I'm done coding with AI

Overview

Brett, a software developer with 20+ years of experience, explains why he chose to stop using AI coding tools and AI chatbots (including products like GitHub Copilot, Cursor, and Claude Code). He frames his decision as both personal (how these tools affected his thinking, skills, and well-being) and broader (their alleged negative impacts on software quality, society, and the environment).

His History With AI Coding Tools

Early Experience (Auto-Complete)

• He tried Copilot-like auto-complete in an editor.
• He found it distracting, disliked seeing generated code, and felt it interfered with his own thinking.

A “Hallucination” Incident

• In 2022–2023, he and a colleague encountered an AI chatbot that claimed a software dependency version existed and suggested upgrading to that version would fix the issue.
• When the version didn’t exist, it reinforced his view that these tools can confidently fabricate incorrect information.

Workplace Pressure

• After switching jobs, he observed teammates using AI editors quickly.
• Later, management mandated AI usage after attending a conference where leaders were sold the idea that AI would revolutionize knowledge work and make non-users obsolete.

What Changed: His “Trial” and Later Withdrawal

Spring 2025 Onwards

Despite earlier concerns, he experimented more seriously:

• Cursor was too slow or hard to run in Linux for him, so he used an AI-enabled editor (Zed).
• Initially, he felt impressed—AI seemed to accelerate development and help him finish projects faster.

Health + Chatbot Advice Incident

• After he became sick at night, he asked an AI chatbot for medical guidance and was told to go to an emergency department.
• Later, a clinician told him he should have rested at home, leaving him embarrassed.
• The experience made him more concerned about how conversational AI can influence behavior.

Growing Doubts

Around the same time, he began to worry about:

• the psychological effects of interacting with chatbots, and
• how easily these systems can shape decisions—especially for people who are not technical.

Claimed Downsides of “Agentic” Coding

Brett argues that as AI tools get better at writing and modifying code with less human involvement, the trade-offs become harmful:

• Developer existential dread and purposelessness

  • He describes a shift from meaningful craft to “feeding” an AI.
  • He experiences anxiety about what his role is if the machine writes most of the code.

• Work becomes QA/review instead of building

  • He reports hearing that developers should:
    • train systems by writing markdown/checks,
    • review large volumes of code, or
    • serve as testers of AI output.
  • He says this isn’t the kind of work he finds satisfying.
  • He also notes that reviewing massive code volumes makes it harder to understand what’s changing.

• Distance from the code reduces care and learning

  • He claims developers become less connected to the codebase,
  • develop apathy toward AI-generated work, and
  • learn less because they aren’t struggling through difficult problems the way they did when coding by hand.

• Impacts on product quality and end users

  • He connects this to “feature bloat” and the creation of slower, harder-to-use software.
  • He cites his view that large cloud platforms (e.g., AWS/Google Cloud) often feel bloated and difficult to use.

• Business risk if teams “de-skill”

  • He warns that if AI tools go down, teams may be unprepared.
  • He believes teams may lose the practice needed for difficult programming work.

Broader Costs: Environment and Community

Brett also claims he learned about AI’s real-world resource costs, including:

• energy use and water usage per prompt,
• environmental impacts from training and operating models,
• data centers affecting local waterways and communities, and
• supply-chain strain (e.g., hardware availability/costs).

He argues these harms are easy to ignore because they’re abstracted away.

His Conclusion and What He’s Doing Instead

• He says he stopped using AI tools completely.
• He returned to hand-coding for fun—making games and even a small engine.
• He reports that coding by hand felt more satisfying, restored his sense of purpose, and reignited his growth as a developer.
• He acknowledges the decision may be risky for his career (including the possibility of being fired or being unable to find work), but he believes there must be a way to exist in software without AI usage.
• He intends to keep sharing programming education and experiences.

In short: Brett presents his decision as a move toward more meaningful craft, better personal agency, and fewer downstream harms—both technical and environmental.

Presenters/Contributors

• Brett (speaker) — Software developer; creator of the video.