Laravel Cloud Office Hours

Okay, there we go. It like took a minute to like do the live thing like to update and show that was live. So I was like, what just happened? Uh, hi everyone. Happy Tuesday. Tuesday. Cinco de Mayo. Taco Tuesday. Cinco de Mayo. Celebrate. That is true. I synced up on Taco Tuesday this year. Bonus. I don't think I'm getting any tacos either. I had tacos last night for dinner, so I'm not doing it again today, unfortunately. But there's no problem with doing it back to back. You're right. But I mean, I have a picky three-year-old and I had to fight her for tacos yesterday, so I'm not going to fight her for tacos today. That's fair. Uh, let's see. Hi, Florian. Welcome. Welcome. I was telling Devon before we went live that you're always on top of the streams. Like I saw you share and you're like yeah he's like five office hours in 10 minutes. Oh look goat. He also zoned in on the background. Tell us how he was on the background monitor. It is Super Nintendo playing uh Lamborgh. I got a uh Raspberry Pi back there running an emulator. That's uh superintendent. They got all left. I think Andy, you might be cutting out a little bit on the end of words. Devin, do you hear it too? Yeah. Okay. It's actually just Andy. He speaks funny. No, he doesn't. That Iowa accent. Yeah. It sounds like the gate on your mic is cutting in when you get a little quiet. which with the AirPods doesn't make any sense. You probably have no gate control, but Andy, stop trolling. That's what Florian said. You just like practiced it just to mess with us this whole thing. Okay, I'm messing with my AirPod settings. Does that help at all? It sounds fine now. Yeah. Okay. I was like aggressively doing it near the end and I was like, am I the problem? Make sure I'm on Wi-Fi. I was on um had some meetings today from the orthodontist parking lot. So, I was on my hot spot on my phone making sure it wasn't a connection. Nice. Who's uh who's the visitor in the background? That is Murphy. Nice. I uh I used to I put a gate up so that my kids don't fall down the stairs, but my dogs used to literally just join meetings. One of my dogs, she'll like just jump in my lap and be like, "Hey, what's up?" My dog is in my lap right now. She's sleeping. Nice. My dog is much bigger than your dog. I'm gonna wake her up. She's a baby. She's like 8 lb. So, not as bad as like jumping in my L. She's so tired. I have a 80 lb Labrador and like a 35 lb Aussie Shep Bernese Mountain Dog mix. Yeah. So, just a little bigger. Yeah, just a little. Yeah, 10X. 10X. I guess we should start getting started. Um, I did have a question for chat though because we were talking about music in like one of our music channels in Slack. And so I want to know what do people listen to when they code? Like do they listen do they listen to nothing? Is it silence? Do they listen to only like instrumental? Like they can't have lyrics or do they like go full send? Full send music. Full send. Full send. And if I'm if I'm happen to be coding on stream, it's full send royalty-free music. That's fair. royalty-free music can hit though. Like some of it's really good. Oh yeah. Oh, there's some uh like YouTubers that have crossed into music and stuff and they've like made a thing that they don't uh DMCA strike uh streamed content. So, some good stuff there. I guess we'll jump into it. So, hi everyone. Welcome to Laravel Cloud office hours. If you're wondering why we're doing this on a Tuesday and not a Friday or Thursday, we switched them to be on Tuesdays now. So now they'll be every other Tuesday at 10:00 a.m. Mountain time. Uh my time zone that most other people aren't. So you might recognize 12:00 p.m. ET or 400 p.m. UTC. Uh so that is our new Laravel cloud office hours time. We're here today with Devon who does the cloud office hours with me, but we're also joined by Andy who you might remember from the last office hours that we did. Um, so Devon, do you want to do a quick intro real quick and then Andy, you can do an intro? Yeah, absolutely. So Devin Garbalosa, solutions engineer here at Laraveville. I work on our sales team. I help customers that are looking to migrate to cloud come up with a technical implementation path. And I am Andy. I work on product on Laravel Cloud. So if you have any features or bugs, you can hit me up and let me know. We love that. And you can also ask anytime during the stream as well, which is what we're here for, is answer questions about Laravel Cloud. If you're confused and you don't know what Laravel Cloud is, it's a fully managed way to host your Laravel applications. And you can ask any questions about cloud during this office hours. And we'll get to that as well. And yeah. Oh, let me share the Slido link again. It is in the description. You can share your questions on the slido. You can also throw them in chat and we'll answer them as well. Let me do that. Florian made a shout out to the exclusive already. The shirt. Yeah, I have It's an exclusive. I have three. Yeah. Did you did you just like count how many I took and you're like I'm going to one up? No, I I made sure I made sure I left there with three. I I wear I so I only wear Laravel these days. So I would see all this awesome merch you keep getting. Hey, you just have to travel a lot. Travel a lot. Yeah, we don't have the travel budget and product. Also, here is the slido link linked here. You can throw your questions in here and we'll be monitoring it. You can also throw them into chat. Hi. Hi. Welcome in Francisco. We have Francisco in chat. We have Florian. We have the whole team here. I think we need to start a petition for Francisco to update his profile picture and also to the Wolverine one. Oh, the Wolverine one's good. It's good. I was like hesitating because at Miami I was hanging out with Franchesco. So my brain's like Francisco, Franchesco, Franc. It's like which one. Um Florian said, "All right, bring them to Boston. I'm taking one." I will make sure that I do not have one of my own that you could steal. But hopefully they'll some will be there. I think that's the plan. It's really nice. So the the cloud logo is actually embroidered and on the back the Laraveville cloud like tag thing here is embroidered and it's real heavy cotton. Like it's it's a really nice it's a nice quality. Like it's different from our other t-shirts we've had. And I told Abby we need beanies that are that color and look like that. Cuz the other beanies were the bright blue and they're a little tight when you have a large head like I do. The same both said the same thing. Is it near here? I think I have them. But the uh the Laravel beanies from the San Francisco meetup were in January. Those ones are