Everything We Got Wrong About Research-Plan-Implement - Dexter Horthy

2026-03-19 · Source: MLOps.community · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Advanced, quick

Summary

Dexter Horthy, a keynote speaker at the Coding Agents Conference on March 3rd, 2026, argues that while "RPI" (presumably a prior AI coding paradigm) aimed to improve AI coding, it inadvertently hindered it by encouraging engineers to outsource critical thinking to agents. Horthy, currently building AI coding agents at HumanLayer, advocates for a new approach he terms "CRISPR." This methodology emphasizes reducing reliance on "magic prompts," increasing structural rigor, and fostering greater human ownership in the coding process. He posits that the fundamental issue lies not with the AI itself, but with engineers' tendency to avoid deep engagement with problem-solving.

Key takeaway

For AI Architects designing agent-assisted coding workflows, you should critically evaluate the extent to which agents are replacing, rather than augmenting, human thought. Implement the "CRISPR" principles by designing systems that demand structured input and clear human checkpoints, ensuring engineers retain ownership and deep engagement with complex problems to prevent the degradation of code quality.

Key insights

Outsourcing thinking to AI agents degrades coding quality; human ownership and structured approaches are crucial.

Principles

• Reduce reliance on "magic prompts."
• Increase structural rigor in AI coding.
• Foster human ownership in the coding process.

Method

The "CRISPR" approach advocates for fewer magic prompts, more structure, and increased human ownership to counter the negative effects of over-reliance on AI agents in coding.

In practice

• Implement structured coding workflows.
• Prioritize human oversight in agent outputs.

Topics

• AI Coding Agents
• Context Engineering
• Agent Orchestration
• Human-in-the-Loop Systems
• Robotic Process Intelligence

Best for: AI Architect, AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• AI Coding Agents, Deconstructed — Effective AI agent systems require structured architectures that separate concerns to manage context and scale workflows.
• Cognition’s Scott Wu says AI coding agents shouldn’t replace humans — AI coding agents like Devin aim to augment human programmers by automating "toil," not replace them.
• Conductors to Orchestrators: The Future of Agentic Coding — Software engineering is shifting from direct coding to orchestrating autonomous AI agent fleets for scaled productivity.
• Applying an Agentic Coding Tool for Improving Published Algorithm Implementations — Agentic coding tools can significantly improve published algorithm implementations across diverse domains within a single workday.
• 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.
• AI Coding Agents Get a Stack Overflow of Their Own — Stack Overflow for Agents provides an API-first knowledge exchange to prevent AI coding agents from repeatedly rediscovering solutions.

Open in AIssential →

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