Lessons Learned From Vibe Coding Plugins  🔗 **Source:** [Lessons Learned From Vibe Coding Plugins](https://www.youtube.com/watch?v=8AULmW_V-hE) > [!note]+ Document Notes > --- > tags: [vibe-coding, wordpress-plugins, ai-hallucination, multi-model-red-teaming, wordpress-hosting] > author: WP Minute Podcasts > source: WP Minute Podcasts (YouTube) > reference: https://www.youtube.com/watch?v=8AULmW_V-hE > date created: 2026-09-29 10:32 > --- > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > # 🏷️ Title: Lessons Learned From Vibe Coding Plugins (August 2026 Member Meetup) > ## 🧭 Table of contents > - [📝 Summary](#-summary) > - [📖 Main content](#-main-content) > - [🗒️ Notes](#️-notes) > - [📚 References](#-references) > - [🗣️ Paraphrase](#️-paraphrase) > - [📰 Newsletter](#-newsletter) > - [✅ Action items](#-action-items) > - [🧪 Hypotheses](#-hypotheses) > - [📊 Methodology](#-methodology) > - [❓ Questions](#-questions) > - [🔮 Future research](#-future-research) > - [💡 Implications](#-implications) > - [🧠 Learning styles assessment](#-learning-styles-assessment) > - [📊 Accessibility assessment](#-accessibility-assessment) > - [🌍 DEIB assessment](#-deib-assessment) > - [✅ Logical validity](#-logical-validity) > - [🛡️ DARVO analysis](#️-darvo-analysis) > - [🎯 Conclusion](#-conclusion) > - [🔄 Refactor appendix](#-refactor-appendix--lessons-learned-from-vibe-coding-plugins-august-2026-member-meetup) > ## 📝 Summary > - **What happened.** The WP Minute Podcasts channel posted a 54:39 recording of its August 2026 member meetup (uploaded 2026-09-28, 79 views at capture). Two hosts and one guest talk: the hosts are named "Eric Caravac" and "Matt Maderos" in the auto-captions (the description says "Eric and Matt"; the second is almost certainly Matt Medeiros, and the first surname is garbled, so verify before citing). The guest is Patrick Gallagher of GridPane (captions: "Grid Pain," "Grid Pane"). The transcript carries no speaker labels; attribution below is inferred from who introduces whom and from first-person statements. > - **The claim.** Patrick argues that AI coding tools now produce working software in minutes, that small open models run on roughly $5,000 of hardware, and that hallucination rates stay high enough (92% for the model Patrick names, 30% for a 27B open model) that a person who can answer for the code still has to sit in the loop. Patrick's working method is to build in one model, then have two or three other providers red-team the result. > - **The WordPress angle.** Patrick says AI site builders start with Next.js or React "0% of the time" with WordPress, offers two hypotheses (dependency surface, training-data sentiment), and concedes that WordPress is the wrong fit for a static brochure site. Patrick predicts utility-plugin demand falls while niche, support-backed products hold value. > - **Why it matters for plugin security and review.** The meetup is the practitioner's view of the exact problem the Plugins team faces: more submissions written by models that "sometimes" follow security guidelines (00:26:17), reviewed by authors who may not understand the output. Patrick's control (multi-model review plus an accountable human) is a process claim, not a tested one. > - **Key stakeholders.** Patrick Gallagher and GridPane (a Pillar Sponsor of The WP Minute, per the editor's note in the description), the two hosts, WordPress plugin authors and buyers, the WordPress.org Plugins and Core AI teams, hosts who move very large WooCommerce databases. > - **Risks and tradeoffs.** Every benchmark number comes from slides the transcript can't show (63 on an "AGI" scale, 52 for Qwen, 92% and 30% hallucination rates). The guest sells WooCommerce hosting and a search-replace tool built in-house, and the sponsor relationship is disclosed only in the description. The WP-CLI failure claim at 00:38:02 is an assertion with no test shown. > - **Bottom line.** The process advice (cross-model review, human accountability, serve a specific customer) is sensible and cheap to adopt. The quantitative claims and the causal story for why models skip WordPress are unverified in this source. > ## 📖 Main content > **The demo opens the argument.** Patrick shows a 14-slide deck that Claude Design produced from one prompt: Patrick told it Patrick had a talk in 30 minutes, to pull the logo from gridpane.com, and to cover benchmarks and hallucinations, and it found a specific post by Jeffrey Emanuel (captions: "Jeffrey Emanuel," spelled correctly) and built a slide around it. Patrick's point is capability, not polish: "design" is the weakest computer-use skill Patrick sees, and even that clears the bar. > **Capability history, told as a personal anecdote.** Patrick studied AI starting in 1998 ("25 years too early") and dates the turn to December 2024, when OpenAI released "01 pro," which Patrick calls the first reasoning model. Before it, models "would just puke out an answer" without checking it. Patrick tested with constraint satisfaction problems (the Einstein or zebra riddle) and got a correct answer after seven minutes. Contested detail: OpenAI's o1 reasoning models shipped earlier in 2024, and o1 pro mode arrived in December 2024, so "first reasoning model that ever existed" overstates it. Patrick's own experience (the December 23rd family test) is the evidence, and it is anecdote. > **Benchmarks and local models.** Patrick checks Artificial Analysis about five times a week. On the slide Patrick reads out, "quad opus 5" (Claude Opus 5, presumably) sits at 63, and Qwen 3.8 27B (captions: "Quinn") at 52, next to a model captioned "GPT 5.6 six Luna." Patrick's claim: a $5,000 home machine runs a model that beats what o1 Pro was two years ago, so businesses needn't hand data to Google, OpenAI, or Anthropic ("we are the product at this point"). Patrick then extends the point to people: Patrick's wife, a physician, should be paid for the prompts, and "actual intelligence" beats model intelligence. > **Why models skip WordPress.** Two hypotheses appear. Matt offers the dependency story (00:16:10): one TypeScript file is easier to pattern-match than a PHP, frontend, backend and database stack. Patrick adds a sentiment story (00:22:41, and Matt picks it up at 00:43:07): the training data is full of people saying WordPress is old or dumb, and the models also inherit people "second-guessing themselves." Patrick also says very little of WordPress code is in training corpora and offers to explain how Patrick knows, but the transcript never returns to it. > **Cost and reliability from an agent run.** On Factory.ai's Droid "missions" feature, a run had gone 34 hours 10 minutes and about $200 in tokens over 36 hours. Patrick swapped the orchestrator to Kimi K3 (captions: "Kimmy K3," from Moonshot), which looped on "and then I'll write the plan" for about 3 hours. Patrick blames training data full of humans second-guessing themselves. That causal claim has no test behind it. > **Hallucination rate.