CJ HESS (Tenex) - Context7 MCP

=== CJ HESS (Tenex) - Context7 MCP ===

(04:02:10): What's up, CJ?

(04:02:11): What's up, Brandon?

(04:02:12): How are you?

(04:02:12): Good to see you.

(04:02:16): I lost you there.

(04:02:17): I said great to meet you.

(04:02:18): Oh, yeah.

(04:02:18): Great to meet you.

(04:02:19): Thanks for having me on.

(04:02:20): I think everyone here is pretty excited to make an appearance.

(04:02:24): Yeah.

(04:02:25): So you're a 10x engineer at 10x.

(04:02:28): Oh, yeah.

(04:02:29): Double 10x.

(04:02:30): Maybe you can tell everybody what 10x is.

(04:02:33): Yeah.

(04:02:34): Maybe like a little bit about yourself and then what you want to get into today.

(04:02:38): For sure.

(04:02:38): Yeah.

(04:02:38): So my background,

(04:02:41): I went to Carnegie Mellon out in Pittsburgh,

(04:02:43): studied computer science,

(04:02:45): got out of school,

(04:02:45): did the whole like I'm going to make a massive startup.

(04:02:49): Did that a couple of times,

(04:02:50): went through it,

(04:02:51): eventually decided I wanted to make some money,

(04:02:54): ended up doing some contracting work kind of around mobile apps is my focus.

(04:03:00): And really started getting into the AI stuff,

(04:03:03): just using it more and more,

(04:03:04): feeling kind of the acceleration and also working as a contractor in an hourly

(04:03:11): capacity.

(04:03:11): I just felt like I was getting so much more work done.

(04:03:17): And that made me start to think about things like,

(04:03:19): okay,

(04:03:20): I'm outputting more,

(04:03:21): but I'm really getting paid less at the end of the day.

(04:03:25): At the same time,

(04:03:26): I was like exploring career opportunities and stumbled across 10X from a Twitter

(04:03:31): post by Alex.

(04:03:32): Basically, the 10X model has two pieces.

(04:03:36): One is traditional AI consulting,

(04:03:39): AI transformation,

(04:03:40): kind of coming into a business,

(04:03:41): finding the opportunities and the best places to plug in,

(04:03:45): whether it's like a pre-made workflow or something custom.

(04:03:47): And then also the side that is more like a dev shop,

(04:03:51): where we're doing general engineering tasks, but we're looking for those 10x engineers.

(04:03:58): And the kind of trick of the model is that we get paid by StoryPoint as engineers.

(04:04:04): So each sprint, we kind of scope it out, and then it's up to us to build it.

(04:04:09): And essentially,

(04:04:10): as tools get better,

(04:04:11): as we get better at utilizing the tools,

(04:04:13): our capacity can expand,

(04:04:15): and we're able to handle...

(04:04:17): projects across a ton of different domains it's such a crazy change in the business

(04:04:23): model of how to run like these sort of service type businesses around engineering

(04:04:28): going from i'm going to pay you per hour in which case you would be incentivized to

(04:04:31): like really take your time there's really like not that much incentive to like do a

(04:04:35): great job unless you want to like keep your job long term to hey ai exists now you

(04:04:41): can get things done really fast the more things you get done the more money you

(04:04:44): make

(04:04:45): And the best way to do that, obviously, is with AI, which is why you're a 10X engineer.

(04:04:51): And we really focus on the output side.

(04:04:54): So it's like, hey, this sprint, what feature can we ship?

(04:04:58): And how can we scope that to fit within this?

(04:05:00): And then how can we knock it out of the park?

(04:05:02): And everyone's incentivized to do better on that output as opposed to take their

(04:05:07): time and just rack up more hours.

(04:05:10): So what's your stack?

(04:05:11): How do you get shit done as a 10X engineer?

(04:05:15): Yeah, so I'm big on monorepos.

(04:05:19): I think that's kind of the way, right?

(04:05:21): Can you explain what that means to everyone?

(04:05:25): For sure, yeah.