awesome. Those are good. Yeah, the red one from like a year ago is really good, too. Yeah. Rub it in. Guess how many. Zero. Wait, what? Oh, I said guess how many beanies I have. Three. And then I have Laracon EU beanies. Oh, I have that one, too. Yeah. Yeah, I don't have any on me. Um Oh, we have a question already. You can start with us before we get in because Andy is going to talk to us about something exciting today. Let's answer a question or two. Question, a general question, but what do you think about the phrase programming is mostly solved by AI? You could not have picked a better guest on the cloud office hours uh for this question too. I know Andy is really opinionated about this one. I'll give my take first. Uh false. It is not mostly solved by AI. I think that the fundamentals of coding um are still an absolute necessary uh skill set to be able to do that uh to code with AI effectively. I think that understanding the scope of what you're trying to do and how you effectively do that is there. Um, fortunately slash unfortunately AI has access to all of the code that was written in the past to whatever the cuto off date was and uh, not everybody writes great code and so it's got a mix of good code, bad code and serviceable code all in there and um, if you don't know what the difference between those are, you can have a lot of foot guns in place. So I would not say it's mostly solved. Now if you have the fundamentals it could write the code for you. I don't write a lot of code anymore because the AI writes it for me and I just give it the scope and the and the process that I want it to follow. So, u that's that's my take on it. Yeah, I think we still have that and in many cases out again the noise gates don't get I've heard about seven words out of those four sentences. Let me uh let me cut there. Let me I don't know if it's because of the dogs barking in the background too. Like if like the gates picking that up and trying to like Those are my dogs barking. Oh, okay. Never mind then. that's not why then. Well, he's doing that. Leah, what's your take? I think my take is, yeah, AI can generate code. You still need to know what you're doing because if you just accept everything AI is giving you, you're going to end up with like slop like people say. Um, you're going to end up with like bad code. It's not going to be sustainable and as you like add things, it's going to break things. So, you need to know the fundamentals like you said. If you know the fundamentals, yeah, the AI can generate the code for you, but you're still acting as a code reviewer. And a lot of times you still have to tell it how you want things and be opinionated with it because it will still do things most of the time in a way that you wouldn't. Of course, it helps using like an opinionated framework like Laravel and it helps that we have Laravel boost that helps with the context for the AI agents as well as we have a Laravel um best practices skill. So having those skills and stuff like that will help the AI write better code, but it's still not to the point where I think if you want to build a business, you would just do that yourself if you don't know how to code using AI. Like you still should know the fundamentals. Yeah, I think that's a great one. And I think the interesting thing that you highlighted there is like the write better code skills and things will help you write better code, but there's still the unknown unknowns of like there's a thousand right ways to do something. And so what is the the most right way for your context is so important to to capture in there. Exactly. Or even Andy's his whole system out. I don't know if he left cuz he's just like, "Let me just rejoin and see if it fixes it." Yeah, normally it's Leah with the computer problems. What is this? I know. I haven't even said, "Oh, no. My mouse once today." Not yet. Not yet. Oh, Auntie's back. Okay, let's see. We back. All right. Is it better? I don't know. Is it good? We're good. So far so good. So, what's your take? Um, yeah. I was saying like context is king and so if you have a year of development experience 5 10 15 20 you can always add that in to what the AI is doing. So I think the AI is just like an enhancement. It's not a replacement. Granted that could change next week. So this whole world is very fluid but I don't think that it's totally solved because I think Devon you made a great point. There's a lot of PHP code that I wrote 15 years ago for WordPress that's on GitHub. And if any of that was used to train AI, I feel sorry for the AI because none of it was probably very good. So the same if that's if that's the context it's using, then it's not going to work out well. So I I don't think it's totally solved yet, but again, that could change tomorrow. I have this uh this one story I tell all the time that my last role, I was reviewing code uh with my manager and we were going through because something was not working as intended. As I'm like reading the code, I'm like, who wrote this? What is like why would anybody? And I literally was like going off the handle about it. I go into GitLab and I looked at the blame and I was like, "Oh, that was me. Oops." So, yep. I definitely feel like, yeah, I apologize for all the bad code that I contributed to LLM knowledge. Let's see. There was also another question um by Rosswing uh question around database backups. Will you be adding more frequent backup options? At the moment I can only see daily as an option. Yeah. So this is a really nuanced one. I'll take the first stab and then pass it to Andy to tell us where it is on on the backlog um of things but or the road map I should say. It depends on what database you're using because like our our serverless Postgress powered by Neon has point in time recovery so you can go back to two seconds ago. Um RDS also has point in time recovery so it's the same. Um and there's there's other options. The other thing is there's always the option to kind of yo process for backups. Um I know spot spati has a package for helping do that and whatnot. Um, but the Laravel MySQL I believe has some some grand plans and this where I'll pass it to Andy uh to add I believe at some point point in time recovery but in the beginning it will be kind of more granular backup windows. Right. Yeah. So we're looking at more manual granular backup windows. Point in time recovery is definitely on the list. Um some different scaling options. I think Taylor's tweeted out a couple previews of what we've been working on lately and hoping to launch in the next few weeks. So, a lot of changes and improvements coming to Laravel, my SQL. Um, and yeah, point in time recovery is high on our