** Patrick reads Artificial Analysis's Omniscience hallucination rate: a named OpenAI model ("GPT 5.6 uh soul," garbled) at 92%, Command A at 14%, Qwen 3.8 27B at 30%. Meaning per Patrick's reading: when told not to answer if unsure, the model answers anyway. Patrick ties it to "helpful helper" instructions and notes humans also guess confidently. > **Method for plugin work.** Two answers to the security question at 00:26:17. First, hold two to four paid accounts and red-team code across providers: build in Claude, have OpenAI's model sanity-check it, have Grok iterate on it. Second, "you need to have a human being in the loop." Patrick's case study is a database search-and-replace tool for very large WooCommerce sites, 24 hours cut to about an hour by Patrick, 10 minutes by Patrick's CTO, and 10 seconds by a later idea, handed between Patrick and a co-founder 12 to 13 hours apart so the code is worked on around the clock. > **Product judgment.** With execution cheaper, ideas matter more (00:35:35). Patrick's search-replace tool matters to "maybe 500 people," and a small site finishes with WP-CLI in two minutes. The strategic lesson is to serve a defined customer who will pay, such as golf course superintendents (two businesses a [00:00:02] Hey everyone, welcome to the WP Minutes August 2026 member meetup. I'm Eric Caravac joined by my friend Matt Maderos and our special guest this week is Patrick from Grid Pain. Patrick, welcome. >> Thanks for having me, guys. Appreciate you. Yeah. So, uh we're going to be talking about uh pros and cons, the uh good, bad, and the ugly of uh vibe coding your own plugins, apps, uh all that good stuff. And Patrick's got a uh a slide deck he's going to show us in a minute. Before we get to that, before we get to that, I do want to just thank all of our uh pillar sponsors, including Grid Pane, uh Pressible, Jet Host, WP [00:00:48] Engine, Hosting.com, 20I, and Wiglot. We thank all of these great sponsors for uh helping us do what we do at the WP Minute because we can't function without their help. So, thank you to Patrick uh for being a sponsor and for being here today. Um, what do you have for us? Uh, yeah. So, I I had some ideas that I that I wanted to share. Um, but I thought it would be useful to demonstrate um to anybody that uh has potentially been living under a rock for a while. Um, just how powerful all of these things are. So, uh, let me see. Let me share my screen here. How do [00:01:32] I do that? I've never done this. Whereby? Oh, here we go. Got it. >> Um, continue with screen share. >> Okay, there we go. >> So, um, so yeah. So, if you're Okay, we can see this. Excellent. So, um, for, uh, again, just a quick little catch up, um, for anybody who hasn't been been paying attention to these things, this entire 14 slide slide deck, um, was one single prompt inside of Claude Design. Um, so if you've heard of Enthropic and Claude, um, you'll you'll know that, um, that these things are getting pretty good. It's actually in my opinion design [00:02:16] is the one area that's that's by far weakest um you know computer use capability that all of these AI systems have now. They've gotten incredibly good at code of you know um almost every programming language. Uh and and so I literally Hey, I've got to give a presentation in 30 minutes. Um make me a slide deck. Grab my logo from gridpane.com. um you know I want to talk about uh benchmarks on you know on artificial analysis uh want to talk you know about these different topics about hallucinations um this guy you know I I literally said Jeffrey Jeffrey Emanuel um has some [00:03:01] pretty good stuff he had something about what's the most accreative thing that you can add and it literally went and found his exact post about this and made this whole slide Right. And so um uh so we can run through these things but but the long and short of it is is that um you can get an enormous amount of stuff done very very um quickly uh with you know any any one of these. Obviously I use Claude um Anthropic for for this presentation um but I have accounts with Google and OpenAI and Grock and I I use all of them. Um, I use all of these tools interchangeably. Um, and so certainly I can talk about what my what [00:03:46] my favorite different use case is. Um, but for just a bit of background, um, and I I think I texted you about this, Matt. Um, I actually I graduated from high school in 1998 and I went into computer science to study artificial intelligence. Um, I just got there 25 years too early. Yeah. like none of this stuff was possible. None of this stuff even remotely existed um when I when I went into school to actually start studying this stuff. Um and it's really only gotten interesting um literally in the last 2 years. um you know everybody's you know everybody chat tpt um and that's they started get you know [00:04:30] pretty popular I guess in 22 23 um but it was actually late 2024 um December I I remember the date exactly um that uh openai released 01 pro that was the first uh open or that was the first uh reasoning model AI reasoning model um that ever existed. And crucially what that unlocked for all of us is before that moment, before 01 Pro, you could hand an AI a question and say, "Hey, what's the answer to this?" And it would just make stuff up, right? And and that's literally what it's [00:05:12] doing. Like like the way all of this stuff works is um is just a really complex uh uh linear algebra equation. um and it's just guessing what to say next. Um but before O1 Pro, um it would just guess and it wouldn't actually look at its answer, right? And so it would just puke out an answer and it wouldn't actually look at that answer and go, "Wait, does that even make sense? Is that even remotely accurate?" Um and so something that my I've got I've got three daughters. my my middle kiddo, we would do we would hand it these they're known as constraint satisfaction problems. And so the most famous constraint satisfaction problem is the [00:05:55] Einstein riddle or the zebra riddle which um is basically it's like you know Glenn doesn't live in the blue house and he smokes dunhills but he doesn't like dogs right and there's all these constraints and the the punchline of the zebra riddle is is who owns the zebra and so basically you you all of these data points of this guy smokes this brand he drinks this thing he lives is two houses down from this guy. And so you have the colors of the houses, the the animals, the drinks, the cigarettes that they smoke. And you have to figure out where all of these things exist. And then you have to that you have to figure out who owns the zebra, right? And so we would we would take we would hand these [00:06:39] constraint satisfaction problems to the AI models and they would just out the most ridiculous nonsense answers ever. Um, and and it would just be like, you know, literally the very first constraint would be, you know, Matt doesn't live in Michigan. And it would go, oh, okay, great. Matt lives in Michigan. And it's like, no, it we literally said Matt doesn't live in Michigan. It's like, oh, you're right. Yep, he lives in Michigan. Because it wouldn't check its