(04:05:27): I think sometimes traditionally,

(04:05:28): especially at larger or older tech businesses,

(04:05:33): you have a back-end team,

(04:05:34): a front-end team,

(04:05:35): you have a DevOps team,

(04:05:36): you kind of have a bunch of people across...

(04:05:39): different stacks and different code bases.

(04:05:43): And really, the problem you run into there is, hey, I've implemented this new API endpoint.

(04:05:48): Now I need to fetch that data from the front end.

(04:05:50): If you're running Claude just within the front end,

(04:05:53): you're going to spend half your time explaining what data is coming in.

(04:05:55): It might not be right.

(04:05:57): It's going to be a lot of back and forth.

(04:05:58): When you're just sitting in the mono repo, you can kind of build the whole data flow.

(04:06:04): You can say, I need this new feature.

(04:06:06): It's going to store this data.

(04:06:07): You're going to pull this data.

(04:06:09): We're going to start at the database,

(04:06:10): build it up through the back end,

(04:06:11): and then we're going to build the front end for it.

(04:06:14): And that just lets us move so much faster,

(04:06:17): both from the coding side with Claude,

(04:06:19): but even just coordination,

(04:06:20): right?

(04:06:20): Like when we're talking about full stack devs,

(04:06:23): they can just move so much faster with these tools.

(04:06:25): What about like, because I mean, you're working with clients.

(04:06:28): I assume you have some really big clients.

(04:06:29): What about those clients that don't have mono repos?

(04:06:31): And I'm sure like some people watching are working in non-mono repos.

(04:06:36): How do you navigate that?

(04:06:39): There's kind of an interesting dynamic between a net new project,

(04:06:45): something that we've started and is a monorepo and is the stack we like,

(04:06:48): or if we're coming in to help an existing engineering team.

(04:06:51): Really...

(04:06:54): in those scenarios we just try to take the best practices things that we've kind of

(04:06:59): built up whether they're skills or prompts internally on how to handle either a

(04:07:04): different tech stack or kind of grab context from everywhere there's some like i'd

(04:07:09): call them like a frankenstein mono repo where you know we're a part of the front

(04:07:13): end and the back end directory so we can still reference them we actually run

(04:07:17): claude in like a higher directory

(04:07:20): And that usually gets us around most of the corners,

(04:07:23): but there's definitely some times where,

(04:07:24): you know,

(04:07:25): maybe we're only interfacing with a backend team and we have kind of the

(04:07:28): traditional coordination bottlenecks that any software team would.

(04:07:32): Yeah, that makes sense.

(04:07:33): Not as efficient.

(04:07:34): So what do you want to demo?

(04:07:37): Yeah, so there's a couple things.

(04:07:40): There's basically a new workflow that I've been playing with,

(04:07:44): and some people internally here have been playing with.

(04:07:47): I'm going to just hop in and present real quick.

(04:07:53): Let's see if this works.

(04:07:58): I think we're going to get an Infinimir here.

(04:08:00): Nice.

(04:08:01): Sweet.

(04:08:02): We are.

(04:08:03): Cool.

(04:08:04): So basically, I spun up this little app kind of as a clawed code guide.

(04:08:10): It's got a basic couple steps of onboarding and a homepage.

(04:08:14): And as I do this, like most people, I'm planning over here in Markdown files.

(04:08:21): And kind of one of the bottlenecks that I've run into in my flow is...

(04:08:27): situations like this where i prefer a lot of visual flows you know ui mock-ups if i

(04:08:34): scroll down here i have rough ui mock-ups yeah and you're doing all this in

(04:08:39): planning mode i assume

(04:08:41): Yeah, yeah.

(04:08:41): So I'll sit, you know, I'll sit in this markdown until this is almost exactly what I want.

(04:08:46): There are definitely times where I even go a step further and I'm like,

(04:08:50): write out all the code file by file.

(04:08:53): Give me some diffs.

(04:08:54): And I almost want to do a code review before we're actually touching the files.

(04:08:57): Yep.

(04:08:59): But kind of the big bottleneck I ran into here was whether it's a system diagram,

(04:09:04): a flowchart,

(04:09:05): a UI mockup.