list. Nice. Yeah. So, that should that should close that gap substantially. There's a good call out though saying that Neon on if you're using the Postgress database through Neon that that already exists though. Yep. Yeah. Yeah. And for private cloud customers, they have RDS available to them. That's already an option as well. Yep. And then I feel like this is a question. It feels relatable or relevant because Devon's been doing um webinars on migrating from vapor to cloud. But someone said or Carlos said, "Hi guys, what's the best way to migrate cues and cache from vapor to cloud? Any gotchas we should know about?" So Q's is an interesting one. Uh, I think the quick answer on that one is drain the queue on the old system and repopulate it on the new system. That's the easiest way. And then I got a hot take on caches. If your cash is not frangible, your cash is wrong. So, uh, you should be able to repopulate that pretty quickly from people using the code and it should be a little bit slower response times in there. Uh, if you don't have that, then you got to go through a migration path. Um, you can connect to the two endpoints and kind of migrate keys over. But uh in a really great um the I think the best way to use the cache is to make sure it's frangible and repopulatable. So that way if there's ever an issue that you have uh even in production where something's going wrong and you can't trace it back, you just clear the whole cache and let it repopulate. Any uh any other maybe not so hot takes on those two options you guys have or those are excellent answers. I will add agreed with the hot take. Yeah, it wasn't too hot. It wasn't spicy at all. The other thing I could add with the cues, technically SQS is supported uh because it is AWS as long as you provide the access key and access secret. So you could just pause Q work um and then just resume off SQS. Um it works very well. Um we find a lot of times that radar vali does make for a better Q driver in current state. Uh but some of the new stuff that we're looking at coming coming soon TM manage cues is backed by SQS is uh it is a very reliable and and very good system. So that would be the other thing with cues if you can't just drain and repopulate is just attach your new instances to use the SQS uh client still. Yeah. And Carlo if you want to try out that new queue let us know. I'll I can hook you up. I know a guy. Yep. Got got the plug right here. What would be the best way for him to reach out about that? Like just on stream here or should he contact you on Twitter? You could open a support ticket and just put that in there and say Andy told me to. Uh or if you want you can reach out to me. My email is just Devon at Larville and I'm Andy. So yeah, either one. Easy enough. Hold on. I'll type this or Oh, if you're typing them, just put it up. Yeah, just kidding. Oh yeah, I'm talking Devin at Laravel. That's it. And just Andy at Laravel. Andy, you don't have to have your last name, right? No, thankfully. I actually got a nice story while we were at Laracon EU with Andy. Uh we were we were hanging out at the bar and he was like, you know, when I was thinking about taking the role, Calvin uh was like, "Hey, dude, Andy, Lar's still available." Not going to say that it was the factor, but right, it seemed like it. It it was definitely helpful for sure. I have the comment pinned with the email so you can reach out to either of them at devonarvel or andyarvel.com. And there was one more question related to the AI stuff and then we'll jump into the fun stuff that Andy's here to talk about. So Don asked, "Do you use LLM code reviews in your development workflow like on CI/CD?" That's an interesting one. Um, I have not leaned too far into those mostly because my workflow has really been let it run, open the PR, and then I review the PR. Um, but I do have like cursors, bugbot, and some other stuff running in there, and I kind of look at them. Um, I feel like those even more so than regular LLM use lack the context that it needs to have. And I haven't found a good way to like load it in there of like I should probably have like some rules file or something in there that like tells it how I like to look at PRs and that would probably be better and I would get better out of that, but I haven't really invested too many cycles into that. It's uh I haven't seen the value for me yet. Um but if your workflow is different than mine, it may be absolutely valuable to you. Yeah, I think depends on the project too. I have um on some of my side projects, I have a review and release skill. Um, and then that does a lot of that and I've to Devon's point kind of coded the rules that I look for when I'm reviewing and writing PRs. U, so I run that skill when I'm ready to go. It does like full code review, runs the tests, updates like a change log file, and then one of the cool things I've been playing with is like this idea of like having the app learn about itself. And so when it's doing the PR and that review, it's writing anything that new the next agent needs to know into the agent's markdown file. And that way the next time we're working on it, it can reference and go back and see like what the a last agent worked on and then anything new that was learned and it kind of compounds from there. Uh that really helps that review and release skill. Um because it kind of has that context that keeps growing over time. Interesting. Open source when? Soon. Yeah. Um I personally haven't used uh LLM code reviews either. Like I like to ask Claude or um Composer like whatever model I'm using to review my code and like if there's a more optimal solution for what I'm doing, is this clean? And then sometimes I'll like ask it to give me a list and I I'll review it and then see what I think. Um, but then I'll just open a PR and I manually review because I'm really big on manual reviews of code still, especially if I'm using AI a lot. I want to like manually go in and make sure I understand things or that like it didn't do anything dumb. But I do think having this like LLM code review kind of in um your workflow and like with CI/CD is a good idea especially like like Andy said it's specific to the project but I'm thinking of like any of our marketing sites how now some people in marketing are opening PRs for that having some kind of like um code review having like rules based on like how you would want the code quality to be and stuff could be super helpful for like Chip and Tim to just quickly go through those PRs. So, I do think it's kind of something maybe we're like going more towards, but I'm still very much a manual reviewer. I think that it's going to be a scaling problem. I talked to a CTO last week