AI Impact on Coding Skills and Developers

In recent years, a growing number of studies have explored the impact of artificial intelligence (AI) on humans after long-term use, especially in software development.
The big question is simple:
Do tools like ChatGPT, Claude, and GitHub Copilot actually help developers learn… or just help them finish tasks faster without real understanding? 🤔
Let’s break down what the research says 👇

Anthropic Study: Does AI Build Skill or Just Speed Up Work? 🧠​

Anthropic, the company behind Claude, published a study titled: “How AI assistance impacts the formation of coding skills”
The research focused on an important question:
Does AI help developers truly learn, or just complete tasks faster?

Experiment Setup 📊​

• 52 developers participated
• Most were Junior or Mid-Level engineers
• All had Python experience
• None knew the Trio library before the experiment
• They were split into two groups:
  • Group A: No AI ❌
  • Group B: AI allowed 🤖

Their task was to learn the new library and use it in real work, including:

• Understanding the code
• Debugging issues
• Analyzing functionality

Results ⚠️​

The AI-assisted group:

• Understood only 17% of the library deeply
• Performed worse in debugging tasks
• Relied heavily on AI instead of thinking independently

The study concluded:

“AI-enhanced productivity is not a shortcut to competence.”

💡 Meaning:
Being faster with AI does not automatically mean you are more skilled.

---

Jellyfish Study: Security Risks in AI-Generated Code ⚠️​

Another study titled: “The Risks of Using AI in Software Development”
found alarming results:

• 40% to 51% of AI-generated code contains security vulnerabilities

Such as:

• SQL Injection
• Cross-Site Scripting (XSS)
• Unsafe dependencies

This raises serious concerns about using AI-generated code without proper review, especially in cybersecurity and production systems 🔐

---

“Code of Silence” Study: Hidden Risks of AI Code 🤫​

A report titled: “Code of Silence - The Hidden Risks of AI-Generated Code”
highlighted that companies are increasingly adopting AI because it:

• Speeds up development
• Reduces costs

But it also introduces new problems:

• Lower code quality
• Reduced application stability
• Lower trust in AI-generated code

Even developers themselves often hesitate to fully trust AI-written code.

---

DORA Report: Insights From 39,000 Developers 🌍​

The DORA (DevOps Research and Assessment) report is one of the largest software engineering studies, involving over 39,000 developers worldwide.

What they found about AI 🤖​

• AI helps with writing and explaining code
• But it also introduces recurring issues:

⚠️ AI Hallucinations​

Sometimes AI generates code that is:

• Out of context
• Incorrect for the project
• Or completely unrelated to the problem

Performance Impact 📉​

The report also found:

• Around 1.5% decrease in software delivery speed
• Over 8% decrease in code stability

---

Academic Research: Lower Security in AI-Assisted Code 🎓​

Studies from:

• Stanford University
• Wuhan University

show that developers using AI tend to produce less secure code compared to those writing code manually.

This includes:

• Weak authentication logic
• Unsafe coding patterns
• Hidden vulnerabilities

---

Veracode Report: 90% of AI “Vibe Coding” Has Security Flaws 🚨​

The Veracode GenAI Code Security Report delivered even more alarming data:

• 90% of AI-driven “vibe coding” projects contain serious vulnerabilities
• Common issues include:
  • Hardcoded secrets 🔑
  • Weak authentication
  • SQL injection risks
  • Unsafe dependencies

Deskilling Problem ⚠️​

Researchers also noticed a growing issue called deskilling, meaning:

• Developers lose core understanding
• Depend too much on AI
• Become less capable of deep debugging

---

Does This Mean AI Is Bad? 🤔​

Not at all ❌
AI is still extremely powerful and useful.
The real issue is not AI itself… It’s over-reliance without understanding.

---

The Real Risk for the Future 🚨​

If AI becomes the primary way developers write code without learning:

• Developer skills may weaken
• Security vulnerabilities may increase
• Software quality may decline
• Deep understanding of systems may disappear

AI is powerful - but without human oversight, it can create serious risks.

---

Final Thoughts 🧠🔥​

AI in programming:

• Speeds up development ✅
• Helps with learning (sometimes) ✅
• But may reduce deep understanding if overused ⚠️

The real skill is not using AI…
It’s understanding what AI produces.