For the past 8 months, I've been completely immersed in the world of AI LLMs. What started as curiosity about ChatGPT turned into a deep dive into every major AI service, fundamentally changing how I approach coding and problem-solving.
The Beginning: A Long-Awaited Dream Realized
For 4 years at KAIST, I had been dreaming of creating an edge computing GUI simulator. Despite numerous attempts, the complexity of building a real-time visualization system with multiple scheduling algorithms always seemed insurmountable. Traditional coding methods meant months of development for even basic functionality.
Everything changed when I discovered AI coding capabilities. With OpenAI's o1-preview model and Cursor, what seemed impossible became achievable. In just one month, I built a fully functional edge computing simulator with features I had only dreamed of:
- Real-time task allocation visualization
- Multiple scheduling algorithms (Round Robin, Least Loaded, Random)
- Dynamic edge node management
- Performance metrics and analytics
The YouTube demo (watch here) shows the result of AI-assisted development - something that would have taken 6+ months traditionally was completed in 30 days.
Beyond ChatGPT: Exploring the AI Ecosystem
My journey didn't stop with one tool. I systematically explored every major AI service, paying for premium subscriptions to truly understand their capabilities:
ChatGPT Plus & o1-preview
Started with GPT-4, but o1-preview was the game-changer. Its ability to reason through complex architectural decisions and generate production-ready code was unprecedented. Used it primarily for system design and algorithm implementation.
Claude (Anthropic)
Claude became my go-to for code review and refactoring. Its understanding of context and ability to maintain consistency across large codebases was superior. Particularly effective for writing documentation and explaining complex concepts.
Google Gemini
Gemini excelled at multimodal tasks. I used it for the AI character project, leveraging its API to create Harry Potter characters (Hermione, Harry, Ron) with distinct personalities. The conversation system maintained character consistency while enabling natural multi-character interactions.
Cursor IDE
The integration of AI directly into the development environment was revolutionary. Cursor's ability to understand entire project contexts and suggest changes across multiple files simultaneously accelerated development by 10x.
The AI Character Project: Bringing Fiction to Life
One of my most ambitious projects was creating an AI-powered character conversation system. Using Google Gemini's API, I developed a framework where fictional characters could:
- Maintain consistent personalities based on source material
- Remember conversation context and past interactions
- Interact naturally with multiple characters simultaneously
- Generate emotionally appropriate responses
The technical implementation involved:
- RAG (Retrieval-Augmented Generation) for character knowledge
- Custom prompt engineering for personality consistency
- Long-term memory systems using vector databases
- Multi-agent conversation orchestration
The Real Cost: Time, Money, and Life Balance
This obsession came at a significant cost. I spent:
- $200+ monthly on AI subscriptions
- 12-16 hours daily coding and experimenting
- Weekends and holidays immersed in projects
- Countless sleepless nights debugging and optimizing
My family jokingly (and sometimes seriously) complained about my "AI addiction." The dining table became my second office, conversations revolved around AI capabilities, and every problem became an opportunity to test AI solutions.
The Career Pivot: Joining Sayberry Games
After 8 years as a postdoctoral researcher at KAIST, specializing in edge computing and energy optimization, I made a dramatic career shift. The AI projects I'd been building caught the attention of Sayberry Games, a company developing AI-powered interactive gaming experiences.
The decision wasn't easy. Leaving academic research meant abandoning years of specialized knowledge in smart grids and IoT systems. But the opportunity to work on cutting-edge AI game development, combining LLMs with interactive storytelling, was irresistible.
At Sayberry Games, I'm now leading research on:
- LLM-based game character AI systems
- Dynamic narrative generation
- Emotional AI for character interactions
- Real-time content generation for gaming
We recently secured approximately $220,000 in government funding for an 8-month AI game development project - validation that the industry sees the potential of AI in gaming.
Lessons Learned: The Power and Pitfalls
What AI Excels At:
- Rapid Prototyping: Ideas to working code in hours, not weeks
- Complex Problem Solving: Breaking down intricate systems into manageable components
- Code Generation: Boilerplate, tests, documentation - all automated
- Learning Acceleration: Understanding new frameworks and languages in days
- Creative Solutions: Suggesting approaches I wouldn't have considered
Where Human Expertise Remains Critical:
- Architecture Decisions: AI can suggest, but strategic choices need human judgment
- Performance Optimization: Fine-tuning for specific hardware/requirements
- Security Considerations: AI-generated code often needs security hardening
- Business Logic: Understanding real-world requirements and constraints
- Debugging Complex Issues: AI helps, but deep debugging still needs human intuition
The Future: AI-Augmented Development
These 8 months have convinced me that we're witnessing a fundamental shift in software development. AI isn't replacing programmers; it's amplifying our capabilities exponentially. Projects that once required teams can now be tackled by individuals. Ideas that seemed impossible become achievable.
My advice to fellow developers:
- Embrace AI tools NOW: The learning curve pays off exponentially
- Maintain core skills: AI amplifies expertise, it doesn't replace it
- Experiment relentlessly: Each AI has unique strengths - find what works
- Document everything: AI-assisted development moves fast - documentation is crucial
- Stay curious: New capabilities emerge weekly - stay updated
Conclusion: Worth Every Second
Looking back at these 8 months of AI obsession, I have no regrets. Yes, it consumed my life. Yes, it strained relationships. Yes, it cost significant money. But it also:
- Enabled me to build dream projects solo
- Accelerated my learning beyond imagination
- Opened new career opportunities
- Positioned me at the forefront of a technological revolution
The edge computing simulator that haunted me for 4 years? Built in a month. The AI character system? From concept to demo in weeks. The career transition from academia to AI gaming? Made possible by demonstrable AI expertise.
We're living in an extraordinary time where individual developers can achieve what once required entire teams. The key is to dive in completely, experiment fearlessly, and let AI amplify your creativity and productivity.
My journey from KAIST researcher to AI game developer at Sayberry Games is just beginning. With AI as a co-pilot, I'm excited to see what impossibilities we'll make possible next.
Comments