who's drowning in PRs because he has a 10 person team and they went from like a few PRs a week to like a few PRs an hour. And so, there's now a bottleneck at the at the PR stage of deployment. And so he's looking at all these different ways of incorporating that in the CI/CD pipeline because now he's the bottleneck because he was manually reviewing all those and now he's getting 10 times the amount. So I think that's coming sooner than later. I haven't played too much with those in the pipeline, but um I think there's definitely something there for sure. Yeah. I think the other mitigation thing that I do in my work my kind of workflow is I lean heavily on plan mode in the beginning and so I kind of know what good is supposed to look like before it comes out the other end. And I don't I wonder if I could find some way like your review and release skill create a scale that outside of the plan it's got to it has to have like a review like checklist better that it checks against and then let Bugbot or something go after that. Yeah, it's interesting. All right. Maybe by next office hours I will be using this. I know. Especially think of I'm like you're right though about the bottleneck being like code reviews now when you can just ship PRs like open PR so quickly. So yeah, now I'm like man now I want to play around with that too even for like my own side projects. Okay. Do we want to get into the fun stuff now? Andy let me reshare here. I can give it to I'm excited for this. So excited for this. This is a big one comes up a lot of people I talk to. Yeah, especially you for sure. Yeah. Okay. Are you ready for me to put the screen up? Yeah. Go ahead. Start with the docs. Um, so I think I've almost actually in a couple days I'll be one year at Laravel and this has been on the road map I think since I started. Uh, we keep kind of pushing it down the road and finally decided to tackle it um, about a month or so ago. So Ryan Chandler's been in the kitchen cooking on this one. Uh, and we're hoping to release it very soon. Um, so what are our back? What's that? I said I was cooking over here. Uh so arbback is a acronym for role-based access control. Uh and so the way we have come at it come at this is this roles and permissions model. So any forge user is going to be very familiar with what this is going to look like because we borrowed a lot of elements from that. Um and so the idea here is um across the different plans on cloud you'll have a different version of arbback. So if you're on our starter plan, nothing's going to change. You'll have your admin and developer roles like you do today. If you're on the growth plan, we're giving you what we call team arbback, which is these predefined roles. So it gives you a little bit more granular control over how your organization is set up. So these predefined roles include admin and developer, like we just discussed with starter. Those are no different than you have today across the entire cloud app. This adds manager, uh, finance, and viewer roles. And so these give very specific pieces of functionality to different kinds of team members. The idea here being if you're in like growth business or enterprise plans, you're probably growing your team, you might have some more requirements. You might have somebody that's really only interested in your usage tab, your billing information, and that's where the finance role comes in. Um, you might have a manager that's kind of either in DevOps or running an engineering team that might want a little bit more control than a developer would get. Oh, we got we got Ryan chef himself. we summoned Ryan. Yes, you could hear his name being spoken. Um, and so those are those predefined roles you're going to get in the growth. Um, and then at the business and enterprise tier, you're going to get what we call advanced arbback. And this is really what you have experienced in Forge where we give you the ability to create custom roles. It adds an an additional predefined role called restricted which I'll share about a little bit more in a minute. And this is where it gets really cool u and where we add a little bit of Laravel cloud flavoring is this application environment level access. So you can imagine a world where I run a team. I hire a contractor. I maybe only want to give them access to a specific staging environment for this app that they're going to be working on as my contractor. Um so they can go do everything they can in staging, but they don't have access to anything else. And so that's where this environment and application level access goes. And then even better is when I add them, that's where this restricted role comes in. They get a restricted role at the organization. And so by default they can't see anything. And this is this idea of lease privilege. So if I add them as a restricted role they get an invitation to your cloud organization and they get there and they can't do anything. I have to go in and explicitly add them to the apps and environments that I want. And this is a big ask from business and enterprise customers because a lot of them are following this lease privilege model where we only want to give access to the specific things that they need to get access to. So that is a high level of what um what's coming very soon and I will jump into a demo. Demo time. So here is my growth org. And so if I go into settings and access what this page already did exist, but now you'll see all of these predefined roles now in there. So I can go and invite people. I can add them to these roles. Pretty easy there. So that's essentially what you're going to see at the growth tier. Now, if we jump over to the business tier, this is where we get into some of the cool stuff. So, let me show you what that looks like. So, first of all, you get again access to all of the uh predefined roles. You'll also see this new restricted role down here. Uh and then I have this custom RO editor. So I can go in and create a role and I can add granular permissions to that role. So there's a gazillion examples. One we were talking about today is what if I want to have a developer role but give them access to update environment variables. I can go in and create a superdeveloper role and add all the developer role permissions and then give them that environment variable permission as well that we only typically give admins by default. So there's a ton of different ways um these custom roles can be used. And again, these permissions, there's a lot of them um and they're growing. So every feature we add, we'll be adding these granular level permissions to it. Um now, if I go into an application, this is where some of the cool stuff happens. So I can go into our