(04:09:07): When I hopped into this view, it's really hard to kind of change what we want to see here.

(04:09:13): You know, let's say in this feature we're building, I want to build a quiz.

(04:09:19): So instead of just having this be like a knowledge base about Cloud Code,

(04:09:23): I'd want to quiz the user about it.

(04:09:26): So we have kind of like some basic state diagrams of what we're going to do.

(04:09:30): If I wanted to come in here and say, OK, I don't really want this to be multiple choice.

(04:09:35): I want it to flow from here to here to here to here.

(04:09:38): It ends up being like a long prompt that I often find gets convoluted.

(04:09:44): So kind of a tool that I'm going to flip back to my browser for.

(04:09:48): that we built internally is called Flowy.

(04:09:52): Basically, it's local Figma type setup.

(04:09:57): But in your IDE, these actually live as JSON files.

(04:10:02): So if I'm here, these diagrams actually correspond to what we see over here.

(04:10:07): And Claude gets a much better place where it can iterate on the flow.

(04:10:11): Okay, so just to say that back to you.

(04:10:13): So this is a file that lives in your system.

(04:10:14): Obviously, it's localhost.

(04:10:17): You're referencing, you're saving it in your directory.

(04:10:19): And you're asking in the plan, hey, reference this diagram.

(04:10:24): Don't actually reference what's in the plan itself.

(04:10:27): Yep.

(04:10:27): Yeah, I'll make often these are supplemental.

(04:10:31): So I'll make the markdown plan and kind of the first few steps are to make these diagrams.

(04:10:36): So it's,

(04:10:37): you know,

(04:10:37): when I was thinking about kind of onboarding,

(04:10:40): it was designed these steps first.

(04:10:42): Then we can talk about UI mockups, what we want it to look like.

(04:10:46): And you just get a much more visual view.

(04:10:49): And then you can even come in here and kind of change up some of the text.

(04:10:52): So say we want this to say guide.

(04:10:54): I can change this here.

(04:10:56): And this will end up saving

(04:11:00): I say that as it doesn't save on the screen,

(04:11:02): but it'll end up saving to that JSON file at least,

(04:11:04): and then Claude gets more and more context around exactly what you want to make.

(04:11:08): Okay, so I just want to read this back to you because this is a new paradigm for me.

(04:11:12): You have built Flowy, which...

(04:11:16): allows you to actually sketch out what an experience is going to be like,

(04:11:21): which is what your onboarding flow screen is.

(04:11:25): And then this is real mock-ups that then in your plan are being referenced.

(04:11:29): So when you say to Claude,

(04:11:30): okay,

(04:11:30): go execute on this plan,

(04:11:32): it's referencing back to onboarding flow and whatever this is.

(04:11:37): so that at the end of the day,

(04:11:38): you have the flow built to match what you built in onboarding flow and also the UI

(04:11:46): to look.

(04:11:47): It looked pretty close to this.

(04:11:49): Yeah, I mean, it's very lightweight.

(04:11:53): I won't claim that the UI in this little vibe-coded app is super complex,

(04:12:00): but it just gives that one step.

(04:12:02): If I'm back in my IDE and I'm looking in this initial plan,

(04:12:08): this UI mockup doesn't give me much signal.

(04:12:11): Yeah.

(04:12:12): And,

(04:12:14): you know,

(04:12:15): when we look at this one,

(04:12:16): I get a lot more signal on what this is going to look like,

(04:12:18): what the feel is going to be.

(04:12:19): And I'm able to come in and,

(04:12:21): you know,

(04:12:21): if I wanted to move this around or delete something,

(04:12:24): you know,

(04:12:24): you kind of have full editing ability.

(04:12:27): So it's almost like as a developer, I've become more of a designer.

(04:12:32): Yeah.

