KATIE PARROTT (Every Editorial) - Claude Code for writing

=== KATIE PARROTT (Every Editorial) - Claude Code for writing ===

(02:13:50): Hey Katie.

(02:13:50): Hi.

(02:13:53): Hell yeah.

(02:13:54): Well, Natalia and Sesh, thank you so much.

(02:13:56): That was great.

(02:13:57): We'll let y'all go and we'll bring on Katie.

(02:14:01): Katie is a staff writer at Every, has a regular column about working with AI.

(02:14:07): So Katie, why don't you introduce yourself a bit and tell us what you're going to present.

(02:14:13): Yeah.

(02:14:14): Bye guys.

(02:14:16): Anyway, my name's Katie.

(02:14:17): I'm a staff writer and AI editorial lead at Every.

(02:14:21): In addition to my column working overtime,

(02:14:23): where as Austin mentioned,

(02:14:24): I write about my experience working with AI and sort of bigger questions about how

(02:14:30): AI is changing labor.

(02:14:31): And then I also work on the operation side with,

(02:14:36): as far as our editorial team,

(02:14:38): thinking about ways that AI can support the content production process.

(02:14:44): And in addition to all of that, I am a baby vibe coder.

(02:14:48): I started by coding actually several think weeks ago.

(02:14:51): So people for people who don't know think week is one is a week per quarter where

(02:14:55): every steps back from the daily grind and just plays.

(02:14:59): And the task this time last year actually was to try an AI tool you'd never tried before.

(02:15:06): Our lead engineer Andre recommended Lovable and I coded about six different apps in

(02:15:12): a sort of two week vibe coding haze and just really got addicted to what I assume

(02:15:18): developers have been in on as far as the secret for forever,

(02:15:22): just the high of creating something and seeing it work.

(02:15:27): And so,

(02:15:28): but obviously in the year since that happened,

(02:15:32): capabilities have just come along so much.

(02:15:35): And like the tools that we use are different,

(02:15:37): the paradigms we use are all different than when I first started coding.

(02:15:41): So what I'd love to demo here is just how I'm using the cloud code,

(02:15:46): cloud code in the cloud app instead of the terminal.

(02:15:49): And then how I'm using the compound engineering plugin to sort of stand on the

(02:15:54): shoulders of giants and get that leg up to create better.

(02:15:59): more useful products with my extremely limited coding capabilities.

(02:16:04): So I will share my screen.

(02:16:06): Amazing.

(02:16:07): And we're going to start.

(02:16:09): So again, as I mentioned, I, can I share my entire screen?

(02:16:18): I built, where, how, window, entire screen.

(02:16:22): Okay, share.

(02:16:23): This way I can just click around and stuff.

(02:16:26): So this is one of the pieces I wrote a couple weeks,

(02:16:29): months ago about how I,

(02:16:31): it was August actually,

(02:16:32): about how I failed at building my first editorial ops project.

(02:16:38): So what that was was an AI editor.

(02:16:40): The idea was that it would take all of the things we know about what makes a good

(02:16:47): every essay,

(02:16:48): train the AI on it,

(02:16:50): and have the AI enforce those rules and provide feedback that a writer can

(02:16:54): implement before it goes to a human editor so the human editor can focus on big

(02:16:59): picture stuff.

(02:17:01): Is this matching our audience?

(02:17:02): Is it structured logically?

(02:17:05): All those big picture things and not telling me for the 50th time that I need to

(02:17:08): provide evidence or stop hedging with actually or I think so much.

(02:17:13): So this is what I built during Think Week in July.

(02:17:20): And you'll see it looks pretty,

(02:17:21): but if I click this every AI editor,

(02:17:24): nothing happens because I don't have a back end.

(02:17:27): I got stuck on the back end.

(02:17:29): I didn't know how like what to do or like how to sort of get there with the skills

(02:17:36): that I had at the time.

(02:17:37): So what I'd like to see if I can do and like this is completely live.

(02:17:40): I'm like just like.

(02:17:42): doing this from scratch on this call is to see if I can get farther in the

(02:17:48): functionality with the Cloud Code app.

(02:17:52): So here is the app.

(02:17:55): And one of my favorite things,

(02:17:57): it's such a small thing,

(02:17:58): but like one of the things that blocked me in the terminal is I just don't have the

(02:18:02): muscle memory and the retention of all the commands you need to navigate from

(02:18:06): folder to folder to make sure you're doing things right.

(02:18:09): And so just the interface here of being able to get this pull down menu is a

(02:18:14): massive unlock for me.