emoji invaders game here. I can go to settings. And you'll see that we have environment and application settings. And so I can go into a this specific environment. So I'm on main right now. And I can add a user um to that environment or I can go to the application and I can add a user to this app who then gets access for whatever role I give them to all the environments in that application. Now amazing. Another cool part of this that um saves you some clicks is if I don't have this new user in my cloud or at all and I go and add them to this application, they'll automatically get added to the org with the restricted role no matter what role I give them here. So I could give uh me developer access and I can share that. Now I have developer access to this entire application. And if I go over to settings, this user will show up as a res. No, it didn't. Oh, I already added that one. Sorry. Bug in the demo. Um, add a new one. Use me. Just throw me in Throw me in, coach. Let's do that. I'm gonna go steal a Space Maid game, guys. We'll add you to We'll add you to Suji here. You see like Devin tweet like 30 minutes after the stream. He's like, "Look guys, I made this game all by myself." Space Invaders. It's like that meme here. I made this. Okay. Now, if I go over here, let's see. It added Deon as restricted to the org. Invite pending. And then in the app, he is a developer role. So, this is the idea of lease privilege. So, if he goes and accepts this invite, he'll be able to log into this org. He won't be able to see all of this stuff. He'll just be able to see that one that I explicitly granted him access to. All right. Um, that's I mean there's a lot of complexity behind this, but this is the beauty of it. We've made this super complex model very simple to use within the cloud um system. One other thing that I will give you a hint at special preview um for the live stream here is one of the similar features that we've been working on is around API tokens. And so today if I go create an API token, I have to be an org admin. I go create that API token and that token then gets everything that that admin can do. Right? In the world of agents, that might not be the best idea. We've seen horror stories of claude de deleting production databases, right? And so we're working on this idea of scoped API tokens, which is going to use this same interface. So I'll be able to go in and create a new API token and only give it these specific things that I want it to do. So maybe I only want it to create environments. I can go create that API token. Now, if the API token goes and tries to delete a database, it's going to get a 403 restricted error because it does not have the scoped access to that. And so, within this new scoped API tokens, you'll be able to grant it these granular permissions, you'll also be able to give it access to a specific app or environment, similar to adding a user. So, it's again this idea of least privilege. So, I could go in and create an API token that only has access to the staging environment with these specific permissions. I can create that, give it an expiration, and then pass that out to my team. Now, these will still only be able to be created by admins, but those admins will then be able to hand those keys off uh however they do that. And um those those keys can't go do anything they want on cloud. I didn't know it was going to be Christmas when Andy showed up today. I did not even know that was in scope. That is amazing. I cannot wait to go tell everyone about that. I didn't know that about the API tokens. I'm so excited. Yeah, we uh we have one person we were talking the same interface here. You'll be able to go in and pick those permissions, set it to the API, and you're good to go. Yeah, I was going to say we had one one person we were talking to that they wanted scoped API tokens specifically because they wanted to host a CTO at a company. He wanted to host a hackathon with his team and have them ship it right to cloud and he wanted them to use the the CLI and whatnot. And uh he was like, "Yeah, but I don't want them to drop production." So, so can't wait to give him that news. That's like a really creative way people are using cloud like you know host hackathons and things. I love that. I love the hackathon idea because cloud's great for that. You know, it's great for getting projects up really fast. So, it's amazing to use for hackathons. Also, Ryan said, "Secret side project. Now the whole world knows." I'm not going to lie, Ryan Chandler has the best side projects. Every single one he's come up with is literally a banger. Yeah. Ryan, if you would want to be on a cloud office hours and you can spoil your own secret side projects. do that, too. Yeah, we're all besties here. There's no secrets. Let's spill all the tea. Just say everything we're working on, right? I get like 10 messages from marketing like what are you doing? What is that? Awesome. This is This is exciting. getting our back out there is going to be a huge unlock for a lot of companies. A lot of excellent workflows. Uh agents too. Just being able to more confidently ship without explicit or uh you know with explicit permissions is going to be a huge unlock for a lot of different use cases for sure. Awesome. So what what other secrets you got over there, Santa? Uh secrets is a good secret, but I'll leave it at that. Little teaser for you. That's a good one. I'm excited about that one, too, I guess. Anything else to say about Arback, too? Like when people can expect it? Anything like that? Uh days. Days. We'll put it at days. Yes, we love things measured in days around here. Yeah. I use that all the time. We have a specific date in mind, but there's a couple little polishing things that we we're working on, but um That's like my new favorite thing. I heard somebody say that one time. Now when I'm talking to people about stuff, I'm like, "Listen, we're measuring it in days versus weeks or weeks versus months." Like, it's just feels so much cleaner. But yeah, right. That's so amazing. Also, we have Graham here. Hi, Graham. Graham, what's up, man? Says, "Good afternoon from London." The time zones always elude me, but I I know it's probably like what, 5:00 p.m. or something. Somewhere around there. I don't know. Graham's always up. I' I've literally had phone calls weird times and I'm like, Graham, it's like 10 o'clock where you are. So, Graham was there for most of the 12 hour stream, too, right? For cloud anniversary stream, I think I think so. And I know it was like very early in the morning by the time we ended because it was like 700 p.m. my time. That was a long day. It was 7 a.m. to 7 a.m. streaming 12 hours. And I did a