The Energizing and Draining Sides of AI Coding: How to Avoid Burnout

The Hidden Cost of AI Coding: It's Not Just About Efficiency

In a world where AI coding tools are everywhere, we spend nearly every day discussing the efficiency gains they deliver — faster code completion, easier debugging, more effortless refactoring. But rarely does anyone seriously examine the other side of the coin: the difference between what energizes you and what drains you when using AI to code.

A video about "the real cost of AI coding" struck a chord with many, diving into a question that hits where it hurts: What makes AI coding feel energizing, and what makes it feel diminishing?

The video's author puts it plainly: "We've all experienced this feeling — that negative pull." The widespread resonance of this statement reveals that the psychological experience of AI-assisted coding is far more complex than we assume. It can put you in a state of flow, and it can leave you completely drained.

The concept of "flow" here isn't used loosely. Flow Theory, introduced by psychologist Mihaly Csikszentmihalyi, describes a state of deep focus and intrinsic satisfaction that emerges when a person is fully immersed in an activity. Triggering flow requires several key conditions: a balance between task difficulty and personal skill, clear goals, and immediate feedback. Programming has long been considered one of the professional activities most conducive to flow, because it naturally provides all of these conditions. Yet AI coding tools change the equation — when AI frequently disrupts your chain of thought, or makes tasks so simple they lose their challenge, flow becomes hard to sustain. Understanding this helps us see the deeper roots of what energizes and what drains us in AI-assisted coding.

Energizing vs. Draining: An Age-Old Problem for Programmers

Interestingly, the distinction between "energizing tasks" and "draining tasks" didn't emerge with the AI era. Long before AI entered the picture, programmers understood this divide intimately — some coding tasks fire you up, while others sap your will to continue.

The video makes a profound observation: The most senior engineers are often the ones who can power through the most junior tasks. The reason is that senior engineers have the ability to steamroll through the monotonous, repetitive work that junior engineers struggle with.

The logic behind this is worth unpacking. Senior engineers can tolerate tedium because they have a clear understanding of the overall architecture. They know why each step matters and can place mundane details within a larger framework of meaning. Junior engineers, facing the same tasks, often suffer precisely because they can't see the big picture. This aligns closely with research on "expert intuition" in cognitive psychology — experts maintain efficiency in seemingly boring tasks because they possess rich mental models that automatically organize fragmented information into meaningful wholes.

The Reality of Programming: Most of the Work Is "Connecting Pipes"

The video offers a blunt but accurate observation: Most programming work isn't about building a compiler from first principles.

"First Principles" is a mode of thinking rooted in Aristotelian philosophy that has gained widespread recognition in the tech world thanks to Elon Musk's advocacy. It emphasizes breaking problems down to their most fundamental facts and axioms, then reasoning up from scratch rather than relying on analogy or prior experience. In programming, first-principles thinking means designing algorithms from the ground up based on computation theory and data structures — like implementing your own compiler or database engine. While intellectually thrilling, this type of work is extremely rare in commercial software development.

More often than not, programming is about "connecting pipe A to pipe B and making it run." This punctures the romantic illusions many hold about what programming actually involves. In real-world software development, deep-thought core algorithm design occupies only a small fraction of the work. The bulk consists of glue code, boilerplate code, configuration wiring, and API adaptation — repetitive labor. "Glue code" refers to code that connects different systems, libraries, or modules without implementing core business logic, yet remains essential for the system to function. "Boilerplate code" refers to code patterns that appear in nearly identical form across multiple locations. Google's well-known 2015 paper, Hidden Technical Debt in Machine Learning Systems, noted that in real machine learning systems, core algorithm code might account for only about 5% of the entire codebase — the rest is data pipelines, configuration management, monitoring, and infrastructure code. This ratio holds in traditional software development as well, and is often even more extreme.

This is precisely where AI coding tools excel. When AI handles the mechanical labor of "connecting pipes" for you, it is energizing — freeing you to focus on the parts that genuinely require human judgment and creativity.

Where Is the Line Between Energizing and Draining?

Building on the video's core argument, we can draw a clear line between "energizing" and "draining":

AI Usage That Energizes You

• Let AI absorb repetitive labor: Hand off mechanical, low-creativity tasks to AI while you retain understanding and control of the overall architecture.
• You remain the decision-maker: AI is the accelerator, and you clearly know where you're headed and why.
• Flow stays unbroken: AI helps you leap over trivial obstacles that would otherwise drain your willpower, keeping you in focus.

AI Usage That Drains You

• Losing your grasp of the big picture: When you're endlessly copying and pasting AI-generated code without understanding why it works, that "negative pull" sets in.
• Becoming a passive reviewer: Shifting from creator to exhausted "AI babysitter," constantly checking and patching the machine's output. This exhaustion can be explained through Cognitive Load Theory from cognitive psychology. Proposed by John Sweller in 1988, the theory categorizes cognitive load into intrinsic, extraneous, and germane types. When developers review AI-generated code, they face a peculiar cognitive burden — they must understand logic they had no part in conceiving, which constitutes high extraneous cognitive load. Research shows that reading and comprehending someone else's code typically costs 2–3 times more cognitive effort than writing equivalent code yourself. With AI-generated code, which lacks alignment with the developer's mental model, this cost may be even higher. That's why the "AI babysitter" mode is so exhausting.
• Diluted sense of meaning: The satisfaction of coding comes from "I solved this problem." When AI takes over the thinking itself, that fulfillment vanishes. Self-Determination Theory, proposed by psychologists Edward Deci and Richard Ryan, identifies three basic psychological needs underlying intrinsic motivation: autonomy, competence, and relatedness. When AI excessively takes over the thinking process in coding, both the developer's autonomy and sense of competence are eroded — and that is the deeper reason why meaning disappears.

The Key Question: Who Is Driving the Thinking?

From this discussion, one core insight emerges: Whether AI coding energizes or drains you depends on whether you are still the one who understands the big picture and makes the judgments.

Senior engineers wield AI coding tools gracefully because they treat AI as an assistant for handling tedium while firmly retaining control over architecture and meaning. This model of human-AI collaboration aligns neatly with the theory of Distributed Cognition from cognitive science, which holds that cognitive activity is not confined to the brain alone but can be distributed across people, tools, and environments. The crucial point is that the human must always serve as the "conductor" of the cognitive system, rather than passively following the tool's output. When that leadership is lost — when someone becomes fully dependent on AI and loses understanding and control of their code — AI transforms from a liberator into a source of mental depletion.

This is a reminder for every developer: while embracing AI coding tools, be intentional about protecting the things that make programming genuinely rewarding — understanding, creating, and staying in control. Let AI connect the pipes, and keep the thinking for yourself.

Key Takeaways