Chris Heilmann === [00:00:00] Hi there and welcome to PodRocket, a web development podcast brought to you by LogRocket. LogRocket helps software teams improve user experience with session replay, error tracking, and product analytics. Try it for free at logrocket. com today. My name is Paul and joined with us is Chris Heilman. Chris is the Vice President of Developer Relations over at WeAreDevelopers. In addition to that, he's an author. He's a speaker. blogging online and he's here to talk about his upcoming talk. That's going to happen at CityJS misplaced. Good AI solutions need good UX. Welcome to the show, Chris. We're excited to have you. Hello i'm excited too, this is good fun. Yeah. This is a very timely thing we're going to be talking about. Cause right before we hopped in here, Chris said there's a lot of rubbish out there. There's a lot going on and it would be good to index what's going on, talk about and figure out like the most best material ways we can move forward. Chris, you probably know, this is [00:01:00] a. Kind of a web development centric podcast, and we have a lot of people that are in that space. So maybe we can hop off into your introduction by talking about how you got into web development and a little overview of what you've been doing in the past few years. okay. Oh dear. I got into web development in 1996 I basically started as a radio journalist and then I realized and I did Computer games as a hobby on 8 bit machines and stuff And then I basically I did the journalism thing or radio and then I realized radio is dead. It's still around now, but it was really dead back then. And then the internet came around and I was like, this is cool. Anybody can become a publisher just by doing something and putting a few random words together and actually just learning HTML, you can start becoming a publisher and people can access your stuff worldwide. So that's how I. Really got into it and I really got excited about it. And as I was that early, I got a lot of companies starting to get attention from on me and say okay, cool. This guy knows HTML, so let's [00:02:00] hire him to do websites for us. Most of the clients back then didn't know what a website was. They just heard from somebody else that it was something. So I had to build a lot of weird stuff and to cut a long story short, I, then I did a lot of stuff for different Banks and I did a lot of things in like enterprise level content management systems back then Like home order television channels one of the first e commerce sites And then I moved to the u. s and worked on e toys, which was like amazon for toys from there I moved to england and then I started at yahoo built I worked on the yahoo front page built yahoo answers yahoo maps And then I went to mozilla and I worked on the browser because I really wanted to see What I could do as a developer to make the environment better, that we work it as well. And that was the browser in that case and worked on Firefox OS, which didn't work out as much as we wanted to. And from there on, I went to Microsoft and basically wanted to kill Internet Explorer because that was the bane of my existence as a web developer for years and years. [00:03:00] So I managed to do that. I've managed to roll out Microsoft Edge and the last eight years in Microsoft. Basically I worked on a developer tools, which were shared together with Chrome that basically are in the Chromium open source core. And I was the principal program manager there. And yeah, in the last few months, I now moved to to work in we are developers because I realized that the job market is right now also still broken. So I want to see what I can do about that. That actually developers don't read job specs that don't make any sense to them. And people don't write. CVs that nobody reads as well. So I think we're good in the platform space. I think the web platform is doing really well at the moment. I think the problem is that people think it's unnecessary to use the web platform, but just generate a lot of code. And I think that's a massive lie that we have to counteract right now. And that also was the inspiration for the talk that I'm about to give there. Like, how do we make sure we use AI. But we use it to to augment what [00:04:00] we do right now to empower developers to become more effective rather than replacing them, because I don't think it's that easy to replace us, but other people disagree, but I think it's also important to realize. This AI revolution that we have right now is a great mirror to see what kind of work should be done by humans and what kind of work can be generated by machines. So a lot of times as a web developer, you build a lot of stuff for the bin. You basically build websites that nobody needs for the moment. It's just like a prototype or it's something that. Somebody wants to prove something with, or that it's a startup that basically pivots every three months. So it doesn't really matter that much how good their web product is, as long as it brings users in. And I wonder if something like that could be done by machines. And it should also show that the craft of being a web developer. Is something that is much more important than just churning out lots and lots of code really quickly. Chris, do you feel like you're in a good position here to comment on this? [00:05:00] From me as being privileged enough to listen to you, I'm like, wow, this guy, he comes from the journalism side, the content side, and he's written every application under the moon. So do you feel like you're in a particularly good spot to see this from a bird's eye point of view , I think so, definitely, because I went through it a few times. I'd survived a dot com crash in 2000, and I survived a few other companies going bankrupt. I saw a lot of And I saw a lot of environments as well, going away. Like I remember when Photoshop was the one thing you could use. And if you don't use it, you're not professional. Nowadays you use