Here is a workflow I’m currently testing with Claude Code.

To be honest, I’ve stopped reviewing the code line-by-line; I just let the AI handle everything. I won’t argue whether this approach is good or bad—everyone has their own perspective on this. However, if you are subscribed to the CC Max Plan, I highly recommend trying this method to build a full application.

Here is how it works:

First, write a spec file describing the application you want to build. Think carefully about the features and logic you need. Once that’s done, ask Claude to interview you about it. I personally used a custom slash command called /interview that I found on X (formerly Twitter).

md
---
description: Interview me about the plan
argument-hint: [plan]
model: opus
---

Read this plan file $ARGUMENTS and interview me in detail using the AskUserQuestion tool about literally anything: technical implementation, UI & UX, concerns, tradeoffs, etc. but make sure the questions are not obvious.

Be very in-depth and continue interviewing me continually until it’s complete, then write the spec to the file.

Claude will ask you a ton of questions, including multiple-choice ones. Just answer them. Once the process is finished, you will end up with a spec file that is far more comprehensive than your original draft.

After you have that finalized file, go to CC, run /clear, and then execute the following command (Note: I am using the Opus model for this):

md
use spec file, use sub agents as orchester and run them with /ralph-wiggum:ralph-loop and create change logs inside change-logs/{slug}.md so that sub agents are aware.
Opus will keep writing code until every feature in the spec is implemented. Based on my tests, the final result is actually quite impressive.

Opus will keep writing code until every feature in the spec is implemented. Based on my tests, the final result is actually quite impressive.

Notes: Since ralph-wiggum:ralph-loop is a Claude plugin, you will need to have it installed. I’ve provided the installation link below.

– https://github.com/anthropics/claude-code/tree/main/plugins/ralph-wiggum

Warning: This method consumes a lot of tokens, so you might hit your usage limits quickly. I suggest trying this on weekends or holidays when you have more time (and potentially reset limits).