own answer, right? And so, starting with 01 Pro, it would act said and revise what it said and keep going back to the drawing board over and over and over again. And so I remember it was December 23rd um and my you know this new model comes [00:07:24] out and it was like oh okay and I'm up in my daughter Desessa's room and I take a picture of the which is the constraints and then I take a picture of the second page which is like the the answer key so that you can plug in you know who goes where and and I hit enter and seven minutes later got the answer right and So it for the very first time it was like oh like we've got something now right we've got something we've got something real on our hands and and obviously things have advanced very very quickly from then um to the degree now where you can literally just say hey make me a slide you know make me a slide deck and it goes and it grabs my logo. I said that I [00:08:08] wanted to talk about, you know, the benchmark trail. Um, which, uh, if if any of you have been following any of these things, this, um, this this is definitely the go-to, uh, site, um, that that you can check out. It's artificial analysis.ai. Um, this is a site that I go to, uh, probably five times a week easily, just to see if there's been any movement basically in this chart. And so what these guys are doing in artificial analysis is they're just building more ever more complex um benchmarks or how we can actually test how good these models are relative to what we want them to accomplish. And so you can see this [00:08:52] chart, this chart is literally less than 24 hours old. Um you've got quad opus 5 at 63. And so basically what they're saying is is that you know based upon what they think is important the the state-of-the-art of the best AI right now is roughly 63% of the way to AGI right like artificial general intelligence. Um I think depending upon which metric um you care about um we're already that um certainly if you're looking at programming like you know making software um every single model that exists right now from any lab and literally models that you could run on your own computer in your own house um [00:09:37] they are better at programming than any human alive. Um and and you can look at people like minus Toralds um who created Linux um and he's like you know is AI better than every programmer I know maybe not but anybody that says that this is not useful is completely full of right they are fooling themselves and so the reason that I added this particular slide it's kind of hard to to see but the the highlighted bullet point here is um Quinn 3.8 27 billion parameters scoring 52. That's the orange line right here in the middle. Um, this is a model that for roughly call it less than $5,000 [00:10:22] in in hardware um you can run this model in your own home. Okay. And uh and it's right there neck andneck with GPT 5.6 six Luna and obviously, you know, Opus and Fable are better than this model, but you need a whole data center to be able to run those models, right? You need millions and millions of dollars to be able to run anything on the far left hand side. And this model that just came out a couple of days ago, right here in the middle, you can run for 5,000 bucks in your own home. Okay. And what's not actually shown on this slide because it wouldn't it wouldn't show up um is 01 Pro would be somewhere over here in the [00:11:04] 20s. Okay. And so that model which was literally the watershed light bulb moment for me um 2 years ago it wouldn't even show on this chart at all. Yeah. Um and you can now run a model that's materially better than that on consumer hardware. Right. And so this is something that I think that certainly every business should know about. um that not only is there enormous capability out there, but you don't have to give your data to Google. You don't have to give your data to OpenAI. You don't have to give your data to Anthropic. Um and you shouldn't. Yeah. Like this is another thing that people aren't quite figuring out is that um we are the product at this point just like with Facebook, just like with anything [00:11:47] else. Um arguably >> same cycle, isn't it? >> Exactly. like like I shouldn't be paying $200 a seat for all of my employees. They should be paying me $200. They should be paying me $2,000 a seat because I have very intelligent people who know their their knowledge their their um their space very very well. Um and every prompt, every iteration of things that we feed into the models, it makes these models better, right? Um and and this is true across every industry. My wife's a physician. There's a company called Open Evidence, I think, and their entire business model is it's free for the doctors to use, and there's advertisers that pay to to show up, but [00:12:31] really, you know, my wife should be paid thousands of dollars a month for just the literal questions that she's asking, you know, because she she has what what I consider to be the most important uh intelligence, which is actual intelligence, right? Everybody's hyped out of their minds about AI and it's like at the end of the day I can tell you for a fact no matter how good these models are um they are nowhere near as good as having access to actual intelligence. Um to be able to talk to have actually built things, people that understand things, people that are scientists, do whatever you name it. Um that's the [00:13:15] that's what you really want to focus on is getting yourself around um actual intelligence if you can. You know, if you're lucky um you'll you'll be able to connect with people who who deeply understand a space um because they're going to be the ones that actually um really create the next unlocks, right? And the most intelligent people that I know are using AI and every single one of them is using it slightly differently. Um they're they're down a different rabbit hole. And so, um, you want to surround yourself with as much of that as as possible. Um, there there's easy ways to do that. You can literally just go follow guys like Jeffrey Emanuel on X, right? So, um, I think that the the highest leverage [00:13:57] thing that I'm doing that's sort of me pissing away time is is literally just trying to find the people that are doing things on the cutting edge in whatever it is that's interesting to me. Um, and just follow them and see what they're doing, right? And so this one single prompt, what's the single smartest and most radically innovative and a creative and useful and compelling addition you can make to add to the plan at this point. This is a really good prompt, right? This is just one little box that you can check to add a little bit more flavor to anything that you're working on, right? Um, and so I highly recommend that. Uh, specifically to WordPress, this is this is kind of scary. So 43% of the internet runs on WordPress. um very very little of all of the code [00:14:44] that makes up WordPress is actually part of the training corpus that's in all of these large language models. Um and if you want I can