lot less that day than you did. I was serious the whole time. I I by the end of it, people are talking about stuff in cloud and I'm just like words don't have any meaning anymore. I I think the only rest bit you got was when Nuno kicked you off stream for an hour. My favorite. He literally showed up and he's like bye Leah. Bye Josh. You can leave now. He's like can I have the hour? And I was like you want to do it alone? And he's like yeah. And I was like say less. I will go like eat food for an hour. Okay. I was right. Graham said 5:40. I'm getting better at knowing. That's not too late then. Dinner time. Nice nice dinner chat. So, does anyone have any questions about the arbback stuff that Andy showed? Anything else in cloud? Um, I know there's another question about AI, which we can get to if there's no cloud ones too. Um, we did have some comments. Oh, yeah. Like like Ashley said, which hi Ashley said custom rolls. So handy. Um Francisco said next level custom rolls. So there was some hype. Is this what you were going to read from Ryan? Uh yeah. Yeah. Yeah. What happens in office hours stays in office hours except for these stay published on YouTube forever. So and when I make YouTube shorts from it and so then you see like a clip and you're like wait that was just supposed to be office hours. And I can 100% tell you whenever a customer asks a question that I answered on at office hours, I share the clip. Nice. As you should though, it's kind of the point of the public office hours. Oh yeah, it's a great great resource. So, hi Angela, welcome in. Hi Helm, welcome in. I guess we can go to the AI question right now. Someone said, "Yeah, based on your real world experience, what is the best LLM model to use right now?" Excluding Claude. I love the tag on that. Excluding Claude. I know. Yeah. I uh I still use Opus personally for a lot. Um mostly because I've tailored my workflow to work really well with cloud cloud code. Um I do like GPT55 right now. Um, yeah, but my I would say this is a hot take, but the more I say it's a hot take, people just agree with it. Um, cursor 2 on I'm sorry, composer 2 on cursor. Yeah, it's good. Pretty good, especially on plan mode. Pretty good and pretty fast. So, definitely definitely highlight that one too, which and that is based on Kimmy, right? If I recall, they trained Composer 2 on I don't know. I'm not I'm not that deep in the AI nerd zone personally. I haven't used Jimmy personally. I've used composer 2 though. Yeah, I think it's composer is based off Kimmy 2.5 if I'm which is which I believe is Deep Seek Under the Hood or something. Anyway, yeah, they're all they're all rappers on each other at this point, right? We got an Ashley answer. I was going to say Ashley would have a really good answer. Ashley said plan 5.5x high and then for code Jimmy that's how I've been doing it too and then um like my review and release skill usually use a cheaper model too because it's pretty it's more deterministic. Um so yeah using one of the those premium models if you will for planning I think works well and then some of the cheaper ones for the actual implementation and review is good. Um, how are you using Kimmy today? Through what? Open code. I need uh I'm going to have to schedule some time with you, Andy, to show me how to use open code cuz uh I don't know what it is. When everybody was having the Claude like problems where they were like running out of their uh a lotments super fast, I was like trying to run out of mine and I couldn't. And then I used Open Code and I blew through like $100 in like three prompts and I was like, I must be broken here, guys. I don't know what I'm doing wrong. So I have not used Kimmy at all really because of that. So yeah, to be fair, I was using open code with GPT 54 codecs. I w x high like I wasn't going with a cheap model at the time, right? Yeah. But yeah, there's like three prompts and I was just like uh I do this on opus max and I get nowhere near this. What am I doing wrong? So I need a I need a pointer. Yeah, there's some magic in how you orchestrate those um to Ashley's point where you often want to switch models based on what you're doing and I it's what I try to do especially on my personal accounts so I don't burn my tokens. She also said interesting I always use is it soda or SOP state of the art is that what that is? Yeah, I think so. Like frontier model basically or whatever. Save the art for planning review. Save the art equals bookends to open weights for coding so far. We'll give your way a go. And Ashley also says open code's amazing. Yeah. And inim said open code is magic. I def I still haven't tried open code mainly because I'm a grandma and I learn one thing and then I'm like this works. I'm just going to stick with it because AI like changes so fast right now that I try not to just like jump to the newest thing. But I do want to try open code. But that's a great reason for using open code because let's say a new model comes out tomorrow. With open code, you're quickly in the interface able to switch models. And so you don't have to like lift yourself out of cloud code and go into codeex because you want to try a GPT model. It's all right there in that same interface. So, it's the same workflow. You can just switch models back and forth, which is why I got into it. And I think uh Ashley mentioned Open Code Go, which is kind of their API gate or their um AI gateway, which I think you can start with five bucks a month and then it goes up to 10, but it gives you access to all of those models through them as like a subscription. And so you can easily switch back and forth. So how I use open code is I connect it to the each API and so I can go from an anthropic model to a open AI model or open router model. So I use a lot of open router that's where I use Kimmy from. But if you go with open code go they kind of encapsulate that all for you and makes it a little easier to switch around models. Yeah I'm definitely have to check that out. I think you just talked me into it too because I do like that's like okay I want to try codeex or whatever. I don't have to like literally go to like codeex like spin all that up. I don't have to have all these different like CLI tools or whatever. I just use open code and I can connect to the models. Yeah. And open code will observe your agent files too. So if you have cloud code skills, it will it'll check an agents directory to see if there's skills in there and then it'll check a cloud directory to see if there's skills in there. And so it does this like tree where it can kind of traverse across all the different 15 configurations for these kinds of things since we can't all agree on an open