Figma, use things in the browser. You don't need to install extra software. I remember when our skill was 90% knowing how browsers mess up and how to avoid that. And nowadays our skill is more like what kind of. Technology. Can we use for a different use case? I've I was very lucky to have worked in Yahoo back then, which is a lot of people underestimate Yahoo. It's always okay, it's like the [00:06:00] search engine that nobody uses, or it was just like back big website back then. But the development team in Yahoo was ridiculously good. Like the whole MPM, the whole notes thing started there. And if it was there at the beginning to see how that happened and also how to see how fast it grew. And how it grew without a thought was something that I think is very important to think about as well We are we right now rely on a lot of on an infrastructure That actually is a lot of packages from other people that we don't know how secure they are how they perform what they do but we basically We focused on developer convenience for far too long and now out of a sudden the developer convenience that is using lots of packages that we don't know is being automated by machines and automated by systems and people go crazy that we can't be replaced. You're like, well, if you work with other people's stuff all the time and you don't even know how you're putting it together, you just hope that it works. Works with bubblegum and duct tape, then you shouldn't be surprised that a machine can do that for you as well, because you become [00:07:00] predictable. I think the biggest thing as a developer to become to have a career and have an interesting career is to stay unpredictable, to be more human in your endeavors and more human, your approaches to things. what does being more human to you mean in this context? To me, it always meant I get very bored if I write code that doesn't have an interface that to humans to basically to to know how people use a product, to make sure it's just a button. Should that be a slider? Should that be a table that gets sorted? Building interfaces, UI together with code was always much more exciting to me than just optimizing a database lookup or optimizing a, an API that actually only talks to other systems that luckily enough has gone away. I remember working in financial products with like soap where you basically, you wrote XML specs. You called APIs with other XML specs, and then by hand you validated if everything went right. And you're like, why did we need humans for that? So the. [00:08:00] To me, writing code is much like writing texts as well. You can have 10 different solutions for the same problem, but you should have after a while, your own writing style. And I think it's very it's very arrogant for people to say there's one perfect syntax. I think. The I've encountered in my career developers that came in sideways into the market that were like architects, like physical architects or biologists. And they actually wrote code completely different to somebody who actually came from a background where they just learned it in uni or came from a math background. And I think the beauty of of our market is still that people can come from all directions. And I think the AI space even opened that up more. Data scientists, I know like 80% of them are not the normal developers that I would have expected. They come from completely different backgrounds and it's exciting that our market gets bigger that way because we need more people in our market, but we didn't we didn't educate enough people in the right way. So this is interesting to see. So [00:09:00] chat GPT must have been one of these catalysts in your eyes that's like helping extend this entry path from other domains. Like you can have somebody that hasn't touched Python code, they can use Code Interpreter now, write a little script to do something for them. I'm curious why you think if in addition to this chat GPT has skyrocketed through in all of our minds, it's commonplace in conversation across all Age gaps right now. What do you think brought it to that spot? There's two sides to that. I'm actually starting with that in my talk. It was a perfect storm of the interface showing the quality of what an LLM can do, a large language model can do, and what natural language processing means. I think they've done a really good job showing a prototype in ChatGPT, what an interface to a system like that could look like. It was much less academic than all the other things that you saw before. We had APIs before that, that you called, we had playgrounds on Amazon [00:10:00] and on Azure as well, where you can use these systems, but chat GPT was more like, Hey, here's a kind of Instagrammy Twittery way to talk to a computer and that always was a scientific dream we always had ever since Star Trek, where we piece the computer do this and the computer did it for us. And the interface was very much like. Gamified to a degree as well. The text didn't just show up, it wrote itself onto the screen and you had to ask the computer more and more to get more things in there and actually get more information out of it. So it was super exciting and super engaging. The danger of it is right now that because it was so engaging, it just got copied blatantly by other systems like Bart looks the same almost every startup that I see that does something like that gives us a chat system and the chat system is to me, I give it another two or three months and people will be bored with it because if it doesn't give you the right information, it's just super costly. It's first of all, super expensive [00:11:00] in terms of energy it provides, it needs to run. It's it. You normally, when you could do a Google search and find a code result in two seconds, you normally talk 10 minutes to chat GPT to find the information. And you don't realize that you spent 10 minutes on it because it's so engaging and it's so gamified to a degree. And I think that's the biggest