tell you how I know this to be a fact. Um but it turn but it turns out that that um if you go to Claude or you go to any one of these models and you say hey make me a website. you go to you go to replet or um lovable and you say make me a website 0% of the time they will start with WordPress. They do not start with WordPress. They start with Nex.js. They start with React. They start with they start with things that are not PHP. Yeah. Um and and this is sort of good, sort of bad. Like um I think I mean [00:15:28] actually what what's you guys' sense on on this? like it you you probably talk to a lot more WordPress product owners than me. Um how how deep down the rabbit hole would you say a lot of them have [clears throat] gone in this context? >> That's a good one. You know, I I have never thought of it that way, but I also didn't assume that it would use WordPress unless I specifically asked. Um you know, I maybe I don't know, Matt, what do you what you think? But maybe it's it's the dependencies of WordPress. um you know if you're using Nex.js JS or something like that, you're just kind of using that one package, right? And and you're not >> Yeah, I I'd assume the same thing. the [00:16:10] dependencies if these things are just learning by pattern matching and it's just so much easier for that for the system to learn you know TypeScript in a single file versus everything that WordPress needs you know to manage it um back front end backend um and a database layer then I assume the training is just like nah we're not going to use all that stuff give us something that's just way easier and it just went the easier path >> yeah yeah so I think another part of it. Um, and I actually literally just saw this last night. So, so I'm using factory.ai. Um, factory droid. They have a they have a function called missions. Um, which I can actually check it right now. I have [00:16:54] a mission that's been running for 34 hours and 10 minutes continuously. Um, and I mean I I've spent probably $200 in tokens in the last 36 hours on that one project. Yeah. Um, so it's very impressive. It's very interesting. But I made a mistake last night. Uh, and I switched the primary uh orchestrator model. Um, so basically the way Factory Droid is doing all its stuff is you have an orchestrator that's queuing off all these workers to go and do hunt down little little things. Yeah. Well, I switched the the orchestrator to Kimmy K3, which is the current most powerful open-source um large language model. Uh [00:17:39] is created by I think the company's called Moonshot. It's a it's a Chinese company. Um uh if you if you haven't been paying attention to this, there's there's like four or five Chinese companies that are all of them at the bleeding edge of um most impressive large language models and they're all open source. Um and so you can literally just download Kimmy K3, you know, you just need $100,000 worth of hardware to even make it work at all, right? Um and even then it's very very slow. But I switched it to that model and then I went and you know and I came back and it thinking and cranking and cranking for like 3 hours and I'm just tracing [00:18:22] through all of the thought like the the reasoning you know of yeah and then I'm going to do this and then I'm going to do this and then I'm going to do this and then I'm going to do this and then it's going but wait I need to think about I need to think more deeply about this and then I'll and then I'll write the plan right And so I just kept seeing thousands of versions of and then I'll write the plan and then I'll write the plan and it just kept second guessing itself and just chewing through all my money, right? And um and so part of part of what it what I realized is that it's just like it's train what do we people's shared public thoughts, right? And what's in all of that [00:19:06] training? It's a lot of people secondguessing themselves. It's just an enormous amount of people kind of playing house, right? And um and I think that that uh it's clearly in in these models, right? And it's like so so one of the things that um that I think is uh really scary that's still true. So you can get artificial analysis also. This is the artificial analysis omniscience hallucination rate. Okay. And so these are all of the most impressive models on the face of the planet. And this is the rate at which they hallucinate an answer. Okay? And you'll notice this [00:19:49] all these black bars over here. These are all of the OpenAI uh frontier models right now. GPT 5.6 uh soul is right here at 92%. Okay, this is the most impressive model arguably of the front runners of the Coke and Pepsi of Anthropic and and Open AAI. This is their most impressive model. And if you tell it, if you don't know the answer, do not answer. 92% of the time it'll go, okay, great. Here's the answer. And it'll just spit out a guess, right? Like it knows that it's not supposed to do that and it still does it anyway. And the reason it [00:20:31] does that is because it wants to be helpful and because it's told you're a helpful helper, right? So, um, like it it wasn't told to play Jeopardy, right? It wasn't told you're gambling with your with with your own money here, right? So, don't guess. Don't ring in and say the answer is 42 if you don't know that the answer is 42, right? And it's like the best one are still Yeah. Command A is 14%. Right. And um let's see there's quad. There's that model that I was discussing earlier. Quinn 3.8 27 billion. 30% of the time it'll still guess incorrectly [00:21:16] even when you tell it not to. Yeah. And so this is why you have these things because um a non-trivial amount of the time it's just making up, right? And and even when you tell it like, "Hey, don't guess. Make sure you know the answer." It'll go, "Yep, great." Uh 30% of the time I'm still just going to confidently make things up, right? And unfortunately, this is this is not just an AI problem. This is a human problem as well, [laughter] right? Like we all forget exactly how things go. We all, you know, we all are like, "Yeah, I think it's that, right? I we confidently guess at stuff. I read this somewhere." And it's like, "Well, [00:21:57] what?" Yeah. Okay. And then what? Right. And so, uh, I think that there's just this very, uh, very scary reality that we're still forced with, um, forced to deal with, which is at the end of the day, you still want to have somebody to point your finger at and say you can't just blindly, you know, turn your life over to these things. Yeah. Um and so I think that having a very strong foundation in understanding just even the fundamentals of computer science um this is this is kind of required right in order to um in order to not shoot your foot off. Um and [00:22:41] so I think kind of back to the original like why does it not pick WordPress? It's like because there's so many question marks around what WordPress is, right? And and there's an enormous amount of the internet that says WordPress is dumb, right? Like don't WordPress is old. Like like I've