(02:18:15): So I'm just going to go into my folders and find the editor.

(02:18:20): And so now I'm here and I'm just going to monologue,

(02:18:22): which if you don't know,

(02:18:23): monologue is Every's Every's voice to text application built by Naveen.

(02:18:30): I'm extremely, extremely voice pilled.

(02:18:33): I wrote a whole essay about it for working overtime.

(02:18:36): And so what I'm going to tell it is,

(02:18:39): can you review this code base using the compound engineering

(02:18:44): plug in and identify opportunities for improvement to the code structure and the functionality.

(02:18:50): As someone with your similar background to this stuff,

(02:18:56): this is also my favorite thing to do is like,

(02:18:58): I'll get stuck and I will ask the agent when I have no idea what's going on to do

(02:19:04): an audit through compound engineering and through the Asian native architecture

(02:19:08): piece that Dan did.

(02:19:09): And I was like, oh yeah, that like, that looks good.

(02:19:11): Let's go in that direction.

(02:19:13): Yeah,

(02:19:13): when I say we're standing on the shoulders of giants,

(02:19:16): like just knowing what the resources that you need are and how to orchestrate them,

(02:19:23): you know,

(02:19:23): Dan has named this the allocation economy,

(02:19:26): right?

(02:19:27): Where you don't necessarily need the skills yourself, you just need to know what to ask for.

(02:19:32): And so that's the thing that

(02:19:34): I find really valuable being in the every ecosystem,

(02:19:37): you know,

(02:19:37): being so close to the metal,

(02:19:39): so to speak,

(02:19:39): on how the coders are building is I'm able to just kind of pick that up and sort of

(02:19:46): run with it.

(02:19:47): So we're herding.

(02:19:49): We're tinkering.

(02:19:51): It occurs to me that this is maybe not the most exciting thing to look at.

(02:19:55): And we're going to always allow- I think this is great.

(02:19:57): I'm curious, you mentioned this being much more accessible to you than the terminal.

(02:20:02): Do you feel like you work different ways with that level of confidence that comes

(02:20:08): from this just makes sense to me when you're playing in this space?

(02:20:11): I definitely am getting more and more ambitious in terms of what I ask for.

(02:20:16): just to see what is possible and where I run up to the limits.

(02:20:22): And again, the limit I run into, it gets farther and farther ahead.

(02:20:26): I'm just able to go further and get more done and get more done more quickly.

(02:20:34): So that's been really, just really encouraging and exciting to see

(02:20:39): So now it's diving deeper into the code quality and identifying specific

(02:20:42): improvement opportunities.

(02:20:44): So it's still crunching and determining.

(02:20:48): I do enjoy watching the,

(02:20:50): the,

(02:20:51): the plethora of different verbs that it comes up with for what do you,

(02:20:56): what do you,

(02:20:56): we've been asking people this,

(02:20:57): like when this is,

(02:20:58): when this is running,

(02:20:59): you know,

(02:20:59): it might take 10 minutes and that might take an hour.

(02:21:02): What's your way to kill the time as the call is doing the work for you?

(02:21:06): Yeah, so I actually wrote a piece about this, about AI-powered multitasking.

(02:21:10): And one thing I'll do is just come in and like work on a draft.

(02:21:13): Actually, our managing editor, Eleanor, is working on this draft that I'm working on right now.

(02:21:19): So I'll check on that.

(02:21:20): I'll do some tinkering on the AI editor that we're building.

(02:21:23): You know,

(02:21:24): I keep myself busy kind of doing this sort of like loop from just across the

(02:21:30): different tasks that I have running because all of them are running more or less

(02:21:33): agentically.

(02:21:34): I might have research running in the background.

(02:21:36): One of the things that I still use ChatGPT for is the pro setting for research.

(02:21:41): I've found that it's still a little bit more reliable in terms of surfacing active

(02:21:46): valid links and not hallucinating those links.

(02:21:49): So that might be a task that I'm doing, editing, as I said, writing.

(02:21:54): And then when I come back for something like this,

(02:21:57): it will,

(02:21:58): we can see here,

(02:21:59): it's got a whole laundry list of problems,

(02:22:01): dead code.

(02:22:03): missing content change, Jason.

(02:22:07): I'm curious, how much sense does this make to you?

(02:22:12): Do you scroll up and see that stuff?

(02:22:13): Do you just rely on the summary as someone who is coming from less of a tech background?

(02:22:19): I try to read the headlines.

(02:22:21): I skim the way I would skim an article to find the stuff that's interesting and important to me.