(04:12:32): yeah and then after running these steps i mean this flow implementation was you

(04:12:39): know a single prompt right and now we have this full flow built out exactly how i'd

(04:12:43): want it wow um is it is it like taking screenshots of that flow and adding giving

(04:12:51): that to claude is there like or is it like turning that into you know some sort of

(04:12:56): front-end code that then it's literally using here

(04:12:58): Yeah, so this JSON file, frankly, Flowy was pretty Ralphed.

(04:13:05): This was me experimenting with a Ralph loop on the weekend.

(04:13:10): But in here, it's very much so just JSON of like, what's our ID?

(04:13:16): What's the shape?

(04:13:17): What is the label going to say?

(04:13:19): Where is it?

(04:13:20): And this is something that I think is really kind of emergent out of the newer version

(04:13:27): generation of models i don't think a sonnet 4 would do this great for example um

(04:13:32): just because of all the spatial you know reasoning it has to do about where

(04:13:36): different things live in the diagram but this really you know i would say is a

(04:13:42): rough set of styles and layout that then claude's really good at looking at this

(04:13:49): and saying okay in react i need this component this component this component here's

(04:13:53): how they're going to stack up and it's going to look like this

(04:13:56): I mean, I can tell you just like for me, and I'm seeing from a lot of the comments too.

(04:14:02): So I'm not technical, but I have a design background.

(04:14:06): So I think very visually.

(04:14:09): Yes,

(04:14:09): I can like look at plans,

(04:14:10): but I run into the same issue as you where like actually describing what I want to

(04:14:14): build.

(04:14:15): When I think about what I want to build, I think about it visually.

(04:14:17): I don't think about it in code.

(04:14:19): So something like this would be amazing for...

(04:14:22): sort of the non-technical designer people to actually be able to build something

(04:14:28): much faster than you would in markdown files.

(04:14:30): Cause it's just such a pain in the ass to like go back and forth with a plan and

(04:14:34): write out,

(04:14:35): you know,

(04:14:36): in words,

(04:14:37): what you want.

(04:14:39): I got tired of looking at these ASCII boxes.

(04:14:42): Yeah.

(04:14:43): You know, the nomenclature is different and you switch the words up by mistake.

(04:14:47): So it's, it's really, it's really cool.

(04:14:51): Yeah,

(04:14:52): and basically what I've sketched out that I'll kick off a Claude and then we can

(04:14:57): talk while it works and actually watch some of the new charts come up.

(04:15:00): Yeah.

(04:15:01): Sketched out this new feature as a quiz.

(04:15:04): And in this first few steps of the plan are basically saying, hey, create the flow diagrams.

(04:15:12): Here's the file you're going to create.

(04:15:13): Let's put it in the flowy directory.

(04:15:16): And here's roughly what it's going to look like.

(04:15:20): So just hopping to Claude,

(04:15:21): I'm just going to say,

(04:15:22): great,

(04:15:22): based on that plan,

(04:15:26): please use the flowy skill and create the diagrams and mockups.

(04:15:36): And in here, I've created two skills for Flowy.

(04:15:40): One is around flowcharts and one is around UI mockups.

(04:15:45): Both of these are kind of just how does Flowy work?

(04:15:50): How did the layouts work?

(04:15:51): How does the JSON work?

(04:15:53): And then also kind of some specifics about what's our goal in a mockup?

(04:15:56): What's our goal in a flowchart?

(04:15:58): Yeah.

(04:15:59): And then just so I understand that in the markdown file,

(04:16:03): in the plan,

(04:16:04): you're saying,

(04:16:05): you know,

(04:16:05): usually skills make this file,

(04:16:07): but you're also giving it a rough outline of what you want before you then go ahead

(04:16:10): and make it.

(04:16:11): Yeah, for sure.

(04:16:12): So I've kind of,

(04:16:14): A lot of the times I'll use whisper flow or something to transcribe or dictate what

(04:16:18): I'm going to say.

(04:16:22): I'll just ramble, you know, just give it as much context as possible.

(04:16:26): And then I have a prompt that roughly creates a structured plan.

(04:16:32): First few steps of it are...