talked to I've talked to people at companies that they're you know they've raised tens of mill tens of millions of dollars and they're working on some cool database thing and I'll tell them, "Oh yeah, well, you know, at Grid Pain, we're doing XYZ and and you know, we help we've got 150,000 sites on WordPress." And they're like, "Wait, WordPress is still a thing?" like I I I didn't realize that it was still a [00:23:23] thing, right? And it's like these are intelligent people who are are building software and they have no idea that 43% of the internet, let's say it's diminished and it's gone down to 40% of the internet. They, you know, there's a lot of people out there that like, no, WordPress is old. It's dead. Nobody does that. Nobody does that anymore. Yeah. Um and it's like, no, there's a lot of people that are that are still doing this, you know. Um, so yeah, I don't I don't know. Let's see. Uh, what are some of the other slides? Yeah, so this is kind of funny. Um, there was a point last year, it was the end of last year where it's really started to get impressive for me. Um, where I messaged my co-founder and I [00:24:09] said, "Hey, um, you know that problem that we have?" Well, I just I just prompted I think it was Sonic 3.7. Um, it was a new anthropic model at the time and I I had something where I needed like a very specific calculator effectively like a fancy calculator and um and I prompted it and it got it exactly right the very first time and I said to my co-founder I don't know whether to save the prompt or the code or neither right because I knew that the models were going to get so much better from there that it's like what's the point [laughter] right like what's the point in saving the prompt because future versions of all of this stuff, it's [00:24:51] like, will you even have to say the prompt out loud, right? Or will you just have a brain computer that's just going, "Well, Patrick's interested in that, and we'll go and calculate what it is that Patrick's going to want to talk about an hour from now, right?" And it's like, I get that an idea like that seems like a sci-fi future that won't exist. And it's like, they're already doing this, right? They're um what's the what's the company? Elon's 16th company or whatever. Um >> yeah. Yeah. [laughter] >> Um what is it? >> Near. >> I want to call it mind eraser, but it's not mind eraser. >> Yeah. Near where you can say,000 and have a computer implanted in my [00:25:34] brain and some people will say yes and some people will say no. Right. >> Yeah. So, I mean, it's interesting when you look at some of these statistics. I mean, that directly flows into what we're doing when we're building stuff, right? I mean, I I'm currently working on a plugin for myself that I've been using Claude for and and you know, I one of my points about it was that I can get a working prototype in 5 minutes, but to actually make it usable, to make it uh as detailoriented as it needs to be for other people to want to use it, you know, it's going to take time. And you're going to have to [00:26:17] to um also understand that the methods that these models use may not always be the best method. Um there are certain things like even with WordPress you can tell it you know well here are the security guidelines >> use those guidelines and sometimes it'll use them sometimes it won't. How how have you kind of navigated that and just sort of verifying what the model's doing and so that it's something that you know you can safely use and and maybe have other people use. >> Yeah. So um so it's very good question. So the the short answer to that is um the or the the answer that's most useful [00:27:00] to everyone I would say is you should have two or three or four paid accounts with all the different providers all the different models. Yeah. And what you want to be doing is you want to be based red teaming whatever code you create you want you know let's say you build it in anthropic in claude you build version one inside of claude you then would want to have open AAI 5.6 GPT 5.6 Saul basically sanity checking that code right and then you could have gro 4.6 six also, you know, um red teaming that code and basically creating iterations um and refining and refining and refining. Um [00:27:46] but then you know unfortunately the um the the sad fact at the end of the day is it still just comes back to um you need to have a human being in the loop that actually knows whether or not this is right? Whether this is brilliance or not. Yeah. Um and and um so a hyper specific example of this is I've been working on something for about the last 18 months that's way off in the weeds um that I I don't actually care sharing it. Yeah. So, so if you are a hosting company for example and you have [00:28:29] really big WordPress websites that have millions of orders um in the database and you need to move that website from point A to point B, one of the things that you might end up doing is um a search and replace in the database to change the URL. Okay. >> Very straightforward. Yeah. Um most people just use a plugin. Some people use WPCLI. Yeah. Um and you know some some people just use the tool right within their hosting platform, right? Um but what I've been working on uh again for about 18 months is I just keep taking the same work that I started after 01 Pro. It was [00:29:15] actually 01 Pro was the first one that was able to advance this idea for me. um that version of that code that I built in December of 2024, every new model release, I have it test that code and make it better and make it better and make it better. And about 3 months ago, I finally handed it off to my CTO and he then made it in order of magnitude better than what I had made because he had ideas around what we could do with with making it better, right? And and ultimately like I had gotten the code so good that it was like I I was convinced there was no no more room to improve it, right? And I had taken something that in a very large [00:29:59] database a search and replace could take 24 hours, literally 24 hours and you don't even know if it's working, right? Like you could be 23 hours and 59 minutes in and not know how far along you are, right? Not know that you're even that you're that close to the diamond in the mine, right? And we took this 24-hour task and um and I took it down to about an hour and then he took it down to 10 minutes and then I had an idea that took it to 10 seconds. And so now we've taken this work that's real like it's a real problem. You can't if you go to any host anywhere they'll go, "Oh I'll see you in 24 hours." That's how long that's going to take. [00:30:42] And now we've compressed it down into literal seconds like real time. And so I guess the the long long story long answer to your question is um for stuff that's really important at the end of the day, you need to have an actual intelligent tool, intelligent human being in the loop. Um and then you still need three different versions of the AI, right? And so it's like the way that my co-founder and I work