Cut the Token Bill on Both Ends
Two small tools that compound: Caveman shrinks what the agent says back, RTK shrinks what your terminal pipes in. More room in the same context window, same model, same prompts.
Topics
ai 
9 posts
Skills Over Agents
Why Claude Code skills beat specialized agents. On-demand context, not the model, decides quality. Build a skill library that travels with your code.
The Levels of AI Adoption
A six-level ladder of AI adoption, from copy-paste prompts to agentic teams and AI-native workflows. Where most companies stall, and how to climb.
Inside the .claude Folder
A hands-on tour of Claude Code's project folder. What rules, skills, agents, hooks, and settings each do, and how they fit together.
Building a Game in Two Days
A casual Telegram message turned into a nine-level Lord of the Rings browser game. Every line of code written by an AI agent. Every creative decision made by a human who grew up on Middle-earth.
Idealism vs Pragmatism
The tension between doing things right and getting things done shapes every decision in software and in life. TDD, pair programming, AI adoption: the best results come from holding both ends of the rope.
Build Your Own Team of Agents
Stop using AI as a single assistant. Organize agents into specialized teams with roles, docs, and parallel execution for real leverage.
MCP: Giving Your AI Agent the Right Context
MCP connects AI agents to your dev tools, databases, and APIs through a standard protocol. Give your agent the context it actually needs.
AI Gives You Speed, Not Quality
AI accelerates your output but doesn't ensure quality. You're responsible for every line you commit. Speed without direction is chaos.