Search for a command to run...
No comments yet. Be the first to comment.
Privacy isn't dead. It's just being rebuilt. The old internet promised us "move fast and break things." We moved fast. We broke things. Now we're realizing what we broke: privacy. Every LLM you use wa
Single AI agents are limited. They make mistakes. They miss context. They take wrong paths. Multi-agent systems don't have these problems. They have different problems. But better ones. The Single Age
You've learned basic RAG. You embed documents. You retrieve them. You feed them to an LLM. It generates an answer. Then you deploy it. And it breaks. Because the real world isn't simple retrieval. Adv
The biggest AI models aren't winning anymore. GPT-4: 1.8 trillion parameters Claude 3 Opus: 200+ billion parameters Claude 3.5 Sonnet: Unknown but smaller Llama 3.1: 405 billion parameters PhiMoe (ope
Muhammad Zulqarnain | Full Stack AI Engineer & Geospatial Developer
16 posts
A blog by Muhammad Zulqarnain — Full Stack AI Engineer & Geospatial Developer based in Turku, Finland. I write about RAG systems, LLMs, Prompt Engineering, Next.js, TypeScript, and geospatial development. Practical insights, deep dives, and real-world AI solutions.
Coding is changing faster than you think.
Not just "AI helps with autocomplete."
AI is becoming the primary developer.
You become the reviewer.
Old model (2020): Human writes code. AI doesn't help.
New model (2026): AI writes code. Human reviews.
Different skill. Different career.
Simple Tasks (100% reliable)
Boilerplate generation
Error fixing
Test writing
Documentation
Refactoring
Medium Tasks (80% reliable)
Feature implementation
Algorithm selection
Performance optimization
Security hardening
Hard Tasks (40% reliable)
Novel architecture
Complex systems design
Ambiguous requirements
Unknown unknowns
Step 1: AI generates code (Cursor, Copilot, Claude) Step 2: Human reviews (What did it do? Is it right?) Step 3: Human tests (Does it work?) Step 4: Human debugs if needed (Why didn't it work?) Step 5: Human ships
Result: 3-5x faster development
Cursor
AI-native code editor
Understand entire codebase
Suggest entire features
Not just line completion
GitHub Copilot
Native integration
Works in any IDE
Fast suggestions
Good for small things
Claude Code Interpreter
Can run and test code
Fixes its own errors
Works with Artifacts
Old skills:
Fast typing
Memory (library functions)
Syntax mastery
New skills:
Understanding requirements
Code review
Testing strategy
Architecture
Communication
Your value: Understanding the problem, not writing solutions.
You spend less time:
Writing boilerplate
Debugging syntax errors
Searching Stack Overflow
Remembering APIs
You spend more time:
Understanding requirements
Designing solutions
Reviewing AI suggestions
Testing comprehensively
Talking to stakeholders
AI can't:
Understand your full system
Know why decisions were made
Design novel solutions
Handle ambiguous requirements
Optimize for your specific constraints
It can:
Generate working code fast
Follow patterns
Generate tests
Suggest common solutions
Developer salary: \(150K/year Developer productivity: 1000 lines/week Cost per line: \)3
With AI: Productivity: 4000 lines/week Cost per line: $0.75
Business wins. Developer might worry.
You can either:
Resist: Pretend AI isn't happening Result: Replaced by people who use AI
Adopt: Become expert at AI-assisted development Result: 3-5x more productive Result: Can tackle harder problems Result: Higher pay, not lower
Project starts "Build a payment system"
AI writes:
Database schema
API endpoints
Test cases
Documentation
You:
Review architecture
Fix edge cases
Add security
Optimize database
Handle integrations
Time: 1 week instead of 4
Prompt engineering for code How to ask AI for what you want
Code review at scale Review AI output faster
Testing strategy Test AI-generated code thoroughly
Architecture Design systems humans + AI will build
Communication Explain requirements clearly
Junior developers (writing boilerplate)
Bug fixers (AI does this)
Code copy-pasters (AI does this)
API documenters (AI does this)
These aren't the future.
Senior engineers (review AI, design systems)
Architects (design for AI teams)
Security engineers (harden AI-generated code)
Performance engineers (optimize AI output)
These pay more.
1. AI-First Codebases Entire projects written by AI Humans handle edge cases
2. Verification Tools Tools that verify AI-generated code Prove correctness
3. Specialized AI Models Models trained on your codebase Understand your patterns
4. Code Search Find code by meaning, not syntax
Coding is becoming pair programming with AI.
You're no longer the primary code writer.
You're the critic, designer, architect.
This is better.
Not because AI is better.
Because you can focus on harder problems.
Embrrace it.kerpower