standard yet. Yeah, that's uh that's helpful. I literally just went through the pain the other day when I wanted to use codeex for something of sim linking all my stuff and then it telling me that my front matter was wrong and then being like, "Okay, how about you fix this for me then?" Yeah. Okay. We did get a question cloud related. It is saying which one is best Laravel cloud or laral vapor which I do believe is like based on your use case but yeah uh it definitely depends based on your use case. Uh where vapor outshines is a very niche application that requires basically infinite spike scale. um which the gap between where cloud and vapor overlap is is is a ton in that space. Um the way that Laravel the framework itself and most apps that people build on top of Laravel go cloud is typically the better use case because it encapsulates a lot of stuff that you don't have to think about for like high availability for scaling your cues in your um your web app differently and having more stuff there. Um, the other thing is that Vapor does have 15 minute timeout limit just because it's on Lambda. And so if you don't want to write code that can pick itself up in jobs and things like that, cloud definitely gets you uh the better kind of working area for those things. Um, essentially Vapor has foot guns and cloud we've tried to alleviate a lot of those, but there's there's definitely use case for both of them. Uh, cloud runs on our infrastructure, Vapor runs on yours, right? So, really depends on your use cases. Um, but cloud definitely fits the bill for more use cases than vapor does. Great answer. And in MSAD also said this is about AI but is composer 2 exclusive to cursor which I believe it is right. Yeah, I believe so. I think it's a proprietary model. Well, technically now it kind of belongs to XAI, but right. Yeah. Oh man, that works. Wild. Yeah, that I don't know. I've really liked Grock u for like personal usage for like what I've you know the first thing I started using chat GPT and other things for was replacing my Google searches right and being with that. And the thing that I like about Grock the most is it's constantly trained. So it's knowledge cut off is now and now and you know and so you don't have that. So for current events and stuff especially it's it's the it's really good and um their interface for like working with code on Grock is also great but you have to go back to the cave the stone ages caveman style and copy and paste the code in and out of it because they don't have like a good unless you're using open code. So, Open Code had Brock Code fast for free over the winter time and it was awesome and it was it was worked well. It was super fast and open code was kind of sub or they were either subsidiz someone was subsidizing it because I burned a lot of tokens with it. Yeah. Um and it was insanely fast. So, I think you can still use Grock code via open code if I'm not mistaken. Curser gets it soon too since they now have that partnership because I would definitely use Grock for a lot more You would think they would. I think it's part of the plan, the grand plan. I never try to guess at what Elon's going to do. So, which I believe it's technically SpaceX, not XAI, but they're kind of all related. So, yeah. Yeah. Hopefully that improves my Tesla, too. Yeah. Yeah. Oh, there was a question. That's funny. I just imagine you talking to Grock through your I do. I have that. So, the other Yeah. Yeah. Cuz I have the I have the 2025 Model 3. So, the other day I was driving somewhere. Quick sidebar here. And I was I was like, I need to like my kids are going to fall asleep, but I don't want them to fall asleep before we get lunch. So, like, you know, whatever. But like my daughter, it was raining out and I was like, she's restless. So, I literally long pressed the button. And so Grock popped up and I was like, "Hey, Grock, can you look for between here and my destination a any, you know, McDonald's or anywhere that has a play place that's open and whatnot." And it's like searching reviews. I found three McDonald's with a play place with reviews within the last 30 days that it's open. Would you like me to route there now? And I was like, "Yeah, route there and then home." And then it literally was like adjusting and I was on full self drive. So it literally just readed me and went. It was amazing. Nice. I'm bullish on full self travel on Tesla. I can talk about it all day, but this is cloud office hours. Someone asked, this isn't cloud specific either, but they asked if we're going to be at Laracon. Laracon US. I don't know. It's going to be a long haul for me, so I' I've really got to consider if I'm going to make the hour drive to get there. Yes. I'll 100% be there. As will. Yes. Yeah. In fact, I'm so committed. It is my son's first birthday and we're bumping up this smash cake a couple days early just so I could be there. So, yes. And I will also be at Laracon for those listening. Laracon US is in Boston this year. Like Devin said, it's July 28th and 29th. The venue is really cool and we've already announced events like around Laracon US this year. Like um the mostly technical guys are doing a pre-conference party. So on day zero and it's LAR prom. So, some people will be dressing up like prom. We're doing dodgeball this year. Literally in a high school gym. That's like the best part. Yes. And if you want to buy a ticket, you can go to laron. us to look at tickets. Oh, I should have cut that up. I like clicked and then immediately clicked off. Double click. I got like I got to like click the windows is strong with this one. Chip said, "Hi, Chip." Uh, he said, "Need a baby blue tux for prom." Yes. I told Ian him and Aaron should dress up like Dumb and Dummer. Dumber. Be awesome. Yes. Blue and orange tux. Yeah. Chip, if you do a baby blue tux for prom, I I'll I'll go to Gentleman's Warehouse and Men's Warehouse, whatever it's called now, and I will uh I'll get I'll go get the beige one or something. I'll I'll join you in on the fun. We have to take Titanic pictures. So, we we have to like do the thing, you know, where the woman stands like that and the guy's behind her. Uh, I'm gonna abstain from that one. No, I mean I think if you're Oh, you I'll take the picture. No, you you and Chip. You and Chip do the Titan post. Fine. I'm not volunteering myself. I volunteer. I'm down if you're down. Um, and then Florian said, "Can I rent a tux in Boston?" Probably. You should be able to. There are probably few cities in the world that you could rent a tux last minute notice and it's Boston and New York City. So, if it's possible, it would be in Boston. I also feel like there even here there's a lot of like men's