Vibe Code Camp Distilled

Katie Parrott - Building with Claude Code and Compound Engineering

Katie Parrott - Building with Claude Code and Compound Engineering

Key Insights

Summary

Katie Parrott is a staff writer and AI editorial lead at Every, where she writes the “Working Overtime” column about her experiences working with AI and questions about how AI is changing labor. In this presentation, she demonstrates how she uses the Claude Code desktop app with the compound engineering plugin to continue building an AI editor tool she originally started during Every’s “Think Week” but got stuck on due to backend complexity.

Katie describes herself as a “baby vibe coder” who started just weeks earlier but has rapidly adopted AI-assisted development. Her presentation showcases how someone from a non-technical background can leverage expert-level plugins and natural language prompts to audit codebases, generate implementation plans, and build functional applications. The talk emphasizes accessibility - from using voice-to-text for coding prompts to asking AI to “explain like I’m five” - while building toward Every’s ultimate goal: a unified editorial operations system that tracks when AI recommendations are accepted versus rejected.

Main Topics

From Lovable to Claude Code: The Evolution of Vibe Coding

Katie started her vibe coding journey during Every’s Think Week (one week per quarter for experimentation) when lead engineer Andre recommended the tool Lovable. She built six different apps in a “two week vibe coding haze” and got addicted to what she calls “the high of creating something and seeing it work” (00:01:25).

However, capabilities and tools have changed dramatically since then. She now uses Claude Code in the desktop app instead of the terminal, finding it far more accessible: “one of the things that blocked me in the terminal is I just don’t have the muscle memory and the retention of all the commands you need to navigate from folder to folder to make sure you’re doing things right. And so just the interface here of being able to get this pull down menu is a massive unlock for me.” (00:04:18)

The AI Editor That Failed (And What Replaced It)

Katie’s original project during Think Week in July was building an AI editor for Every’s content team. The vision: “it would take all of the things we know about what makes a good every essay, train the AI on it, and have the AI enforce those rules and provide the questions that we have. So that a writer can implement before it goes to a human editor. So the human editor can focus on big picture stuff.” (00:02:54)

The problem: “I got stuck on the back end. I didn’t know how, like what to do or like how to sort of get there with the skills that I had at the time.” (00:03:36) She managed to build a pretty frontend but couldn’t implement the actual functionality.

Quote: “I frantically trying to five code something to show at demo day and ultimately wound up having to Theranos it with a fake, fake functionality.” (00:11:26)

The practical solution: “We realized that a cloud project just set up with custom instructions and some documentation worked for editing our pieces as well as what I was building and more reliably because it wasn’t built on an app that I had to maintain. It was built on - it’s Claude’s - it’s Anthropics job to keep Claude projects functional.” (00:11:56)

Standing on the Shoulders of Giants: The Compound Engineering Plugin

Katie’s strategy centers on leveraging expert knowledge embedded in plugins: “When I say we’re standing on the shoulders of giants, like just knowing what the resources that you need are and how to orchestrate them. Dan has has named this the allocation economy. Where you like you don’t necessarily need the skills yourself. You just need to know what to ask for.” (00:05:32)

Her typical workflow when stuck: “this is also my favorite thing to do is like I’ll get stuck and I will ask the agent when I have no idea what’s going on to do an audit through compound engineering and through the Asian native architecture piece that Dan did. And I was like, oh, yeah, that like that looks good. Let’s go in that direction.” (00:05:18)

In the demo, she asks Claude to: “review this code base using the compound engineering plugin and identify opportunities for improvement to the code structure and the functionality” (00:04:59)

Reading Code as a Beginner: The ELI5 Approach

When Claude returns technical analysis of code problems, Katie has a clear strategy: “I try to read the headlines. You know, I skim the way I would skim an article, like to find the stuff that’s interesting and important to me. But ultimately what it is going, and this is another one of my favorite things to ask is like, I should have said this in the original prompt, but I’m a newbie coder. Can you explain these changes to me like I’m five?” (00:08:29)

This “explain like I’m five” approach is “one of the ways that I sort of help myself level up in my understanding of different things. Just like the different terminology, the different concepts.” (00:09:16)