danger that we have with that, that this kind of like a communication with the computer or with the machine is now a hot thing to use. And it's, we've been there before. I remember when like Siri came out and then when Cortana came out and. Bixby on Samsung. Some people use that one as well, and we all thought this would be the number one thing to talk to our computers and use our computers. We would wake up in the morning and say, Alexa, what's my schedule like? And in essence, we don't use them that way. We asked them for a 10 minute timer when the pasta is cooking. And that's about it. The luster or the magic of of Cortanas and series and these kind of things actually [00:12:00] gone a few months after they rolled out there. And it was interesting because I remember when Microsoft in their keynote said that every app will be an extension to Cortana in the future, or everything will be talking to your personal assistant and that Corona came. And out of a sudden we realized the it market is not ready to actually roll out a piece of software that actually shows numbers who has been vaccinated or not. So we tried to jump into that sci fi world of like ubiquitous computing and voice recognition. And then we got disappointed really quickly with it because it didn't do the things that we needed to do. And I think we will get the same with chat GPT when people start thinking, okay. I talk to this thing now and I get five lines of code. Now I have to copy and paste them somewhere else. And then I paste them in my development environment. And then I realized it's full of bugs because the chat GPT interface doesn't do any validation for me. It just tells me the things. The other danger that I see in it is that it's it's eloquent. It's really good. [00:13:00] It's an English, especially it's super eloquent. So I keep calling it like mansplaining as a service. These systems say the most. outrageously wrong things, but they say it in a very convincing manner. So you're just standing there and they're like, Oh, I feel stupid probably by thinking that's wrong. But no, it's just plain wrong because these machines make mistakes because they've been trained by humans as well. So, I think the next evolution of that will be get away from that chat GPT model and integrate AI into the environments that are specialist environments already. And this is where I think it becomes much more efficient to use that kind of functionality. I love that ChatGPT has broken open the doors to, to basically make everybody understand that machine learning and LLMs and these things are there. I'm worried that the success of ChatGPT as an app itself has blinded a lot of people to try to imitate the success of the app and not the success of the system that runs it.[00:14:00] So, gitHub does actually some really good job in educating people about the inner workings of Copilot and other things out there as well. And OpenAI has a few training courses out there as well. But a lot of it is still Oh, this is just cool magic. And I'm worried about Yeah, I'm worried about that. I'm worried about that. Like generative AI is thinking everybody's an expert by pressing a button like, Oh, I can make a great image by telling a mid journey, a good prompt. I can actually edit a video by using an editor that does automatically all the removal of stuff by me, but that doesn't mean, the craft, it just means you know how to use that system and that system can go away any second if there's not much money coming in it. And chat GPT and others bleed money at the moment. So that's something else we might worry about. Your comment about having these LLMs and other AI powers in domain specific areas, having that be one way for them to shine is definitely, I'll just remind our listeners right here that the [00:15:00] podcast is brought to you by LogRocket and LogRocket's definitely taking this approach a little bit. With their product, you can surface all sorts of patterns, errors. Interactions that you might not have seen in your console or in your node project and allow you to spend more time developing and less time like combing through those error logs, those console logs, or events that you might be missing in your DOM. So if you want to solve issues faster and, take advantage of this new technology in this type of specific domain, go over to logrocket. com today and you can try the product out for free. Chris, I want to zero in a little bit on this theme I feel like we're digging up here, which is shoehorning this interaction, the chat interaction into everything. Do you feel like that is the poster child about Why we might be limiting ourselves with AI and I love that you brought up co pilot too because co pilot. I recently got the ability the privilege to join their like co pilot X sort of [00:16:00] thing and it's more similar to chat GPT than my old co pilot the old co pilot I was like there's something going on in the background and it's you know Maybe reading this file of that file vectorizing it who knows but now it's just a chat interface And it feels awfully similar to the chat GPT. And I don't know, do I use chat GPT or do you use co pilot? I don't really know the difference now. So do you feel like this. gravitation towards chat as the phenotype is shoehorning us into less capabilities. Is that sort of the poster child of what you would fix? I think it's something that we need to do right now to get the people, if a certain feature becomes fashionable, then when you release a new product, you have to use that feature, even if it makes no sense, you have to get the people that way people would expect there to be to be a chat, if it says it's AI driven. The fun thing with Copilot, and I've been using it since the very beginning, and the Copilot X is just coming out right now, is that it's great that it [00:17:00] has this chat interface as well, but it's even better that it's inside my Visual Studio Code or my editor, there's five others that it