(02:22:25): This is another one of my favorite things to ask is,

(02:22:32): I should have said this in the original prompt, but I'm a newbie coder.

(02:22:35): Can you explain these changes to me like I'm five?

(02:22:39): Love it.

(02:22:42): Oh, wait.

(02:22:42): Okay.

(02:22:43): This is not Monologue, I should say.

(02:22:45): This is not Monologue's fault.

(02:22:46): This is something with the Claude app,

(02:22:49): and now it's not letting me find my previous chat,

(02:22:52): but if I just drop this in,

(02:22:54): so.

(02:22:55): And this will tell me like in simple terms, okay, what needs to change and why is it important?

(02:23:01): And this is one of the ways that I sort of help myself level up in my understanding

(02:23:06): of different things.

(02:23:08): Just like the different terminology, the different concepts, it's celebrating now.

(02:23:13): And that will give me a more accessible version.

(02:23:16): But,

(02:23:17): you know,

(02:23:18): I could probably,

(02:23:19): like,

(02:23:19): reviewing just these key findings would probably also work as well.

(02:23:24): You know, it wants us to split this into components, that makes sense to me.

(02:23:30): Missing backend,

(02:23:31): yeah,

(02:23:31): like the backend,

(02:23:32): I just,

(02:23:33): again,

(02:23:33): that's where I got stuck trying to,

(02:23:35): like,

(02:23:36): build in the functionality on the backend so that it has all of Every's style

(02:23:41): guidelines and stuff.

(02:23:43): All of that is just some stuff is a little bit more complicated.

(02:23:50): So I'm going to say like,

(02:23:53): all right,

(02:23:54): I'd like you to make a plan to implement these features,

(02:23:58): these fixes using compound engineering best practices.

(02:24:03): So again,

(02:24:03): I have the compound engineering plugin plugged in and there are like slash commands

(02:24:08): that you can use that do a lot of these things.

(02:24:11): But I found that with this,

(02:24:13): with the desktop app,

(02:24:15): I mean,

(02:24:15): just using plain language,

(02:24:17): I'm able to get it to invoke those skills.

(02:24:23): And so it's entering plan mode.

(02:24:26): and it's going to explore the code base more thoroughly to understand the existing

(02:24:30): patterns and create a solid implementation plan.

(02:24:33): And then it should be,

(02:24:35): if my understanding of compound engineering is correct,

(02:24:38): there are multiple different agents that Kieran has built into this plugin that

(02:24:44): perform different tasks,

(02:24:46): approach it from different perspectives,

(02:24:48): and you get the brain of a

(02:24:51): expert software engineer helping you build your app the way that it should be built

(02:24:56): and so this is again like you know like this is just miles better than anything

(02:25:03): that i could have done back in july when we were in upstate new york for think week

(02:25:07): and i was frantically trying to vibe code something to show a demo day and

(02:25:11): ultimately wound up having to

(02:25:13): Theranos it with a fake functionality.

(02:25:17): And I'm hopeful that, as Dan said, he's hoping to ship something by the end of today.

(02:25:24): I'm hopeful that I can get this stood up and see if it's more functional for us

(02:25:30): than our current solution,

(02:25:32): You know,

(02:25:32): as I mentioned,

(02:25:33): I failed building this last year and we realized that a Claude project just set up

(02:25:38): with custom instructions and some documentation worked for editing our pieces as

(02:25:42): well as what I was building and more reliably because it wasn't built on an app

(02:25:46): that I had to maintain.

(02:25:47): It was built on, it was, it's Claude's, it's Anthropics job to keep Claude projects functional.

(02:25:54): And so we've been having success with that,

(02:25:55): but something that we're looking to do from an editorial perspective is just get

(02:25:59): more visibility into when we accept the AI's recommendations,

(02:26:03): when we reject them,

(02:26:04): when they're more valuable,

(02:26:06): when the human input is more valuable.

(02:26:07): So my dream version of this would be that all of that happens in the same place,

(02:26:13): the editorial feedback,

(02:26:14): the measuring and tracking.

(02:26:17): I just want a one app to rule them all for every editorial.

(02:26:21): where all of these different tasks that are involved in content formation are being

(02:26:27): captured and we're getting that visibility into when we lean into AI and when we

(02:26:33): don't.

(02:26:34): Yeah,

(02:26:34): like a full production OS for us,

(02:26:37): but also one that could be replicable for anyone else in publishing.