AI-Powered Multitasking and Avoiding Burnout

When asked how she manages waiting for long AI processes, Katie describes her parallel workflow: “I actually wrote a piece about this about AI powered multitasking. And one thing I’ll do is just come in and like work on a draft… I’ll do some tinkering on the AI editor that we’re building… I keep myself busy kind of doing this sort of like loop from just across the different tasks that I have running because all of them are running more or less agentically.” (00:07:19)

For research tasks, she still uses ChatGPT Pro: “One of the things that I still use ChatGPT for is the pro setting for research. I found that it’s still a little bit more reliable in terms of surfacing active valid links and not hallucinating those links.” (00:07:50)

On avoiding burnout: “I am treating this like a video game and letting myself get burnt out and not setting like healthy boundaries on it. I probably should.” (00:13:15) However, she also invests in analog activities: “I make time and kind of discipline myself to like go off and do one of those grass touching type activities… play board games, which I really love because they’re tactile and not a screen… do Bible study… And I find that that helps calibrate and keep me level.” (00:13:38)

The Vision: A Unified Editorial Operations System

Katie’s ultimate goal extends beyond the current editor tool: “something that we’re looking to do from an editorial perspective is just get more visibility into when we accept the AIs recommendations, when we reject them, when they’re more valuable, when the human input is more valuable.” (00:12:16)

Quote: “My dream version of this would be that all of that happens in the same place, the editorial feedback, the the measuring and tracking. I just want a one app to rule them all for every editorial where all of these different tasks are being that are involved in content create formation are being captured. And we’re getting that visibility into when we when we lean into AI and when we don’t.” (00:12:25)

As Austin summarized: “like a full like production OS for us, but also one that could be like rubricable for anyone else in publishing.” (00:12:50)

Voice-First Coding with Monologue

Throughout the demo, Katie uses Monologue (Every’s voice-to-text application built by Naveen) to give commands to Claude. She describes herself as “extremely, extremely voice pilled” and wrote a whole essay about it for Working Overtime. (00:04:39)

Growing Ambition with Growing Capabilities

Asked if the accessible interface changes how she works: “I definitely am getting more and more ambitious in terms of what I ask for just to see what is what is possible and where think where I run up to the limits. And like, again, like the limit I run into, it gets farther and farther ahead. Just that I’m just able to go further and get more done and get more done more quickly.” (00:06:21)

Actionable Details

Tools Mentioned

Workflow Demonstrated

  1. Navigate to project folder using Claude Code’s visual folder navigation
  2. Use voice-to-text (Monologue) to give natural language prompts
  3. Ask Claude to review codebase using compound engineering plugin
  4. Request “explain like I’m five” versions of technical output
  5. Generate implementation plans using compound engineering best practices
  6. Let Claude enter “plan mode” where multiple specialized agents review the code from different perspectives
  7. Cycle between multiple parallel agentic tasks while waiting for long processes

Key Prompts Used

Technical Context

Quotes Worth Saving

“When I say we’re standing on the shoulders of giants, like just knowing what the resources that you need are and how to orchestrate them. Dan has has named this the allocation economy. Where you like you don’t necessarily need the skills yourself. You just need to know what to ask for.” (00:05:32)

“one of the things that blocked me in the terminal is I just don’t have the muscle memory and the retention of all the commands you need to navigate from folder to folder to make sure you’re doing things right. And so just the interface here of being able to get this pull down menu is a massive unlock for me.” (00:04:18)

“I definitely am getting more and more ambitious in terms of what I ask for just to see what is what is possible and where think where I run up to the limits. And like, again, like the limit I run into, it gets farther and farther ahead. Just that I’m just able to go further and get more done and get more done more quickly.” (00:06:21)

“My dream version of this would be that all of that happens in the same place, the editorial feedback, the the measuring and tracking. I just want a one app to rule them all for every editorial where all of these different tasks are being that are involved in content create formation are being captured.” (00:12:25)

“I am treating this like a video game and letting myself get burnt out and not setting like healthy boundaries on it. I probably should.” (00:13:15)

← Back to Index ← Natalia Nityesh Nat Eliason →