supports. So I have the context of the code next to it. I don't have the full page taken over by it, and I don't get just thrown into the chat interface. Even cooler, and I'm going to show that in the talk as well, is that you can actually do a Command I. Inside your code and on the line that you are, you can open the chat prompt there and it actually only chats with you about the part of code that you highlighted rather than just going into the editor and having to copy and paste it back. And when you actually have a chat. with the system and you generated code that you wanted to have. It doesn't just throw the code in like any other editor would do, but it shows you the diff view as well. So you can even see the you can see that the changes that it made to the rest of the code. And if everything gets called and everything lines up, and then you can actually just say, okay, this is good enough, or you have to tweak it a bit. And then you actually put [00:18:00] it into your code. So that extra editorial step is something that a lot of people, a shortcut. Because it feels like it's slower and it feels like the magic of AI doesn't do everything for you. But that's the most important part to look things up and make sure that the code that you generate actually works. The other thing that it does, and it has a weird name for it. What was it called? Code referencing is also another one of CodePilot X that just came out, which if you write some code or you, and it does an auto completion for you. It actually and it finds that it's public code It actually shows you which other products use that public code and what license it is so you're not running into the problem if you work on a local government website or you work On a government initiative that basically you cannot just nilly willy use code of any license out there But in an audit it would actually break the whole product So, this case, it's looking up where it came from, I think is super important as well. So I like that command i the most. Basically, you get into your code and you bring the chat inside your [00:19:00] workflow. Rather than just going out of your workflow, going to the chat, find some information and getting it out there. One place where chat is really good is ideation. If you just want to start thinking about something or a solution, then you start chatting to the system and get that way out of the way. It's a bit like rubber ducky programming that we used to talk to our rubber duck. And now we just talk to a chat AI system that gives us immediate results as well. But I think the danger, of course, is that if we just take the first result and everything is good, then it all becomes a bit samesy in the end as well. If it's great code, that's wonderful, but who knows that? I think the problem with chat systems right now is that they're so limited in one response. That it feels like going to Google and hitting, I feel lucky and hoping that the first result is the best one that you get. So the the immediacy of a chat system also for, it's not necessarily the best way to get some code because you want to compare four or five approaches, four or five results before you actually get the one that you want to have.[00:20:00] So I'm really excited about what Copilot is doing. And it was I spent the last. Two weeks for my talk, reviewing all the stuff that they've done. And yesterday they wrote a blog post about it. So I'm like, great. So all the research that I've done they published with their own design team, but luckily enough we match. So I've seen the right thing. So I'm very happy about that. Yeah, you would you say that copilot is definitely taking on this challenge of Making the powers of LLMs into a domain specific space. That plus and minus code diff it's a refinement step that is built into the tool. Is that sort of the direction that you want to urge people to develop in? I think one of the biggest successes in the last 10 years at Microsoft was VS Code. And VS Code is, was a success, not because it was the perfect IDE that does everything for you, but it was a damn good editor. It was hackable, so you couldn't not hackable in terms of security, but you can write H T M L C S as a [00:21:00] JavaScript to extend your editor, to change the editor to your needs. And that's a lot. That's what a lot of companies right now forget when they roll out AI systems. They just roll out a product and then they put a shoehorn open, AI's APIs into it. Great example is I had one company that sent me an invite a month ago that's oh, we're gonna do an AI driven. Text editor for writing. And I'm like, Oh, cool. Yeah. So I'm going to try it out. And the main editor is terrible. It doesn't do copy and paste properly. So as a text editor, it completely fails on the first, in the first step. So basically it doesn't help me to do the job that I came for, which is writing. And it doesn't matter if I can now press a button and open, I can give me five paragraphs of random stuff that I wouldn't write that way. So that doesn't make it a good editor. Make sure that the first. The core of the tasks that people came for came for us to think that works and this is why integration into VS code with GitHub Copilot was a very [00:22:00] simple way to get that working and I've got other examples. One thing I've been using for years and years is Hemingway. Hemingway is a text editor that allows you to, that while you type, shows you how complex your sentence is. And it colors it in the background and says this sentence is too long. You should probably cut it up into three sentences. This sentence is like is like. not active. It's passive voice. You use too many adjectives. So it actually takes the writing style of Hemingway as the as the inspiration. And this was always really cool because I just wrote something and then I put it into Hemingway and then I went through all the like colored things and deleted those that I didn't want to have. And now it actually uses OpenAI