Learn
Short, honest guides. We build a workspace that runs Claude Code, Codex, Gemini CLI and friends side by side all day, and these pages are where we write down what we learn: what the terms actually mean, which tools hold up, and how to keep agents under control. Dated founder notes live on the blog.
The definition, where the phrase comes from, how it works in practice, where it shines, where it bites, and how to do it without shipping chaos.
9 min read BasicsWhat separates an agent from autocomplete: the read, plan, edit, run loop, what agents do well in 2026 and where they still fail.
8 min read ToolsClaude Code, Codex CLI, Gemini CLI, Copilot, Aider and more, compared honestly from running them side by side every day.
10 min read How-toA practical first-week setup: guardrails, outcome prompts, checkpoints, and the beginner mistakes we see constantly.
8 min read How-toThe terminal closed mid-task. How to bring the session back in Claude Code and every other CLI agent, and lose nothing next time.
7 min read ToolsThe full stack, layer by layer: CLI agents, AI-first editors, prompt-to-app builders and the visibility layer that keeps it sane.
9 min read AdvancedPatterns that work, file collisions, cost control and orchestration: a field guide to multi-agent coding on one codebase.
9 min read AdvancedPromote one agent to a boss that plans a goal and delegates to the rest automatically: queueing, crash recovery and auto-hiring.
7 min read Claude CodeWhat SKILL.md files are, where they live, and how to import a whole library at once from a folder or another agent.
6 min read How-toSet the font, size and Nerd Font glyphs for Claude Code in any terminal, or with a built-in per-agent picker.
6 min read PlatformRun Claude Code in a 3D workspace on Linux: install the AppImage or the .deb on Ubuntu, Mint, Arch and more.
4 min readThe Termi Protocol gives every CLI agent a body, a desk and a room. Every file read, command and change becomes something you can actually see.