(04:16:34): usually something i outline but if not i have it infer like what diagrams would be

(04:16:39): helpful for this one i said you know i want to know the navigation flow which is

(04:16:44): roughly like this right we're going to enter the main app going to go to the learn

(04:16:48): tab there's going to be a quiz card that's going to take us through these different

(04:16:53): pages this is great this is kind of exactly what i was looking for and described um

(04:16:59): the other one i had mentioned was this gameplay flow

(04:17:03): And this is, hey, what question are we on?

(04:17:06): Did we get it correct?

(04:17:07): What happens when we get it correct?

(04:17:09): Kind of this whole visual mapping of what's actually going on in this quiz.

(04:17:13): And the mock-up one is going to take Claude a second,

(04:17:16): but when that pops up,

(04:17:17): it'll actually show kind of the screens that we're going to build.

(04:17:20): Yeah.

(04:17:21): Can you talk about, like, just how impactful voice-to-text has been for your flow?

(04:17:26): It...

(04:17:29): So I was a hater of all the transcription type tools for a while.

(04:17:36): But only recently did I start using Whisperflow.

(04:17:40): And I found...

(04:17:43): One, the speed is awesome.

(04:17:44): Just being able to kind of ramble and talk much faster than you can type makes it

(04:17:50): so much quicker to interact with Cloud.

(04:17:52): I didn't realize how much of my time was truly just writing prompts.

(04:17:56): And the other piece is I feel like when I'm typing,

(04:17:59): There's some inherent laziness where I don't want to spend the time to type out

(04:18:05): this perfectly exact prompt.

(04:18:07): I'm fine if it infers a couple of things.

(04:18:09): But if I'm just rambling, I'm way more specific.

(04:18:12): I'm way more detailed.

(04:18:14): I'll give it context that might even not be related from what I think,

(04:18:18): but it actually helps the model solve the problem.

(04:18:20): So it's been a pretty big unlock for me.

(04:18:23): When we're here in the office, I don't do it as much, but absolutely clotting on the weekends.

(04:18:28): I am I'm only talking.

(04:18:30): Yeah,

(04:18:30): I think that I found it to be a huge unlock to every has a voice to text tool

(04:18:36): called monologue that that our team uses.

(04:18:39): And one thing that Kieran just built in compound engineering that's really cool is

(04:18:43): I will monologue into

(04:18:46): his compound engineering skill,

(04:18:48): um,

(04:18:49): or plugin and then ask it to interview me back,

(04:18:52): interview me on like things that it doesn't understand.

(04:18:55): Cause I'm rambling as well.

(04:18:56): Anyway, let's get into these mockups.

(04:18:59): Yeah.

(04:18:59): Mockups are finished.

(04:19:00): Um, basically again, this is kind of looking pretty similar to the app we've built.

(04:19:07): Um,

(04:19:08): Nothing too,

(04:19:09): too complex with this feature,

(04:19:10): but something about like seeing it like this as opposed to seeing it in ASCII just

(04:19:15): makes me so much more confident the model is going to actually do it right.

(04:19:19): Yeah.

(04:19:21): And then, you know, let's say I want to get rid of this box.

(04:19:26): I can just delete this.

(04:19:28): I'll delete the inner box and then I'll come back over to Claude and I'll say I

(04:19:35): made some changes to the mock-up.

(04:19:39): What did I change?

(04:19:41): I usually like to ask something like this just to confirm understanding,

(04:19:45): just to make it reread the file and see what I've changed.

(04:19:52): And hopefully, Claude's not down again.

(04:19:55): Great.

(04:19:55): So I don't know.

(04:19:57): It didn't look like it read any files, but somehow it knows.

(04:20:01): But it's noticed that I removed it.

(04:20:04): That's great.

(04:20:05): Awesome.

(04:20:06): So now I'm just going to say, based on the flow charts and the plan, build it.

(04:20:16): And now we'll kind of just sit back and end up seeing this come to life.

(04:20:22): I'll click into here and then hop back over to these flow charts.

(04:20:26): But once Claude's done there, we should have that quiz feature kind of built out on this app.

(04:20:31): Awesome.

(04:20:33): Is there anything else that you want to talk about in the meantime?

