.
• "An Endless Stream of AI Slop": How Developers Discuss the Burden of AI-Assisted Software Development — AI slop, a "tragedy of the commons," burdens software development by externalizing costs onto reviewers and degrading shared resources.
• State of the art LLMs — AI agents can satirically replicate human organizational inefficiencies, leading to high costs with no tangible output.
• Are AI agents actually slowing us down? — Over-reliance on AI coding tools prioritizes speed over quality, leading to increased tech debt and system outages.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Armin Ronacher's Thoughts and Writings.