together now is thankfully for us this um he's he lives in China. So he's always 12 hours off from me. He's off 13 hours when we're doing daylight savings time cuz they don't do daylight savings time. Yeah. So every day um when I'm getting ready to [00:31:27] go to bed, he's getting up and vice versa. Right. And so I'll hand him what it is that me and my clankers have been working on that day. And then he and his clankers work on it and they hand it and we hand it back and forth, right? Um and so we're using at minimum always two different models that are sort of sanity checking each other. And then we're we're still having to jump in though and go, "Wait, stop Kimmy K3. Stop secondguessing yourself. Stop doom scrolling." Right? Stop going, "And then I'll make the plan and then I'll make the plan and then I'll make the plan." Right. Um and um and so yeah, like the uh what's the what's the meme? It's like the astronauts pointing the gun at the other astronaut, right? And it's like [00:32:10] it's always been that way. Um, I think I think at the end of the day, the reason that I'm not concerned about AI taking my job is at the end of the day, the reason that a company will pay me a lot of money to host their WooCommerce site is because we will be the last astronaut at at the end of the line that has a gun pointed at his head. And it's like, yeah, we'll we will be responsible, right? If something breaks, we will go, okay, this is what broke and this is why. Now, we may be fully assisted and we are fully assisted by AI models, right? But it's like I have a head of engineering that lives in Dallas and I have a CTO that lives in Wooi, China, and I have another guy in another place [00:32:54] and you know, and I have access to other people where I can go, yeah, but why? And then understand it and explain and and make our own internal models better, right? Um, and so there is no there is no um perfect answer and and and it was a real light bulb for me last night thinking about like um at the end of the day nobody is going out on the internet like I think that people are going out on the internet and writing what they what they feel right and they're and they're secondguessing themselves publicly but I think that there's a lot of people who run podcasts or run companies or build build things or write books or get on [00:33:39] stage and perform in front of people where they're petrified to get on that stage and and if you if they had to write their thoughts out publicly, they would go, I don't know what I'm going to do, right? Like I I might screw up. I don't know if I can hit that high note, right? But they go on stage anyway, right? And and there's not a lot of training data out there for that, right? There's not a lot of people who who who are writing about like, "Yeah, I was terrified every minute of every day, but I still just ate another bowl of glass, right? I still just decided to press record. I still decided to get on this podcast with this totally thrown together slide [00:34:23] deck that I had no earthly reason to ever present, right? and and who knows if anyone will ever see it and if anybody cares and it's like but yeah we're still doing it anyway, right? >> Absolutely. >> Well, I think it's interesting you talk about the human factor. I mean because that that is still important, right? I mean you we need to have humans still guiding the ship. I mean it's like to me it's kind of like being the general of an army or something. you still have to have somebody that knows what they're doing at the top to to make sure things go in the right direction and to kind of uh have that critical thinking, right? Because AI is not at a point where it's going to do it for us. And especially if [00:35:06] we're making products, if we're expecting other people to use what we're building, you know, there has to be some level of expertise and a a conscience about it. >> Yeah, 100%. I mean it's like like the thing like there's a there's this cliche in in um in venture investing in particular in in VC backed businesses that you ideas are right execution is everything I are execution is everything and um and every venture capitalist everywhere has has muttered that phrase right and it's like well now you can farm a lot of execution off like an enormous amount of execution And so you could make the argument that [00:35:51] now ideas are really important, right? And it's like we could build the the most amazing combination Bose wave radio uh freestanding urinal umbrella, right? We could make that. But who would want it, right? And like why? Like who would actually say like yeah I bought the combination BoseWave radio you know catheter thing [laughter] it's like no no one wants that right and so you still have to you still have to understand like what keeps people awake at night basically like I think that that's what great products actually solve you know like making the software [00:36:34] making the boxes appear and the boxes get checked making those things is easier now than it ever was but um like that thing that I described that database um search replace tool. It's only important to like maybe 500 people on the planet. Like it really like the reason that it doesn't matter to 99% of people that are building WordPress websites is because their database is so small that it doesn't matter, right? You can do the search and replace through WPCLI and it'll be done in 2 minutes, right? And so like who cares if it's if it, you know, if it gets done in a in a minute versus 3 seconds, who's signing up? Nobody. Nobody cares, right? But [00:37:19] it's like if you're building and managing and maintaining WooCommerce websites that have billions of dollars transactions and history in them, not only does it matter, but it's really important that it's correct, right? It's not good enough that it's just super fast. it's verifiably correct and and we actually literally found ourselves in a place with that specific tool that none of the other tools that exist we um none of them do it well enough for us to benchmark against it. So like if you use WPCLI does it wrong in like all these weird edge case ways. So like if you if you [00:38:02] compared what our output was versus the right one WPCLI 24 hours later WPCLI would have it wrong and then you'd go well you didn't do it right and it's like no actually actually WPCLI does it wrong you know in this very specific way and it misses when you have escaped URLs encased with you know you know serialized strings encased within encased within encased within you know and so um so I mean I think at the again at the end of the day. It just really comes down to who who do you serve, right? What do they need and do are they willing to pay, right? Like are they willing to actually part with their money in order to get the thing that [00:38:45] that you're going to give them, you know? And I think all the fear around um oh AI is going to take my job or like you know we see these collapses of um you know utility plugins in particular utility plugins. The demand for