(04:20:36): Otherwise, I have a question.

(04:20:40): Well,

(04:20:40): I do want to know a little more about the transcription tool you guys have built

(04:20:44): and how it's been the most impactful for your engineers.

(04:20:52): Yeah, so it's an engineer on a team named Naveen.

(04:20:56): He built out Monologue, which is a Mac app.

(04:21:01): We're also launching iOS February 9th around then.

(04:21:07): And I'd say the powerful tool that he built that our team loves is a feature called modes.

(04:21:13): So basically,

(04:21:16): It's like an editor can go over what you say depending on which product you're using.

(04:21:20): So if you're in ChatGPT versus Warp versus Cursor,

(04:21:24): you can ask it to sort of like edit on your behalf in very specific ways.

(04:21:29): So the team really likes that.

(04:21:31): It also just like sort of is like this beautiful skeuomorphic design that like I

(04:21:35): personally love.

(04:21:37): And then one feature that it has that has been really helpful for me,

(04:21:40): I used it this morning on a hike,

(04:21:42): is a really good notes feature.

(04:21:44): So if I have an idea while hiking or just doing anything,

(04:21:47): you can create a great note with a transcript and summary that I like that feature.

(04:21:53): Yeah.

(04:21:55): But yeah, our team is just like very, very voice-filled, for sure.

(04:22:00): When Naveen is gonna be coming on,

(04:22:02): somebody will drop it in the chat,

(04:22:03): I forget what time,

(04:22:04): but he's coming on like,

(04:22:05): I think around like four or five to talk through monologue.

(04:22:09): Yeah, that's super cool.

(04:22:11): I do love the idea of almost passing prompts through an LLM.

(04:22:15): In a way, that's what we're doing in plan mode, right?

(04:22:18): It's,

(04:22:19): hey,

(04:22:19): I want to build this,

(04:22:20): go learn some more,

(04:22:21): and then write a big markdown file that I'm basically going to use as a prompt for

(04:22:25): another agent.

(04:22:26): Yeah,

(04:22:27): I mean,

(04:22:27): take what I'm saying and make it much better,

(04:22:29): much more thoughtful,

(04:22:30): much more built out.

(04:22:31): Yeah.

(04:22:32): Yeah, it's super powerful.

(04:22:33): So question for you, unless this is actually built the quiz yet.

(04:22:37): Yeah, I was just checking.

(04:22:38): Looks like Claude's thinking about this one, but be there soon.

(04:22:43): So what's next for Flowy?

(04:22:45): I'm seeing a lot of people are interested in it.

(04:22:48): It seems like a weekend project.

(04:22:51): What would you build as a next step for it?

(04:22:53): And would you ever want to bring it to market?

(04:22:56): Yeah, there's definitely a couple ideas around making this available.

(04:23:01): I really have spent a weekend on this and have found it impactful,

(04:23:06): so definitely want to carve out time to build it further.

(04:23:09): I think there's kind of some great examples.

(04:23:13): If I look at Figma, basically I would want

(04:23:18): 70% of the functionality of just the design of Figma and then have Claude be able

(04:23:27): to just work with the JSON off that.

(04:23:29): If we can get to that point,

(04:23:31): there's no reason these mockups can't even get higher and higher fidelity and then

(04:23:37): you know, like Figma's dev mode, this almost just operates as that already.

(04:23:42): And just hands.

(04:23:44): What's stopping you from like using Figma's MCP to, to do what you're doing at Figma.

(04:23:49): Um, there's kind of two pieces.

(04:23:53): One is speed.

(04:23:55): Um,

(04:23:56): I guess I'd call it three.

(04:23:57): There's speed and I'm anti-MCP.

(04:24:01): I find that skills and CLIs are often just as capable and also kind of give you way

(04:24:09): less context bloat.

(04:24:13): I'm liking the skill paradigm more and more because of that.

(04:24:16): But also that MCP is good and

(04:24:22): from what I've found, for writing.

(04:24:24): So it'll kind of make a design really well.

