Practical AI. Real Projects.
Built After Hours.
I’m James. In the daytime, I manage technology and software projects in Singapore. At night, I build independent web applications, test frontier AI models, and automate real-world workflows.
This site is where I distill what actually works in AI, large language models, coding harnesses, and systems engineering — grounded by live, working applications rather than promotional hype.
Singapore Blue Hour · Project management by day, software craft by night
Featured Public Applications
Live, free-to-use web platforms built to solve genuine learning and recreational needs.
Professor Turtle
AI exam practice for Singapore students and parents
Generates syllabus-aligned practice questions for Primary and Secondary school topics. Features automated open-ended answer grading, handwriting & essay feedback, and instant worksheet uploads with Google SSO.
Puzzle Arena
7 multiplayer puzzle and board games with room codes & bots
An interactive real-time arena featuring Sudoku, Killer Sudoku, Nonogram, Word Search, Property Tycoon, Manor Mystery, and Scrabble. Powered by Fastify, React 19, Socket.io, and append-only event sourcing.
Learn AI Without the Noise
Concise, high-density guides designed to be read in under 90 seconds each. Practical mental models, architecture blueprints, and real-world trade-offs.
LLMs Demystified
Tokens, context windows, RAG, and prompt grounding explained with zero hype.
Coding with Agents
How to harness Claude Code, OpenAI tools, and OpenCode with strict guardrails.
Workflow Automation
From repetitive tasks to reliable pipelines: triggers, synthesis, and human review.
AI in Trading & Finance
Research, strategy modeling, backtesting discipline, and why risk limits come first.
Favorite Tools & Daily Drivers
Tested frameworks and tools I rely on for daily development.
Latest 90-Second Reads
Short, focused essays on engineering, AI boundaries, and system reliability.
AI for Busy Professionals: The Useful Mental Model
Treating LLMs as probabilistic interns rather than infallible gurus gives you predictable leverage without costly hallucinations.
Working with Coding Agents: Claude Code, OpenCode, and Clear Guardrails
Why top-down decomposition, tight review loops, and deterministic verification make coding agents 10x more reliable.
AI in Financial Trading: Research, Backtesting, and Why Risk Always Wins
Using machine intelligence for market research while keeping mathematical backtesting and hard circuit breakers non-negotiable.