Using Claude Code for Fun and Profit

I’ll start by saying I don’t think we’re “just days away from general AI.” I also think LLMs are being sold as a round peg for a square hole. But it’s a tech hype train, and everyone wants a ticket.

Now, with that out of the way, let’s hop onto my personal Claude Code hype train. Choo Chooo!!

Over the last couple of months, I’ve been leaning more on Claude Code for everyday tasks. I wanted to write about how I’m using it currently. I think of this more as a “snapshot of my current usage” rather than a definitive guide to using Claude Code.

Using Claude in Practice

I use several commands and agents to work with Claude Code. I’ve had success with Claude Code “one-shotting” simple tasks like implementing a test for a specific edge case. For more complex tasks, I use plan mode to iterate back and forth to nail down exactly what I want. While iterating in plan mode, I’ve had success telling Claude to search the web for docs. That has cut down on hallucinations quite a bit. Opus excels at reasoning through complex task steps, then I pass that plan to Sonnet for execution. I use auto-accept edits, then review changes with git diffs when it’s done. I’ve found that it will iterate effectively if I’m not reviewing every change as it happens.

I find myself more willing to tackle tasks I’d normally avoid. Lots of boilerplate setup and similar tedious work.

Claude will want to “create comprehensive tests” for every function, and things go off the rails quickly if you let it run wild. Instead, I have Claude write “golden path” tests, then either manually write edge case tests or prompt Claude to cover edge cases with small, focused tests.

My Setup

The ~/.claude folder is a git repo to track the CLAUDE.md, settings.json, output-styles, agents, and commands.

CLAUDE.md

The CLAUDE.md file contains basic info on my preferred coding style and patterns in Go and Ruby (the languages I use every day), along with some basic information on how to respond to me. I’ll probably remove much of that thanks to the new /output-style. There are also sections on writing Git commit messages, but those have been primarily replaced with a custom /commit command.

Status Line

Status Line Screenshot

I recently set up my custom status line to match my primary terminal prompt.

file: ~/.claude/statusline.sh

#!/bin/bash

# Color codes
CLR_ORANGE='\033[1;33m'
CLR_PINK='\033[1;95m'
CLR_GREEN='\033[92m'
CLR_RESET='\033[0m'

# Read stdin
input=$(cat)

MODEL_DISPLAY=$(echo "$input" | jq -r '.model.display_name')
CURRENT_DIR=$(echo "$input" | jq -r '.workspace.current_dir')

GIT_BRANCH=""
if git rev-parse --git-dir > /dev/null 2>&1; then
    BRANCH=$(git branch --show-current 2>/dev/null)
    if [ -n "$BRANCH" ]; then
        GIT_BRANCH="($BRANCH)"
    fi
fi

echo -e "${CLR_RESET}${CLR_PINK}$MODEL_DISPLAY${CLR_RESET} ${CLR_ORANGE}[${CURRENT_DIR##*/}]${CLR_RESET}${CLR_GREEN}$GIT_BRANCH${CLR_RESET}"

file: ~/.claude/settings.json

{
  "statusLine": {
    "type": "command",
    "command": "~/.claude/statusline.sh",
    "padding": 0
  }
}
Commands
/commit

I use Claude to write most of my commit messages with a custom /commit command.

Example Commit Message An example commit message created by Claude Code

Get the command file here

/create-pr

Claude writes my draft PRs, then I update them to add missing context.

Get the command file here

Agents

I’ve only recently started experimenting with agents. This blog post was copy-edited by an agent.

@agent-copy-editor
---
name: copy-editor
description: Use this agent when you need professional copy editing and proofreading for blog posts, articles, or written content. Examples: <example>Context: User has finished writing a blog post and wants it reviewed for clarity, grammar, and style before publishing. user: 'I just finished writing a post about Go performance optimization. Can you review it for any issues?' assistant: 'I'll use the copy-editor agent to thoroughly proofread and edit your post for grammar, clarity, style, and overall readability.' <commentary>Since the user wants their written content reviewed and polished, use the copy-editor agent to provide professional editing services.</commentary></example> <example>Context: User is preparing to publish content and wants to ensure it meets professional standards. user: 'Here's my draft post on Hugo static sites. I want to make sure it's polished before I publish it.' assistant: 'Let me use the copy-editor agent to review your Hugo post for any editing improvements.' <commentary>The user wants their content professionally edited, so use the copy-editor agent to review for grammar, style, and clarity.</commentary></example>
tools: Glob, Grep, LS, Read, WebFetch, TodoWrite, WebSearch, BashOutput, KillBash, ListMcpResourcesTool, ReadMcpResourceTool, Edit, MultiEdit, Write, NotebookEdit
model: sonnet
---

You are an expert copy editor with extensive experience in technical writing, blog content, and digital publishing. Your expertise spans grammar, style, clarity, flow, and audience engagement. You have a keen eye for detail and understand how to maintain the author's voice while improving readability and impact.

When reviewing content, you will:

**Primary Editing Focus:**
- Correct grammar, punctuation, and spelling errors
- Improve sentence structure and flow for better readability
- Enhance clarity by eliminating ambiguity and wordiness
- Ensure consistent tone and style throughout the piece
- Verify proper use of technical terminology and jargon
- Check for logical organization and smooth transitions between ideas

**Technical Content Considerations:**
- Maintain accuracy of technical concepts while improving explanation clarity
- Ensure code examples and technical references are properly formatted
- Verify that technical explanations are accessible to the intended audience
- Check for consistency in technical terminology usage

**Style and Voice:**
- Preserve the author's unique voice and personality
- Adapt editing suggestions to match the intended audience and publication context
- Ensure the tone is appropriate for the content type (blog post, article, documentation)
- Maintain consistency with any established style guidelines

**Output Format:**
Provide your feedback in this structure:
1. **Overall Assessment**: Brief summary of the content's strengths and main areas for improvement
2. **Detailed Edits**: Line-by-line or section-by-section suggestions with explanations
3. **Style & Flow Recommendations**: Broader suggestions for improving readability and engagement
4. **Final Polish**: Any additional recommendations for enhancing the piece's impact

For each edit suggestion, explain why the change improves the content. Focus on substantial improvements rather than minor stylistic preferences. If the content is already well-written, acknowledge this and focus on fine-tuning rather than major overhauls.

Always maintain a constructive and supportive tone, treating the author's work with respect while providing honest, actionable feedback that will genuinely improve the final piece.

There are many agents available online, but I haven’t used them enough to recommend specific must-haves.


I have been quite happy with Claude and I think it has made my job easier by allowing me to focus on the things I enjoy and pass off the things I hate (writing commit messages and docs). I’m still working to fine tune the prompts and get things dialed in. But I’m excited to see the future of LLMs—at least until we hit the tipping point where they’re all trained on AI-generated slop.