(04:24:27): But if I need that to come into Claude and be read,

(04:24:30): I've struggled to have it kind of do a one-to-one match.

(04:24:33): So I almost want the specificity of that full JSON file to actually build this out

(04:24:38): in the project I'm working on.

(04:24:40): Yeah.

(04:24:41): I mean, it just seems like such a no-brainer for Figma to figure that out.

(04:24:46): In the meantime, I want Flowy, so I don't know.

(04:24:51): Maybe you can get access for the Everyteam.

(04:24:53): Hey, I'll spend another weekend on it.

(04:24:56): Perfect.

(04:24:58): Well, looks like we're here.

(04:25:00): All right.

(04:25:01): Let's see if we can pull it off.

(04:25:03): This is looking solid so far.

(04:25:05): We're going to start the quiz.

(04:25:07): Oh, and we failed.

(04:25:08): You're not sharing your screen now.

(04:25:11): Oh, well, that's good because it threw a bug.

(04:25:15): Okay, it happens.

(04:25:18): Are we sharing this screen?

(04:25:19): Yeah.

(04:25:20): Nice.

(04:25:22): If I start it,

(04:25:24): it looks like there's a navigation bug,

(04:25:26): so it appears that whole navigation flowchart didn't work out.

(04:25:31): What would you do in this scenario?

(04:25:33): Walk us through how you'd solve this.

(04:25:35): Yeah.

(04:25:36): The first step, just kind of from a mobile background, is I'm going to reload this real quick.

(04:25:42): Seems like it could have been a hot reload issue.

(04:25:45): Going to try it again.

(04:25:47): Failed this time.

(04:25:50): So I would often say, I would grab this error.

(04:25:53): I'm going to give this right back to Claude.

(04:25:56): And I'm going to say, this error is happening when I press get started.

(04:26:05): Then I'm actually going to point it back to the charts and the plan because there's

(04:26:09): kind of two places where I find the models trip up.

(04:26:12): One,

(04:26:13): they don't follow the plan,

(04:26:15): particularly if anyone out there is still using Claude Sonnet 4.5.

(04:26:21): I usually only pull it up if I need the million context window.

(04:26:25): But that guy loves to just not follow the plan.

(04:26:29): But if there's a discrepancy there, this won't end up working if that plan was correct.

(04:26:34): And then also,

(04:26:36): if the plan was wrong and it implemented that,

(04:26:39): kind of sending it on a bug fixing mission.

(04:26:42): So I'll give it the error.

(04:26:43): I'll tell it when it happened.

(04:26:44): And I'll say, reference the original plans.

(04:26:48): Did we follow them correctly?

(04:26:53): If so...

(04:26:55): retrace the navigation flowchart and the code and tell me what is causing this

(04:27:08): error in the chat.

(04:27:12): I often like to add an in the chat to have almost a side conversation

(04:27:17): Claude is historically eager and just giving it that prompt without that,

(04:27:22): it'd probably start writing a bunch of code.

(04:27:23): But, you know, I kind of want to look at this.

(04:27:26): At this point,

(04:27:26): I'd probably hop back into the IDE and start digging through some of these files

(04:27:30): and,

(04:27:31): you know,

(04:27:31): seeing if there's anything obvious that's going wrong.

(04:27:35): But if we're talking a pure like vibe coding check, this is how we would do it.

(04:27:41): Again, Claude doesn't seem to have read anything, and it seems to have written code anyway.

(04:27:46): Yeah.

(04:27:50): Would you ever use, like, Ralph Loop here to just have it keep running until it figures it out?

(04:27:56): Because, like, this seems pretty... This is probably a pretty simple solve.

(04:27:59): Yeah.

(04:28:00): I've...

(04:28:01): So I've experimented with Ralph.

(04:28:04): I need some lessons from Ryan Carson, but I can never get it to exactly do what I want.

(04:28:12): It almost feels like a little bit of an abstraction.

(04:28:14): That's great when I'm building something like flowy where I'm not super opinionated

(04:28:19): and I'm kind of trying to get to a MVP on a side project.