as well in the background. Now you can actually long click on one of those paragraphs and say, let's fix it for me. And instead of just fixing it immediately, It will pop up a little interstitial window where it actually shows you the old text and it writes the new text on in a secondary pain under it. And that way tells you like, you could rewrite it that way. And then you say, [00:23:00] okay, I rewrite it that way. So it didn't take over. It didn't take the task of writing away from me, but it helped me making my writing better. You can also do the you can just hit slash and then say keep writing from here and then it indexes the rest of the text and then keeps writing and also tries to write in a style that is almost like the rest of the text. And I think that's another thing that that a lot of systems do wrong. They just get like the AI results open AI results and say they sound great, but they always sound like a Californian that has been like over excited about things. And it keeps repeating the same content as well. So writing texts in Copilot, for example, is possible, but it's actually not as good as you think it is because a lot of repetition and a lot of like hollow phrases that just don't sound like you. There's an interesting thing in GitHub Copilot as well, if you use it, if you open more tabs on the side with code, it references that code first, rather than just getting the rest [00:24:00] from the LLM itself. So it recognizes its context. And I think that's the thing that a lot more tools should be using. So if you say Hey, you're a writing tool, upload 12 of your other documents, and then I can recognize your writing style and later on roll it out that way. So I think that's a very. Interesting approach to that. What sort of domains, Chris, do you feel are going to lend themselves? really well that are ripe for this sort of in depth interaction pattern with LLMs. Something like a writing style or code. These are text editors are really interesting because you can immediately have a refinement window. You can insert the human into the process there and you can get higher order outputs that you might not get with ChatGPT. What other domains do you think are going to lend themselves well to that? Well, I think graphical editing and video editing is another one. That's a big one. Video editing, I'm using video tap and I'm getting super excited about them right now. What that one does is that you could, or Descript does [00:25:00] similar things as well. You upload a video. And then it actually generates a transcript from your video, and then it uses that transcript and at OpenAI to basically generate a blog post from the video. So it actually turns into paragraphs, turns into sentences, finds out, pull quotes. If you talk about a a.com domain it automatically makes the link for you. And it also generates. Tweets from your art, from your video, like short bits that it finds that are more interesting. And even clips where it says okay, you can highlight parts of the text in the in the transcript. And then it generates a video only from that part of the text that you highlighted. So that's something to do like teaser videos on Instagram or for the big one. And that's something I did by hand in the past. And again, it's like you start from the video and you start, you stay in that context. You don't just, Randomly in invent things around it. And I think that's a very powerful way of doing it. The other thing is graphical user interface. Of course. I'm, [00:26:00] I've got friends in the company called Locify. What they do is, and I'm super excited about that is that. As I said before, we don't do Photoshop anymore. We use now things like Figma or we use all kinds of other editing tools in the browser. And we also don't paint designs any longer. We actually have components in our designs as well. So what I always found it a weird thing that we have components in our designs. We have components in our code. Why do we keep redefining those? So what I think like Locify does as well is take your take your designs and automatically recognize what could be a component and say is this a component like a login? Is this a component? Is the button? Is this a table that should be sorted? And then generates the React components or the view components or all the other components for you. So I love that because it means reuse. And I always hated that chasm that's in between designer and developer. That like designers make something perfect and then they send it over to the [00:27:00] developer and we reinvent something that looks closely to what it is and then wonder why nobody is happy. So having the integration of AI tools into design suites that actually recognize what certain components are and already generate stub code for you. I think that's a super powerful thing that is interesting. To where it can go in the near future, because that's one thing that we've been always dreaming of the, what is what you get dream of a dream weaver or front page. That was all a really bad idea to a degree because we started with a graphic. We didn't start with a component or we didn't start with a structure, but if you if you. Use something like Locify or other systems that do similar things. It actually validates your your effort as a designer to use, to create components or reusable components in design as well, rather than just paint every product of product page of your website. And Chris, we're getting up to our time here. I have one [00:28:00] bigger question to round us out with at the end before we go into our outro. I know you're big on user experience, and you said yourself at the beginning of the podcast, you love an interface, you love working on an application that has that human element to it. Do you, where do you think the best user experience gains can be made? Do you think it's. Something like integrated apps and VS code. Do you think that there's a reinvention that's going to happen in ways