utility plugins is going down. It's like, well, yes, because five years ago, there were literally hundreds of thousands of people that they didn't know how to solve the thing that they wanted. And so, they went and they looked in the plug-in directory and they found one little thing where it's like, oh, if I check this box, then it'll do this thing, right? And it's like, so now if you are the, you know, nth edge case on on an add-on on an add-on for gravity [00:39:28] forms, well, I don't need that anymore cuz I can just tell Claude, "Hey, can you make my Gravity Forms do that thing?" But things like Gravity Forms are still going to exist and they're still going to be very important because I'm not going to go, "Wait, what are we what are we doing? We we're we're gambling our whole business on something that somebody that I've never met vibe coded in 5 minutes versus somebody who built a company to millions or tens of millions in revenue and they have hundreds of thousands or millions of customers all across the globe. And all of those people every single day are going, "Hey, wait, that's broken." And then somebody picks it up and fixes it and makes it better, right? And provides support. Um, yeah, I'm not going to [00:40:12] gamble my business on on question marks on the on the on, you know, this this chart right here, which is, you know, even if I'm even if I'm doing everything right, you're saying that there's a 20 or 30% chance that it's completely wrong and you told me to to bet my life on it anyway, you know. Um, and then I'm sure you guys have seen this, like you take it to the next level, right? where what's what's the fundamental what's the biggest reason that one back and forth between it's no it's that it's that there right here is cuz that plugin over there [00:40:55] is doing it wrong [laughter] right and it's like eventually somebody has to be that last astronaut and go okay it was it was my fault I'm sorry here's how I fixed it here's how we're going to make it so that this doesn't happen again in the future, right? And the closer I think you can get to being that for whoever you serve, that um that will always have value. I think like I can't at the end of the day, I can't see somebody um saying, "No, I'll just I'll just gamble it all on AI. I'll just gamble every aspect of my life, my health, my wealth, my business, my sanity. I'll just gamble it on on this chart right here. Like, no. You're gonna [00:41:41] you're going to want people that are using these tools, but they're also going, "Yeah, this is our responsibility. Like, we will take responsibility for for this thing." Um, and so I think that I think that while it it will potentially get harder for people to make a narrow plugin that that has value in that way, I think it's I think it's a lot easier to to think of yourself set of a particular group of people, right? and and then you know just lean into that heavily where where it's like oh I serve golf course superintendent right this seems like a narrow segment [00:42:24] this seems like this couldn't possibly be a big business I can name two different businesses that are doing tens of millions of dollars in revenue annually and all they do is literally just serve golf course superintendent right like I can't name a single golf course superintendent but there's tens of thousands of them in the United States, there's tens of thousands of them in Europe, in Australia, right? And it's like that narrow tiny little segment that nobody thinks about is is a multiund million or billion dollar vertical. >> Matt, did you have a something you wanted to say? Yeah, on the um and we'll you we'll wrap this up uh and then we'll just get any [00:43:07] questions for folks that are in the audience, but um the thought of the AI training coming from people second-guessing themselves. I I'm wondering also because I had always thought about it like oh when it's training it's training it from the technical level. I hadn't I mean I knew that it was but I hadn't really thought about it in the fact that people could be criticizing WordPress. Therefore, the AI model is saying we'll never use WordPress because people are either criticizing or second-guessing it versus just like what you and I said, Eric, not to throw you into the same thought process as me, but just from like a strict dependency issue, [00:43:50] >> um, of just like, okay, I just need this code versus I just need this code and by the way, thousands of people have said WordPress sucks. So therefore therefore the model says ah we'll probably not I'll probably not use this >> but curious h how has and with that in mind like how has that played out for for grid pain or your take on people wanting to use WordPress these days? You know people people are just never happy. This is just like an internet thing. People are just never happy with like the progress of of WordPress no matter what happens. >> I [clears throat] see so little positive [00:44:34] sentiment around WordPress and I don't know if I'm just living in my own bubble because I am a WordPress podcast publisher, you know, but the Studio 2.0 thing, I just put a video out the other day about Studio 2.0 0 and you can now build data agentically and it's it's like having cloud like a cloud desktop but for WordPress. It's how I see it. I posted it in dynamic WordPress Facebook group and people are like this still sucks. This is still slow like this. It's like what what is anybody ever happy about anything that comes out and knowing that this is like beta version and it's going to get better just like the models on this chart used to. We used to hate the words that Chad [00:45:19] GBT would create and the images that Chad GBT would create. And now people are just like, "Yeah, just give it to me. Let's go. It's fine." Like now it's good now. Don't even think about it anyway. Prompt engineering. We don't even think about it anymore. [laughter] We just go and take it. So anyway, >> correct. The point the question I have is like how have you seen the this fight against WordPress affect either your business or your outlook on on WordPress as a whole? >> Yeah. So, so, uh, I guess the the short version I would say and sort of going back to like why would the AI not pick this, right? From a pure like just mathematical, you know, strictness [00:46:04] computer science standpoint. Um, WordPress is not a good idea for a lot of things. Okay? Like like when when you if you have a lemonade website, right? You have a lemonade stand website. It's a mom and pop brochure site. It changes almost never, right? WordPress is not the right answer. It just isn't. Okay. And so, um, I do think that there is going to be it's we're already seeing it. It is going to be easy, um, for people to choose to not use WordPress anymore, right? But every single day I also see scenarios where somebody vibe codes the lemonade stand [00:46:48] website or the quilting website for Nana, right? And then Nana comes back with 18 questions about, well, how do I do this and how do I do this and how do I