(02:26:41): I got one more question before you go,

(02:26:43): Katie,

(02:26:43): because I think a lot of people,

(02:26:45): especially with no code backgrounds,

(02:26:46): are really vibing with this,

(02:26:48): especially seeing your approach to it.

(02:26:51): There's a question that I'm feeling too a lot, though, that's like...

(02:26:54): How do you manage the daily daycare of Claude when it's doing this stuff and

(02:26:59): running it without letting yourself get burnt out?

(02:27:02): We both got that question.

(02:27:03): I'll say I have no good answer for it.

(02:27:06): I am treating this like a video game and letting myself get burnt out and not

(02:27:10): setting healthy boundaries on it.

(02:27:12): And I probably should.

(02:27:14): Have you found a better approach?

(02:27:17): Honestly,

(02:27:18): my big thing is and what I really appreciate that AI has kind of freed up the

(02:27:22): energy and mental bandwidth for me to do is just to invest in like in my analog

(02:27:27): skill,

(02:27:28): like analog interests.

(02:27:29): So like Ash showed how when she's got downtime while an agent is working,

(02:27:33): she reads a print book.

(02:27:35): People who've been following me for any amount of time know that I'm involved in a

(02:27:38): project of reading the Bible.

(02:27:40): So I'll do Bible study.

(02:27:42): I'll play board games,

(02:27:43): which I really love because they're tactile and not a screen and you touch real

(02:27:47): things.

(02:27:49): I make time and kind of discipline myself to go off and do one of those grass

(02:27:54): touching type activities.

(02:27:56): And I find that that helps calibrate and keep me level.

(02:28:00): But everyone, sometimes to your point, you just have to give in and let the...

(02:28:05): let the fixation take over, and I have done more than a few nights.

(02:28:09): I don't know how healthy it is, honestly.

(02:28:11): I like what you're saying of taking a little walk.

(02:28:14): Yeah, once in a while, I'll live code till 2 a.m.

(02:28:17): This might be one of those days when I'm so amped by everything that's happening

(02:28:19): right now,

(02:28:20): to be honest.

(02:28:21): I'll probably be live coding all day today just because all of this is just making

(02:28:25): me so excited.

(02:28:27): Hell yeah.

(02:28:28): Well, thanks so much, Katie.

(02:28:28): That was awesome.

(02:28:30): We appreciate it.

(02:28:30): Please read Katie's column.

(02:28:32): It's a must read every week and we'll see you soon.

(02:28:36): All right.

(02:28:36): See you later.

(02:28:38): Thank you.

(02:28:38): All right.

(02:28:39): Dan is back.

(02:28:42): I'm going to tee it to you and go back to producing this thing.

(02:28:47): One second.

(02:28:49): Thank you so much.

(02:28:50): I meant to go outside, but I could not tear myself away from watching the live stream.

(02:28:54): So it was pretty good.

(02:28:57): There was some amazing stuff.

(02:28:58): Thank you, Austin, for hosting.

(02:29:00): If you're just getting here,

(02:29:03): Welcome.

(02:29:05): We are doing this all day.

(02:29:06): This is Vibe Code Camp.

(02:29:09): Austin just hosted a really,

(02:29:11): really awesome hour-long session with a bunch of our internal people from Every.

(02:29:15): Every is the only subscription you need to stay at the edge of AI.

(02:29:18): You should check us out, every.to.

(02:29:21): you get ideas.

(02:29:22): So a daily newsletter about AI apps.

(02:29:24): So we have four apps that we build internally that you get access to and training.

(02:29:28): So things like this,

(02:29:29): where we show you in in-depth deep dives into our workflows and the workflows from

(02:29:33): the most interesting,

(02:29:35): smartest and most productive builders in the world.

(02:29:38): That was a really awesome session.

(02:29:39): Thank you so much to Katie.

(02:29:41): I love,

(02:29:42): I love everything that you do and,

(02:29:43): and,

(02:29:43): and love how inspiring you are to everyone who is,

(02:29:46): Maybe non-technical and starting out for the first time.

(02:29:50): Now, I have the distinct pleasure to introduce a very, very good friend of mine.

(02:29:59): Before I do that,

(02:29:59): let me just say real quick,

(02:30:01): if you're interested in the agenda,

(02:30:02): we are going to be streaming all day today.

(02:30:04): We've been up for about almost three hours now.

(02:30:08): We'll be streaming until like 6 p.m., 6.30 p.m.

(02:30:11): Eastern.

(02:30:13): And if you're interested in the agenda, go to every.to slash agenda.

(02:30:18): And I'm going to bring on my good friend, Nat Eliason.