From one loop to a professional, multi-tool workflow

Free preview 3 min to read

The loop was never the whole job

"AI tools: introduction (Cursor & Antigravity)" named the loop underneath every agentic tool: read a file, write a change, run a command, read the result, repeat. Its example had an agent add published? to an Article model: open app/models/article.rb, write the method and run the test suite. "AI tools: Claude & Gemini", the course before this one, went a layer deeper, into the model doing the reasoning inside the loop. Neither covered what "professional" adds once the loop and the model are settled: a workflow built around project context, staged review, and more than one agent running at once.

This app's MEMORY.md already made a cameo

"AI context: @Codebase & rules", an intern course, used this project's MEMORY.md as its running example and quoted its Layer Contract, Service Object Pattern and Forms with Namespaced Models sections. This course doesn't repeat that. What that course didn't get into is what happens when the file an agentic tool should read automatically doesn't have the filename any tool looks for. MEMORY.md isn't CLAUDE.md, AGENTS.md or GEMINI.md. Lesson 2 comes back to that gap.

Two tools, for two different reasons

This course uses Cursor, introduced at the intern level, and Gemini CLI, which these AI courses haven't touched yet. They aren't chosen to repeat the Claude-versus-Gemini comparison, which was about the model inside the loop. Cursor and Gemini CLI differ somewhere else. Cursor is an IDE, a full editor with a visible diff and a review flow in its interface. Gemini CLI, Google's open-source terminal agent, runs the same read, write, run, iterate pattern from a shell prompt. It's scriptable and pipeable, and it can run in a CI job with no window to click through. It's the same loop in two different working contexts.

What "professional" means in this course's title

It doesn't mean fancier prompts. It means three specific things. First, giving either tool this project's context correctly, beyond having a rules file exist somewhere (lessons 2, 3 and 5). Second, reviewing a plan before any file changes, instead of only reading a finished draft (lesson 4). Third, reviewing work that ran while you weren't watching, such as a cloud agent's branch (lesson 6). Each is a documented capability, checked against Cursor's and Google's current documentation while writing this course.

What this course covers

Lesson 2 goes past the intern course's basic rules mechanism into Cursor's decision logic for which rules apply where, and the gap this project's MEMORY.md falls into. Lesson 3 introduces Gemini CLI, checked against its documentation and open-source repository. Lesson 4 covers Cursor's Plan Mode, a staged review workflow for tasks bigger than one file. Lesson 5 works through an example from this app's own conventions. Lesson 6 covers cloud and parallel agents, and reviewing work you never watched happen. Lesson 7 pulls it together and closes with a challenge that models lesson 2's rule-matching logic.

Lesson 1 of 7 · free

Keep going with lesson 2

Cursor's rules system, past the basics

Covers Cursor's four-branch rule logic, User Rules limits, nested AGENTS.md precedence, and why MEMORY.md does not auto-load.

A free account also gives you the quizzes, your progress and notes, and the discussion.

Type to start searching...