We should be more tired than the model

2026-05-28 · Source: Tech Blog on ✰Vicki Boykis✰ · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Intermediate, quick

Summary

The author describes a personal struggle with agentic code generation, experiencing a loss of control and diminished skill retention despite producing code. This phenomenon is contrasted with the human brain's active engagement of short-term, working, and long-term memory during manual coding, which fosters understanding. The article posits that the "slot machine" user experience of AI code generation, offering instant solutions, undermines skill development. To counteract this, the author has implemented several deliberate strategies to reintroduce "friction" into the development process. These include writing initial code implementations for agent review, using agents to clarify code sections and retrieve documentation, evaluating multiple agent-suggested approaches, discussing agent outputs with colleagues, delaying agent use for 20 minutes, and actively studying fundamental data structures and academic papers. This approach aims to build a stronger personal foundation for coding rather than passively relying on AI models, ultimately enhancing long-term proficiency.

Key takeaway

For Machine Learning Engineers or Software Engineers adopting agentic code generation, recognize that over-reliance can diminish your coding skills and understanding. To maintain proficiency, integrate deliberate "friction" into your workflow by initially writing code yourself, using agents for review or clarification, and discussing their outputs with peers. This approach, though seemingly slower initially, will solidify your foundational knowledge and improve your long-term effectiveness with AI tools.

Key insights

Agentic code generation can hinder skill retention; deliberate friction improves long-term learning and tool mastery.

Principles

• Skill retention requires active cognitive engagement.
• Instant solutions can impede deep learning.
• Adding friction enhances understanding.

Method

The author's method involves integrating AI agents into a deliberate workflow: initial manual coding, agent-assisted review/documentation, multi-approach evaluation, peer discussion, and foundational study.

In practice

• Write initial code, then ask agent to review.
• Use agents for documentation retrieval.
• Discuss agent outputs with a peer.

Topics

• Agentic Code Generation
• Skill Retention
• Developer Productivity
• Human-AI Collaboration
• LLM Workflow
• Cognitive Load

Best for: NLP Engineer, Software Engineer, Machine Learning Engineer, AI 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.
• The Growing Dependence on AI Tools — And Why It’s Risky — Over-reliance on AI for coding accelerates development but diminishes fundamental understanding and critical thinking.
• How I deleted 95% of my agent skills and got better results — Nick Nisi, WorkOS — Agent performance improves by enforcing actions with code, guiding with specific "gotchas," and rigorously measuring outcomes, not by comprehensive instruction.
• 10 Lessons for Agentic Coding — Agentic coding is transforming software development, requiring new approaches and skills.
• Breaking the Spell of Vibe Coding — Vibe coding can induce a "dark flow" state, similar to gambling, leading to misperceptions of productivity and skill development.
• Faster Code, Slower Understanding: A Survival Note from a Year of Agent-First Engineering — AI shifts the engineering bottleneck from code production to comprehension and maintenance, demanding new optimization strategies.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Tech Blog on ✰Vicki Boykis✰.