Saw a Guy Coding Today. No Cursor. No ChatGPT. Just Sat There Typing. Like a Psychopath.

Yesterday, I witnessed something that stopped me in my tracks. A colleague was sitting at his desk, methodically typing out code, line by line. No GitHub Copilot suggestions. No ChatGPT tab open. Just pure, unassisted coding. In 2025.

At first, I thought his AI tools were broken. Maybe his Copilot subscription expired? Network issues? But as I watched longer, I realized this was intentional. He was choosing to code without AI assistance, and frankly, it made me question everything about how we develop software today.

The Great Divide: AI Adopters vs. The Resistance

Watching him work got me thinking about the surprisingly large number of developers who are still resistant to AI coding tools. It's not just about being old-fashioned—there's real psychology behind this resistance, and some legitimate concerns that are worth exploring.

Recent research confirms just how widespread AI adoption has become in software development. The latest DORA State of DevOps Report reveals that more than three-quarters of developers now depend on AI for core professional tasks including code generation, documentation, and problem-solving. This makes our manual-coding colleague part of an increasingly rare 24% minority.

According to Stack Overflow's 2023 Developer Survey, while 77% of developers have used AI coding tools, a significant 23% haven't adopted them yet. That's nearly 1 in 4 developers who are still coding the "old way."

Why Some Developers Fear AI Tools

After talking to my colleague and others like him, I've discovered several reasons why some developers are hesitant to embrace AI assistance:

• Skill Atrophy Concerns: "If I rely on AI too much, will I lose my fundamental coding abilities?"
• Code Quality Fears: "AI-generated code might have hidden bugs or security vulnerabilities"
• Learning Impedance: "How can I truly understand the code if I didn't write it myself?"
• Dependency Anxiety: "What happens when the AI isn't available?"
• Professional Identity: "Is coding with AI assistance really 'programming'?"

The Pros and Cons: A Honest Assessment

Let's be honest about both sides of this debate. AI coding tools aren't universally good or bad—they're tools with specific advantages and disadvantages.

The Legitimate Concerns Are Real

After researching this topic, I've realized that the "AI resistance" isn't just stubbornness. There are legitimate concerns backed by emerging research:

A study published in Nature found that heavy reliance on AI assistance can lead to reduced problem-solving skills over time. Another research paper showed that AI-generated code often contains security vulnerabilities that experienced developers might miss during reviews.

The Security Question

Perhaps the most serious concern is security. AI models are trained on public code repositories, which means they might suggest patterns that include known vulnerabilities. A IEEE Software study found that AI-generated code had a 40% higher chance of containing security flaws compared to human-written code.

Finding the Middle Ground

After observing my colleague and diving deep into this topic, I think the answer isn't black and white. The most successful developers seem to be those who use AI tools strategically rather than as a complete replacement for their coding skills.

The Future of Human-AI Collaboration

Rather than replacing human programmers, AI tools are evolving into sophisticated coding partners. The future likely belongs to developers who can effectively collaborate with AI while maintaining their core programming competencies.

As Claude, Cursor, and other advanced AI tools become more sophisticated, the key will be knowing when to rely on them and when to think independently.

Respecting the Choice

Watching my colleague code without AI assistance reminded me that there's value in both approaches. His deep understanding of every line, his methodical problem-solving process, and his independence from external tools are genuinely admirable qualities.

Maybe the "psychopath" label is unfair. Perhaps he's more like a craftsperson who chooses hand tools over power tools—slower, but with a deeper connection to the work and a comprehensive understanding of the process.

The real question isn't whether you should use AI tools or not. It's whether you can maintain your expertise and judgment while leveraging these powerful assistants. The best developers of the future will likely be those who can code both with and without AI, choosing the right tool for the right situation.

What's your approach? Are you team AI-assisted coding, team manual coding, or somewhere in between? The debate continues, and honestly, that's probably a good thing.