The latest articles on DEV Community by JessYT (@jessyt). https://dev.to/jessyt https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3929396%2F258b3757-050e-4d7f-9578-f524506551bc.png https://dev.to/jessyt en JessYT Wed, 13 May 2026 14:15:09 +0000 https://dev.to/jessyt/three-months-with-claude-code-what-35-jobs-and-26-killed-ones-taught-me-3alf https://dev.to/jessyt/three-months-with-claude-code-what-35-jobs-and-26-killed-ones-taught-me-3alf <h1> Three Months With Claude Code: What 35 Jobs and 26 Killed Ones Taught Me </h1> <p>Bottom line up front. <strong>AI support cut 90% of my writing time.</strong> Three months in, I built 35 automation jobs and killed 26. Volume alone doesn't make money — that part I confirmed the hard way.</p> <ul> <li><strong>3 months in</strong></li> <li> <strong>35 jobs</strong> active</li> <li><strong>26 killed</strong></li> <li><strong>90% writing time saved</strong></li> </ul> <h2> 01 · What I have to show for 3 months, in numbers </h2> <p>Start with the tangible stuff. This is the automation surface I've built up over three months. Every morning, launchd runs the lot without me touching anything.</p> <ul> <li> <strong>35 LaunchAgent jobs</strong> (26 killed)</li> <li> <strong>12 project folders</strong> (5 blogs + α)</li> <li> <strong>3 custom skills</strong> (only SKILL.md loads)</li> <li> <strong>19 slash commands</strong> (<code>/run-daily</code> &amp; co.)</li> </ul> <p>On top of that: <strong>13 CLAUDE.md files</strong> (1 main + 12 per-project) and <strong>3 hook scripts</strong>. Every job ships with a wrapper, plist, and log file — so the actual file count is well past 100.</p> <h2> 02 · Five things that actually save tokens </h2> <p>In Claude Code, context tokens are the bill. After three months, these are the five habits that stuck.</p> <ol> <li> <strong><code>/compact</code> — the moment the context warning fires.</strong> When the alert pops, hit it immediately. Keeps the gist, drops the rest.</li> <li> <strong>Split out agents — separate roles for detail and deep work.</strong> Main session is the coordinator; details go to a sub-agent. Keeps the context from getting polluted.</li> <li> <strong><code>/clear</code> — every time the task switches.</strong> Going from blog work to infra debugging? Clear, no exceptions.</li> <li> <strong>Keep CLAUDE.md slim — only the must-reads.</strong> It auto-loads every session. Push the rest into separate files and reference them only when the task calls for it.</li> <li> <strong>Skill split — load only SKILL.md.</strong> My three custom skills sit in their own files. The full text stays out of context until needed.</li> </ol> <h2> 03 · The three I actually use every day </h2> <p>Out of 19 commands, 3 skills, and a pile of plugins, these are the three I genuinely reach for daily.</p> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Rank</th> <th>Tool</th> <th>How</th> <th>ROI</th> </tr> </thead> <tbody> <tr> <td>A · Daily publish</td> <td><code>/run-daily ×5</code></td> <td>per-blog <code>.claude/commands/run-daily.md</code> </td> <td>edit one line → all 5 blogs pick it up</td> </tr> <tr> <td>B · Writing quality</td> <td><code>blog-style-guide</code></td> <td>9 base patterns + 4 evolved patterns in one file</td> <td>tweak the guide once → every post follows the new rule</td> </tr> <tr> <td>C · Design</td> <td><code>frontend-design</code></td> <td>installed 4/26, generates v2 visual components</td> <td>3–5 polished components per post, on demand</td> </tr> </tbody> </table></div> <p>The <code>frontend-design</code> plugin in particular changed what I can ship. Before it, I was hand-rolling CSS and the readability suffered. Now Hero, VS, and impact blocks come out clean in one pass. <strong>Eleven days in, the lift in visual quality is immediate.</strong></p> <h2> 04 · Automation ≠ revenue — the hardest lesson </h2> <blockquote> <p>"I assumed automated posts would mean automated income. Three months later, it's the opposite."</p> </blockquote> <p><strong>Root cause</strong>: automation cuts repetition cost. Revenue is built on trust. You can ship 5 posts a day, but if every one reads like a textbook, readers click and bounce.</p> <p>Look at the asset count again. <strong>35 active</strong> + <strong>26 killed</strong>. Those 26 are the ones I built on the assumption that volume alone would pay off.</p> <p><strong>Early assumption (26 killed jobs):</strong></p> <ul> <li>"Auto-publish daily"</li> <li>Generic / textbook tone</li> <li>Conversion flatlined</li> </ul> <p><strong>After the pivot (35 active jobs):</strong></p> <ul> <li>Experiment logs + lived experience</li> <li>Drafts → my own review</li> <li>Reads trending up</li> </ul> <p>Automation cuts repetition cost. Revenue compounds from trust. Two completely different mechanisms — I had to kill 26 jobs to internalize that. Volume without trust is worse than zero posts.</p> <h2> 05 · CLAUDE.md ≠ memory — keep them separate </h2> <p>Two different contexts auto-load every session. Mix them up and things break.</p> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Criteria</th> <th>CLAUDE.md (project file)</th> <th>Memory (auto-memory)</th> </tr> </thead> <tbody> <tr> <td>Holds</td> <td>Operating rules, publish rules, incident logs</td> <td>Meta info, persona, where external systems live</td> </tr> <tr> <td>Update freq</td> <td>Often (every time something breaks)</td> <td>Occasionally</td> </tr> <tr> <td>Trust</td> <td>High — source of truth for ops</td> <td>Supporting — meta only</td> </tr> </tbody> </table></div> <p>The split rule is simple. <strong>"This needs to work in the next session"</strong> goes in CLAUDE.md. <strong>"Who the user is, where they work"</strong> goes in memory. Don't trust memory for operating rules — always pin them in CLAUDE.md. I caused several incidents over three months by getting this wrong.</p> <p>One more thing: don't put it in either place. <strong>Concrete values (cash balances, prices) or transient state.</strong> They go stale fast — push them to a file or a database instead.</p> <h2> 06 · Pull quote </h2> <blockquote> <p>"You still need a human's lived experience. Trust beats volume. But AI support did cut 90% of my writing time."</p> <p>— Eddie, 3 months with Claude Code</p> </blockquote> <h2> 07 · Six follow-ups I'll go deeper on </h2> <p>This post is the index. Each topic gets its own case-study post next.</p> <ol> <li> <strong>Token saving deep dive</strong> — each of the 5 habits, in practice</li> <li> <strong>Slash commands deep dive</strong> — how <code>/run-daily</code> differs across 5 blogs</li> <li> <strong>3 custom skills deep dive</strong> — why I built them, how they're shaped</li> <li> <strong>Running CLAUDE.md</strong> — how I let incident logs accumulate</li> <li> <strong>frontend-design plugin</strong> — the 30-day review</li> <li> <strong>Hooks, memory, agents</strong> — the rest of the stack</li> </ol> <p>Productivity and trust aren't substitutes. <strong>AI buys back the time. The human fills in the trust.</strong></p> <p><em>Disclaimer: personal 3-month retrospective. No ads, no affiliates.</em></p> <p>Originally published at <a href="https://jessinvestment.com/three-months-with-claude-code-what-35-jobs-and-26-killed-ones-taught-me/" rel="noopener noreferrer">jessinvestment.com</a>.</p> aicoding automation claudecode developerproductivity