do this? And then they go, "Well, I should probably actually vibe code this back into WordPress because then I can point her to YouTube and there's 18 million videos on YouTube. how to make the right and so um certainly there are people who would say you know WordPress sucks yeah doesn't matter what you push out it's not fast enough it's not good enough it's not agentic enough like everybody everybody has their complaints right and I was actually talking to Matt Cromwell about an hour ago and he said [00:47:30] hey to you Matt and then also he wanted me to make sure that you knew that you he said hey make sure that Matt knows that sucks, but he said hi. Um, and we were talking about just like everybody has this negative vibe, right? And it's like I think what people need to do is just surround themselves with people who are actually not saying those things, not doing those things, right? It's like if when I talk to the people that I know that have successful businesses, they're very very busy, right? And um and they're in WordPress, right? And it's working. And I can assure you there are literally millions of dollars that are going to be transacted today on websites that are on [00:48:14] WordPress and Woo Commerce and they're running on grid pain and they're happily paying me more than they paid me last month. Right. Because number go up and to the right. Yeah. But I think that I think that the the majority of people on the internet commenting on the internet think about what that person is doing, right? They're not they don't run a podcast. They don't own a business. They don't have a team of employees. They don't they don't have um the same inputs and outputs that that we do, right? Um, and so it's very very easy to just, you know, sit in a sit in a glass house and [00:48:56] chuck rock. Um, out there and get punched in the teeth and pick yourself up and keep going, right? And and nobody has time that's really busy and is winning to run over to X and go, "Hey, by the way, I'm winning." Right? And and even if you did, nobody wants to hear that. Nobody wants >> That's the level of business I want to get to though, Patrick. I just want to sit on Twitter all day and be like, I'm winning here. I'm winning there. >> Right. Right. No, but but the the thing about it, it's like nobody would want to hear, >> you know, Kobe or or or or Michael Jordan going, you [00:49:41] know, he's dunking on everybody and then runs over to, you know, whatever it would have been then, MySpace. He runs over to MySpace and goes, "Hey, I'm dunking on everybody." It's like he's too busy dunking on everybody, you know, and and and and it's bad enough that he's dunking on everybody. Nobody wants to hear him rub it in, right? No one wants their faces rubbed in it. And so, um, I think that that's why I think that like in-person events are going to become more important, not less important. You know, like I think it's I think it's going to be more important that you're putting yourself in the right room with the right people talking [snorts] about the things that are working, right? Um, and because there is just so much negative sentiment [00:50:25] everywhere, right? There's there's, you know, when you look at the 8020 of anything, there's 8020 of 8020 of 8020, right? There's like one small fraction of a percent of people that's doing exceedingly well and nobody wants to hear what they have to say because everybody else is struggling and sad and it's not working and you know, and so um so I think it's just get in better rooms, right? Get in better rooms with better people and talk about >> cooler I guess. Stay out stay out of Facebook groups is what I'm hearing. >> Well, right, right. Like I mean certainly certainly you could, you know, there's a version of me that, you know, uh, Word Camp 2024 happens and the [00:51:09] keynote pops off the way that it pops off and then, you know, just downward spiral of bad from there. And it's like, no, I myself and a lot of other business owners in around the WordPress space, we kept our heads down and we just kept working and we kept working on what we knew we could we could do and we kept delivering value to our customers, right? And it's like I think if you can do those things, you're always going to be able to find success, you know, in in anything you set anything that you really commit yourself to and you just don't give up. if if there is gold to be had there. Again, maybe the BoseWave radio, you know, umbrella um vibrator is not a [00:51:54] great idea. >> Let's go back to urinal, please. [laughter] >> Yeah, whatever it is. Yeah. Um so, so yeah, I mean, I think that I think that's um if if we just look at the regular discourse that all of the models are trained on, there's a lot of negativity. There's a there's not a lot of positive things being said about about WordPress, right? There's not a lot of positive things being said about anything, right? >> And so just pay less attention to that stuff, I guess. Listen to more podcasts like this one. >> Yeah. [laughter] >> Well, well, let me wrap up the uh recorded portion of it on this. I will say that just because there are loud voices out there do not mean they are [00:52:37] correct voices. And that's something I think we all have to learn because we we pay so much attention to these people that just blasts out negativity at every turn and sometimes they're doing it for profit. So you have to remember that in all aspects of life. Uh but Patrick, I thank you for coming to uh join us and um with the well-prepared slides. Um [laughter] how can people connect with you? Uh, so I'm in terms of uh the easiest place to get me is probably on X. Um, so just at Gridpane. Um, you can also go to gridpane.com and there's there's my fat head in the bottom right. There's this [00:53:20] thing called video ask that you can click on and um I don't see all of those but pretty much every single one where somebody actually submits a video message of themselves as a real human being saying hey this is my problem you know or hey this is my question. Um I I'm pretty sure I answer 95% of those or I see 95% of those and respond to them just like this. And so, you know, if you have a question about WordPress, particularly anything on Woo Commerce, that's where we we tend to have the biggest impact. But if you have questions about agency owners that are winning, that are succeeding, happy to talk about that. Happy to talk about uh parenting, golf, not politics. [00:54:06] [laughter] Uh but, you know, uh yeah, that's that's the best way to get me. reach me on X or just um send a video message and um happy to connect and help if if I can. >> Awesome. Well, uh thanks to you. Thanks to everyone for uh checking out our member meetup this month. Visit us over at thewpmin.com/subscribe. Grab our newsletter. Support the work we do here at the WP Minute. Okay, we're going to turn off recording now and if anyone has any other questions, we'll go to that., Lessons Learned From Vibe Coding Plugins, WP Minute Podcasts, 01m3pspbgv9x3hj3avyyf0a8tr.