(04:28:23): But if it's

(04:28:24): you know, some production system, I'd be terrified of Ralph.

(04:28:29): I feel like I use a lot of that human in the loop control.

(04:28:31): Okay.

(04:28:33): So perfect for flow on the weekend, not for a 10 X client.

(04:28:37): Yeah.

(04:28:37): Yeah.

(04:28:38): That's it's, it's been fun to play with.

(04:28:41): I like the idea of kind of getting to the world where,

(04:28:45): particularly the bash loop style,

(04:28:47): Ralph,

(04:28:48): where it's really spinning up brand new agents,

(04:28:50): it's re-anchoring them,

(04:28:51): and they're tackling one piece of the task.

(04:28:54): But in terms of giving up that control, I'm still not fully comfortable with that yet.

(04:29:01): Yeah.

(04:29:02): Okay.

(04:29:04): Let's see if Paul is able to fix it.

(04:29:08): Final try.

(04:29:08): We did it.

(04:29:10): Nice.

(04:29:12): Wow.

(04:29:12): Let's see if I can get these right.

(04:29:16): Bang.

(04:29:17): This is a little Claude tutorial for everyone watching.

(04:29:21): Yeah, this is cool.

(04:29:22): Ooh, this one?

(04:29:23): Nice.

(04:29:27): I don't know this one.

(04:29:31): Yeah.

(04:29:32): Ooh, okay.

(04:29:33): There we go.

(04:29:34): I am jumping back in.

(04:29:36): Hello, everybody.

(04:29:37): CJ and Brandon, how's it going?

(04:29:40): Dan, do you think you could pass this exam?

(04:29:42): I don't.

(04:29:42): Probably not.

(04:29:43): Probably not.

(04:29:44): I don't think I'd pass this exam.

(04:29:46): I was close.

(04:29:49): CJ, thank you so much.

(04:29:50): That was great.

(04:29:52): CJ, thank you.

(04:29:53): Awesome.

(04:29:53): Thanks for having me on.

(04:29:54): Have a good one, man.

(04:29:56): Brandon, thank you for hosting.

(04:29:57): I will see everyone later with Naveen and Yash.

(04:30:02): Sweet.

(04:30:02): See you soon.

(04:30:04): I am back.

(04:30:06): Just a quick note to say that this podcast is brought to you by hydration.

(04:30:14): If you are vibe coding,

(04:30:16): it's very important that you drink water and maybe even feed your agent water as

(04:30:20): well.

(04:30:22): And if you're just getting here,

(04:30:23): we are about to enter one of the most fun conversations of the day.

(04:30:28): But before we do,

(04:30:31): Uh,

(04:30:31): you should subscribe to every every is the only subscription you need to stay at

(04:30:35): the edge of AI.

(04:30:37): When you subscribe to every you get three things, ideas, apps and training.

(04:30:40): On the ideas side, we have a daily newsletter about AI.

(04:30:42): It has everything you need to know to keep you up to date on what's new, what's interesting.

(04:30:47): We test all the models before they come out and give you a day of vibe check to

(04:30:50): tell you what to use and what not to use.

(04:30:52): We also have a suite of apps that we've built.

(04:30:55): So we have four apps, Cora, Sparkle, Spiral and Monologue that help you work better with AI.

(04:31:00): Everything from doing your email and AI email assistant to an AI ghostwriter to a

(04:31:04): speech to text app.

(04:31:06): And we also do trainings.

(04:31:08): So if you're if you like this live stream,

(04:31:10): we do these sorts of camps twice a month for paying subscribers.

(04:31:16): We have the whole team come on and demo their workflows,

(04:31:20): the latest tools that they're using,

(04:31:21): how they're thinking about building things.

(04:31:22): We also have people from the top AI companies in the world like OpenAI and Cursor

(04:31:27): come on to show you their workflows.

(04:31:28): It's all for one subscription.

(04:31:30): Go to every.to slash subscribe to sign up.

(04:31:33): What are you waiting for?

(04:31:34): OK, now I have two amazing people, Logan and Amar.

(04:31:40): Welcome.