Skip to content

Latest commit

 

History

6 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 

Repository files navigation

The Loop: agentic coding, defined properly

Read it here: https://colincalnan.github.io/agentic-coding-lessons/

A short course for developers who already ship code and have been using coding agents daily without ever being handed the vocabulary. Eight lessons, a glossary, and a self-check.

The argument

Every term in this field is a label for one part of a single diagram: the agent loop. So the page opens with that diagram, makes it clickable, and hangs all the vocabulary off it.

An agent is a model in a loop with tools. That one sentence is the whole field. Everything else is zooming in on one box.

The load-bearing lesson is the one on harnesses, because it is the most under-explained word in the space:

The model is maybe 80% of the outcome and the harness is the other 20%, and the 20% is what differs between products. The same model in two different harnesses behaves like two different developers.

If you run Codex CLI or Claude Code, you already operate a harness, drive an agentic loop, and do context engineering. You just have not been given the words.

The eight lessons

00 The model An agent is a model in a loop with tools. Three components, three families of jargon.
01 Harness What the software wrapped around the model does on every turn: assembles the prompt, advertises tools, executes calls, manages the context budget, enforces policy.
02 Agentic A spectrum of who holds the loop, from autocomplete to delegated background agents. In the chat era, you were the harness.
03 Tools and MCP Tool calling is structured text the harness acts on. MCP is the portable packaging format.
04 Context The window, context engineering, compaction, RAG. Why your AGENTS.md is context engineering.
05 Subagents Context isolation first, parallelism second.
06 Rails Sandboxing, approval modes, prompt injection, framed in infosec terms.
07 Evals SWE-bench and friends, and why a ten-task private eval beats any leaderboard.

Then a one-line-per-term glossary and five questions with instant feedback.

How it is built

One self-contained HTML file. No build step, no dependencies, no analytics, no network calls. Open it locally, or drop it on any server to share with a team.

Related

Companion to Loops: from prompter to loop designer, which covers designing and running agent loops in practice. This one covers what the words mean. Read this first if the vocabulary is the blocker.

Licence

Copyright (c) 2026 Colin Calnan. Free to read, link, and share.

About

The Loop: agentic coding defined properly. A short course on harnesses, agents, context and evals for developers who already ship code.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages