Ever felt like the worst coder in the world? One such developer set out to prove that even the most inexperienced can build an agentic AI to dominate a leaderboard. This journey isn't just about automation—it's a story of grit, learning, and the unexpected rewards of embracing failure. Below, we dive into the highs and lows of this ambitious project.

Who is the 'Worst Coder in the World'?

The self-proclaimed 'Worst Coder in the World' is a coding newbie who openly admits to struggling with even basic programming concepts. Despite years of dabbling, they never felt proficient—until they decided to tackle an advanced project: building an AI agent to crack a competitive leaderboard. This title reflects their honesty about their skill level, not a lack of effort. Their story resonates with many who feel imposter syndrome in tech. The goal was to learn by doing, embracing mistakes as stepping stones rather than roadblocks. As they put it, being the 'worst' is a starting point, not a label. For more on the project's motivation, see why build an agent.

Why Did a Coding Newbie Decide to Build an Agent?

The motivation was twofold: solve a real-world problem and accelerate learning. The leaderboard they wanted to crack required repetitive tasks that could be automated. Instead of grinding manually, they saw an opportunity to create an AI agent that could handle the grunt work while they focused on understanding the code. More importantly, they believed that building something tangible—even if imperfect—would teach them more than endless tutorials. The agent became a vehicle for learning: each bug, each failed attempt, each rerun forced them to research, experiment, and improve. It wasn't about being the best coder overnight; it was about proving that anyone can make progress with persistence. Check out the challenges faced during development.

What Does 'Going Agentic' Mean in This Context?

'Going agentic' refers to building an autonomous software agent—a program that can make decisions and take actions without constant human input. In this project, the agent was designed to interact with a online leaderboard system, perform tasks like submitting entries, and analyze results to improve its strategy. Unlike simple scripts, an agent has a degree of independence: it can adjust its behavior based on feedback. The term has become buzzworthy in AI circles, but for this newbie, it simply meant creating something that could 'think' and act on its own within a narrow domain. The result wasn't perfect, but it demonstrated the core ideas of agentic AI: observation, action, and learning. For a deeper dive, see the technical stack used.

What Were the Biggest Challenges?

The challenges were numerous and humbling. First, the coder had to grasp basic Python syntax and libraries like Selenium for web automation. They often got stuck on simple bugs—indentation errors, missing imports, or incorrect XPath selectors. The leaderboard itself added complexity: it had anti-bot measures that required the agent to mimic human behavior with random delays and mouse movements. Another hurdle was debugging asynchronous processes; the agent would sometimes hang or crash without clear error messages. Perhaps the hardest challenge was staying motivated when everything seemed to fail. But each obstacle taught a lesson: how to read error logs, search for solutions, and ask the right questions. The journey was messy, but that messiness was the real education. Read about the rewards and lessons.

What Rewards and Lessons Came from the Experience?

Despite the struggles, the rewards were immense. The agent eventually worked, cracking the leaderboard and climbing ranks—a proud moment for someone who couldn't write a loop a month earlier. More importantly, the coder gained practical skills: they could now debug confidently, understand API calls, and appreciate how AI agents are structured. They learned that failure is not the opposite of success but part of it. The biggest lesson? You don't need to be a genius to build something useful—you just need to start, iterate, and never give up. The project also sparked a passion for coding, turning a former 'worst coder' into someone who now enjoys building tools. The full journey is detailed in the original story, but this Q&A captures the essence.

What Technical Stack Did the Agent Use?

The agent was built using a simple but effective stack: Python as the core language, Selenium for browser automation (to interact with the leaderboard UI), and a custom loop with condition checks to handle decisions. For data storage, a lightweight SQLite database tracked past submissions and results. No fancy frameworks like LangChain or OpenAI APIs—this was a DIY approach. The coder even wrote basic rule-based logic for strategy (e.g., if rank drops, try a different submission time). While not cutting-edge, the stack was chosen to minimize complexity and maximize learning. The entire codebase was less than 500 lines, proving that agentic AI doesn't always require massive infrastructure. It worked because it was tailored to the specific task.