The most valuable step in writing with AI isn't the prompt — it's compounding your edit feedback
Katie took AI from career coach all the way to a writing plugin: load the context in first, then save every round of edit notes back into the system and let it compound. For her, AI isn't only about output speed — it's a support system that steadies her emotionally and routes around the things she can't get herself to do.
The argument · tap a timestamp to hear it
Hire AI as a coach for the interrogation, not to think for you
After being laid off from an editing and ghostwriting job at a crypto company, Katie tried the client side, the agency side, and freelancing, and still felt she had run out of road on her own ideas. A real career coach was too expensive — she estimated around $150 an hour — so she turned to ChatGPT at $20 a month instead. She wasn't looking for something to think on her behalf; she used it as a surface to externalize thinking against: describe her situation, have it push back and ask follow-up questions, and above all pull her out of her own head when she was prone to catastrophizing. She later came to call this a ‘thinking tool’. She even used it to decide whether to accept Every's invitation to write a column — ChatGPT told her to try it, and that's the path that brought her here.
— Katie ParrottOne person shipping five channels runs on context, not typing speed
Freelancing in early 2025, Katie badly overcommitted: 8 blog posts, 3 ebooks, 24 LinkedIn posts, 24 posts on X, and 16 Instagram posts, all inside a two-week deadline. Getting through it wasn't about writing fast; it was what she only later learned to call context engineering — first writing out the company's brand messaging, product details, audience, and differentiation point by point, then letting the model generate. Building that context up front was laborious, but once the foundation was laid the production speed afterward became very fast. She admits she overcommitted the same way again six months later — the method did not cure the habit.
— Katie ParrottWithout the underlying information, style constraints just spin in place
On what goes into a project first, Katie is clear: not rules like ‘use these words, avoid those’ or ‘keep the reading level low’, but a base dossier on who the reader is, where the pain is, how the product maps to that pain, and who the competitors are and how you differ. She says this is the same instinct she had writing style guides in content marketing — only now it gets fed to the AI. Once the foundational material is in place, then it's time for constraints on tone, word choice, and syntax. Her position on ‘style’: the word-choice arguments will go on forever, but without the underlying information, style constraints just spin in place.
— Katie ParrottThe model is the kitchen; only a person can bring in fresh ingredients
The ‘data’ she means isn't vague background — it's firsthand material that either happened after the model's knowledge cutoff or exists only in the real physical world: research the company did itself, third-party reports, specific personal experience. If you only ask the model to write a ‘style guide’ out of what it already knows, what comes out is commoditized information. So the job she assigns to people is the ‘last mile’: bringing the new material the model doesn't have into the system. Her metaphor: the model is the kitchen and the workflow is the chopping and cooking, but the fresh ingredients have to be carried in by a human — the quality and freshness of the inputs is finally what decides whether a piece is worth reading.
— Katie ParrottAI as supportive technology is as powerful as AI as a production tool
Katie talked about a layer she rarely discusses publicly: she has bipolar disorder, which brings high-energy stretches and also very, very long lows. For her, the value of AI isn't only ‘high output’ — it's reducing the friction of daily life so that being a functioning person comes more easily. She had put off finding a primary care doctor for three years; in the end she had Codex find one nearby who was in her insurance network and accepting new patients, and book the appointment. Her inbox is automated too, filtering out the mail that actually needs a reply instead of leaving her to dig through piles of subscription notifications every day looking for a ‘bomb’. In her own words: AI as supportive technology is as powerful as productive technology.
— Katie ParrottA career coach isn't one prompt anymore — it's a project
That simple ‘be my career coach’ prompt has since become a standalone project inside Codex: her own dossier and role positioning, brand-positioning intelligence on Every, article performance data exported from the CMS, a validation folder that collects nice things readers have said, and her Q2/Q3 OKRs. A lot of that material she didn't compile by hand: following Alex Duffy's approach, she has the AI interview her, drawing the ideas out of her head and organizing them into documents. Increasingly she just talks to it out loud by voice, letting it set priorities, do project management, and maintain a board. She's even forgotten what's on the board — she just asks when she needs the due date for some deliverable.
— Katie ParrottWriting should work like engineering: correct the same problem only once
Compound Writing started as a fork of Kieran's Compound Engineering: in engineering, the system should remember every piece of feedback so the same problem is corrected once and handled correctly next time — writing should work the same way. She mapped the engineering version's brainstorming, planning, working, and reviewing onto brainstorming, outlining, drafting, and editing for writing, and designed three kinds of review: a substantive edit for overall structure and argument, a line edit for sentence-by-sentence polish, and a final publication-ready pass at the end. She no longer opens a web chat window; she runs the system in the Claude or ChatGPT desktop app, where the style guide, example articles, and other context all live, and the person only needs to bring in the fresh information. As of the recording, the plugin was already downloadable from Every's GitHub.
— Katie ParrottThis plugin is a gym, not a ghostwriting machine
She calls the plugin a gym: the goal is to make the user stronger, not to have AI write in their place. She has broken writers she loves into reviewer skills she can call: Vonnegut on starting as close to the end as possible, making every sentence earn its place, and giving the reader a character to root for; Hitchcock on the bomb under the table — if you let the audience see the countdown first, one second of shock becomes five minutes of suspense. She built those frameworks directly into the editing steps, so when a draft is done you can invoke Vonnegut, Hitchcock, Sorkin, or Sedaris to look at it. She thinks non-professional writers may benefit even more, because they don't have an existing attachment to ‘the way I used to write it was better’ in the way.
— Katie ParrottIn their own words · checked verbatim
Clarity, it turns out, doesn't arrive gift wrapped from a digital assistant or even a human coach. It's something I had to dig out for myself, question by question and prompt by prompt.
Natalia Quintero5:13
But when you have that environment set up, you can just kind of run and like run fast.
Katie Parrott10:28
So the data is the unique insight, the data point that the model doesn't have access to yet, the personal experience, all of those things that are working are ultimately going to make the writing unique.
Katie Parrott17:45
AI is just as powerful as a supportive technology.
Katie Parrott22:50
So compounding is the idea that every piece of feedback that you give to AI should feed back into the system to improve the next output.
Katie Parrott32:09
if there's a bomb under the table and it explodes, that's a one moment of surprise. But if you have people sitting around the table and you show the audience the bomb under the table and say this bomb will go off in five minutes, now you have five minutes of suspense.
Katie Parrott40:19
what I want to challenge the AI community with is thinking about how we can expand those kinds of opportunities to more people and more kinds of people in more different places.
Katie Parrott46:32
Figures
| The real career coach cost Katie compared against her ChatGPT subscription | about $150 an hour vs $20 a month | 4:12 |
| How long she put off booking a primary care doctor | 3 years | 23:53 |
Glossary
- Compound Engineering
- Having AI remember every piece of feedback so a correction is given once and applied automatically thereafter, with output that keeps getting better.
- Compound Writing
- Katie's port of compound engineering to the writing process: a plugin that accumulates style, structure, and editing feedback.
- context engineering
- Loading brand, audience, competitor, and other background information into the system before the model writes, setting the AI's boundaries.
- supportive technology
- AI that doesn't produce the content but lowers the friction of living and executing, making it easier to actually get things done.
- MCP
- A common interface standard connecting models to external data and applications; Katie uses it to wire small apps into the tools she already uses.
How to listen
Reporters and editors doing content, newsletters, or long-form pieces, plus indie developers building AI into a personal productivity system; especially useful if you're stuck with AI output that reads like everyone else's and want a workflow of your own.
The teaser before 1:04 and the sponsor read at 21:48 can be skipped; the main body runs valuable from 8:26 to the end.