that we haven't yet seen of how we can interact with these AI models? I think the biggest gain will be to to augment what we, what people already use. It's always the danger of the great thing about, as we said before, chat GPT and the amount of app users it had was the big thing where everybody gets excited. Cause that's what the money people want to see. They don't really care what the thing does, as long as you can show a lot of ads in it or whatever. So the idea that I think where it becomes more useful for end users, the AI bits that we have right now [00:29:00] is augmenting and analyzing what we have at the moment. And you can do that in clever ways. Like you could you could, I think I'm quite sure Google Docs will. Change immensely the next few weeks office does similar things as well, but I'm very worried about that. The branding bit in this case Microsoft tries to shoehorn being chat into everything right now. And again, it feels okay, go back to another part of the page and go back to where you are. So, analyzing what people already use and making it easier for them to use is probably the biggest gain. And again, that's nothing new. If you think about it, the first smartphone that came out, the biggest joke was when people said oh, it doesn't have a keyboard. Nobody's going to use that one. Nobody's going to type things that doesn't have a physical button. And nobody's going to write text. And then we have predictive texting in our mobile phones. And after a while, the phone actually recognizes what words you're supposed to write and what you've written autocompletion for you. [00:30:00] I think the auto completion of texting on mobile phones is a great example how that will become a normal thing for us. I hope that in a year's time we don't even realize what's AI generated anymore because it feels so natural to what we do anyways. It just makes us more effective. I think that dream of saying like I'm using AI to become an artist and to be able to to write great texts without with a press of a button, is a lie that we keep trying to sell to people like you are. You have enough creativity and you have enough knowledge in you. I think it's just a matter of your tools helping you to avoid blockages and to avoid falling into the same traps of doing the same thing over and over again. So hopefully this is when the hype has. We will actually realize that it's people can become writers, people can become coders, people can become video editors, people can become designers. If the tool aids them along the way and helps them to [00:31:00] even out the problems that they create or even out the issues that they create rather than just be one. That generates things for you because right now fully generated stuff is still very obvious and very annoying. Great example was the secret invasion marvel tv series where the intro was made by an old version of wave of I don't know the google video thing and it just looked horrible and and they did it on purpose to show It's like a secret invasion kind of thing But we will get tired of this very quickly as well The same way we got tired of like dog filters on instagram and all these things So I think the when it comes to professionalism Integrating these things in the tools that people use anyways and making it super affordable to do it Will be a great way to get everybody more Involved in the creation process rather than just thinking it's magic. I love your last comment there, making sure it's affordable because as you hinted at earlier in the [00:32:00] chat, Or in our conversation here, chat GPT bleeds millions. What's the sustainability of this? Are we really, can we really build a reliant workforce on something that drains? Probably not. So it's going to be interesting about how that aspect of tools are accounted for. as we move forward. I hope that local Hosting will be much better as well. I was part of the w3c group on machine learning on device And mobile phones do that funnily enough really well And computers to a degree as well. So I hope that instead of having one centralized LLM in the future, we could have a distributed ones. Like it could be like through a, through an own protocol like BitTorrent, where we could actually share these systems amongst all the users that they have. So the more users a product has, the more percentage of their computer you could use to actually generate things for them. So they don't have to burn the main server every time. Ah, sounds the theme of decentralized computing and [00:33:00] systems coming, hinting at us again. Chris, it's great to hear your perspective on these things, especially like. From yourself as an individual being in the industry for decades, you've seen the web get built and you've seen it change and hearing comments about well, it's gotta be what they already use. It's it's so simple when you say it. But it's hard to see that when there's a lot of, when you're in a forest of excitement. So thank you for reeling me as well I'm sure a lot of other listeners in when you're looking at all the AI stuff coming out I'm sure people, if you want to hear more about what Chris is cooking up You're gonna give this talk at City JS. Do you know when that talk is coming? That's eighth of next week, basically eighth of September. Don't know how long the video will be then. But yeah, I will publish posts about it as well, as soon as it's out there. Awesome. And where do you publish? yourself and where do you blog? It's christianheilman. com is where where basically the central is. And then I just [00:34:00] spawn out on dev. to and everywhere I can LinkedIn. But basically it's christianheilman. com is the center where I am a code point on Twitter or X or 10, as it's called now, I think we should call it 10 just to mess with him. And I was also on mastodon at toot cafe, a code point at toot cafe. So, I'll be available there. Awesome. Well, Chris, thank you again for your time. It's been a pleasure. You're welcome.