Six months ago, I was skeptical about AI coding tools. Today, I use them daily. This shift didn't happen because AI got dramatically better — it happened because I finally understood what these tools actually are and, more importantly, what they're not.
Let's talk honestly about AI coding assistants, the anxiety they're causing among developers and what this technology actually means for our careers.
The Elephant in the Room
Every developer I know in Nairobi has asked this question: "Will AI replace me?" It's a legitimate concern. When you see GitHub Copilot write entire functions from a comment, or watch ChatGPT debug code in seconds, the anxiety is real.
Here's what I've learned after months of using these tools: AI won't replace developers. But developers who use AI will replace developers who don't.
That's not a threat. It's an opportunity.
What AI Coding Tools Actually Do Well
Boilerplate Code Generation
This is where AI shines. Need a React component with state management? A CRUD API endpoint? Form validation? AI can generate the skeleton code almost instantly. I recently built a dashboard that required about 15 similar components. Instead of copying and pasting code for an hour, I described what I needed to Copilot and had working templates in minutes. Did I need to review and modify them? Absolutely. But the time saved was massive.
Code Explanation and Learning
This is underrated. When I encounter unfamiliar code — maybe I'm working with a new library or maintaining legacy code — I can paste it into ChatGPT and get an explanation in plain English. Last week, I had to work with a complex regex pattern. Rather than spending 30 minutes decoding it, I asked ChatGPT to explain it. Two minutes later, I understood it completely and could modify it for my needs.
Debugging Assistance
Here's a real scenario: I had a bug in a React component where state wasn't updating correctly. I described the issue to ChatGPT, pasted my code and it immediately identified that I was mutating state directly instead of using the setter function. Could I have found this myself? Yes, probably in 20-30 minutes. The AI found it in 30 seconds.
Documentation and Comments
Most developers hate writing documentation. AI is actually good at this. Point it at a function and it can generate clear documentation explaining what it does, what parameters it takes and what it returns.
What AI Coding Tools Do Poorly
Architecture Decisions
AI can't design your application architecture. It doesn't understand your business requirements, your scale considerations or your team's capabilities. These decisions require human judgment and experience.
Creative Problem Solving
Give AI a well-defined problem and it does great. Give it a novel problem that requires creative thinking and it struggles. I recently needed to optimize a data-heavy dashboard for slow connections — a real issue in Kenya. The solution required understanding the user's context, creatively implementing progressive loading and making UX trade-offs. AI couldn't do this.
Understanding Business Context
AI doesn't understand why you're building something. It doesn't know your users, your market or your constraints. When a client asks for a feature, AI can't tell you if it's a good idea or suggest a better approach based on their actual needs.
Code Quality and Standards
AI-generated code often works but isn't always good code. It might be inefficient, insecure or violate your team's coding standards. You still need to review, refactor and improve it.
The Real Impact: Productivity, Not Replacement
After using AI tools for six months, my productivity has genuinely increased — I estimate by about 30-40%. But it hasn't made me a better developer. It's made me a faster developer. There's a difference.
What I'm faster at: writing boilerplate code, debugging syntax errors, learning new libraries, converting designs to code and writing tests.
What I still need to be good at: understanding requirements, designing solutions, making architectural decisions, code review, performance optimization, security considerations and team collaboration.
How I Actually Use AI Tools
GitHub Copilot: enabled in VS Code for auto-completing repetitive patterns, generating test cases and writing common utility functions. I don't blindly accept its suggestions. I read every line it proposes.
ChatGPT: my debugging partner and learning assistant for explaining unfamiliar code, debugging specific issues and learning new concepts.
My workflow:
Understand the problem (human work)
Design the solution (human work)
Write critical logic myself (human work)
Let AI help with boilerplate (AI + human review)
Review and refactor (human work)
Test thoroughly (human work)
The most important steps are still 100% human.
The Skills That Matter More Than Ever
With AI handling more routine coding, certain skills have become even more valuable: code review, system design, problem decomposition, communication and domain knowledge. These are the things AI genuinely can't replicate.
For Junior Developers: A Warning and Opportunity
The threat: if your value is writing basic CRUD operations and following tutorials, AI can do that.
The opportunity: AI can accelerate your learning dramatically. You can ask questions without feeling stupid. You can get explanations instantly. You can experiment faster.
But you need to actually learn, not just copy and paste.
Bad approach: "I need a login function" → paste AI code → deploy → move on.
Good approach: "I need a login function" → ask AI for explanation → understand the approach → write it yourself → use AI to check your work.
The Accessibility Angle
AI is democratizing coding knowledge. In Kenya, where access to quality programming education can be limited, AI tools provide a form of mentorship that wasn't available before. A student in Kisumu can get the same quality of code explanation as someone in Silicon Valley. That's powerful.
My Honest Take
Am I worried about AI? No. Am I adapting to it? Absolutely.
AI coding assistants are like calculators for math. Calculators didn't make mathematicians obsolete — they made them more productive and allowed them to focus on more complex problems.
The developers who resist AI will be at a disadvantage. But the developers who blindly rely on AI without understanding what it produces will be equally disadvantaged.
The sweet spot is using AI as a powerful assistant while maintaining your skills and judgment.