warehouse and Joseph Banks and stuff like that. So, yeah, you could definitely get a nice suit, but the tux rentals, they usually like order those out. So, you'd have to get lucky that they got one in your size on the shelf. You can maybe look into it online, too. right. Like you could reserve it ahead of time online. So here's here here's what it is. Usually you could rent these things like months in advance. So you just got to find a reason to come out here to the US for a day, get fitted. Doesn't matter where you go. So Men's Warehouse is a giant chain call across the country. There's definitely one in Boston. So go somewhere, go visit somebody down in Florida or something. Get sized up and pick it up in Boston or Denver. Well, you go to Denver. Yeah, but we're about to have snow in May, so maybe maybe go to Miami or something. It says to snow today and tomorrow. it's like why not? We have Luigi. Hi, Luigi. He said, "Usually you have to give them some lead time for tailoring and alterations." That's true. So, like Devon said, you'd want to go in advance to do for the right money. I am sure there is a tailor that does tuxedos in Boston somewhere that will get it done in a night just just for what it's worth. If it's that important to you, said, "Oh, wait. Okay, I'll visit Justin in Florida. Yeah, you can see the horses. That's the cows. That's actually why I suggested Florida, but I wasn't I wasn't trying to dox him like that. Shout out Florian. Florian did it. I at first I was like, "Oh, no. Did I do it?" And then I was like, "Wait, no, that was Florian." Uh and then Chip said, "Too bad no snow in January." I know. We had the warmest winter in Denver. And then it's like, "You know what? Why don't we just snow in May?" Yeah. I lived out in Phoenix, Arizona for a year. And uh the one year I was out there for Christmas, I was flying back home. It was 35 and snowing in Phoenix because of some weird stuff. And I flew home and it was 72 degrees at Christmas in Connecticut. And I was just like, "This is terrible." It's backwards, too. I love that. Let's see. I'm trying to go through to make sure I didn't miss any questions, but I don't think I did. If anyone has any last minute cloud questions before we start wrapping up. So anyone's our vet coming out. I need a date. No, don't make them drop a date. He said days. Days. Yeah. Yeah. Days. Yeah. So like tomorrow we have to keep the suspense. Maybe. I don't know. Why don't you check? Why don't you go to cloud.lar.com tomorrow, sign in, and see if it's available. That's that's how you There you go. Yeah. I mean, to be fair, it is available to me already. I was in the I know. I know. I know. You're trying to pressure Andy into giving me a date. I'm I'm excited for that one. You'll see be a busy day. Customer notifications for me. I'll tell you that. But I have linked the link to the cloud homepage. I I have it pinned. I pinned the link to the cloud homepage. You can go to cloud.lair.com. either make a new account or sign in so you know whenever arbback is available on cloud. Every time I hear arback, I I hear arback. A shared back. It It is It is kind of the same. It's kind of like a turtle shell. You know what I mean? Right. I won't share my back. I'll be very nice and I won't share my back cuz I have scoliosis and a slip disc. You don't You don't want to share my back. Oh, my back hurts just thinking about that. That pack's bad. It was born bad and it's still bad. So, we won't share mine. We'll share Andy's or someone. maybe Ryan's. Ryan's young. We'll share Ryan. He is. Yeah, he lives, too. So, So, it's a safe bet to share. Oh, someone said, "Since Laravel 13 introduced a built-in AI SDK, can Laravel become the future of AI?" Laravel is the future of AI in my opinion. But yeah, no, that's definitely uh I mean, go look at the laravel.com homepage. It's the best framework for people and agents. That was uh that was a uh definitely intentional decision we made going into Laravel 13 with AISDK and other things that released is that Laravel um you know the the TLDDR on it is Laravel's been around for a long time and it's held opinions for a long time and you know what works really well with opinions? LLMs. So Laravel code that comes out of LLMs is very good compared to a lot of other languages and frameworks. And of course we have Laravel boost which helps with that. We have like we've created skills like I talked about earlier where um we have the best practices skill for Laravel like specifically for Laravel. Um and we're current like constantly working on ways to improve um aentic workflows and stuff for people who are doing agentic coding and using Laravel. So I agree with Devon that Laravel is the future of AI and it is a great stack to use if you are coding using AI especially like if you don't know how to code too for people who want to do vibe coding and stuff like that Laravel is fantastic for it as well and we have multiple like clips and videos talking about how great Laravel is to use with AI. Taylor um Taylor Artwell the creator of Laravel has talked multiple times about how Laravel is a great framework to use with AI and like Devon said how Laravel is the future of AI and agentic coding. So agree with that and Florian said you can even create MCP apps now with Laravel. Sure can. Still still got to work on the one that I made for Andy that doesn't work yet but feel good. popped up. Yeah. Well, what I found out is the uh the way that I have it authenticated doesn't work with the way that we need to do things yet. So, someday soon. For now, just keep using my really cool Slackbot. the future is here. Chip, I cannot wait to meet those two beautiful dogs. I was just gonna say I love Golden Looks like. Yeah, Chip will be on stream. Hold on. I gotta double check. Discard this when Oh, Chip will be on stream next week. Next Thursday. So, maybe we can get Chip to show off his dogs on stream. Yeah. All right. He said, "Come on over. It's boating season." Chip's one of the new team members I'm excited to meet at Larcon. Oh, you haven't met Chip? I have not met Chip yet. I did not get invited out to marketing's outing. Yeah, I met Chip. I I go to so many I forget who's there or not. That happened to me in uh New York when Abby came down and said hi to me and then went over to Luigi. Like, oh, who are you again? To meet him for the first time. I was like, you don't know Luigi? I was like, wait a second. That's right. I go to a lot of things. I see Abby a lot. I like said hi to Luigi. I hugged Luigi and he's like, "Why not?" And I was like, "Wait, we haven't met before." And he's like, "No." So, my bad. Yeah, Floren, there's going to be a lot of new people to meet at Larcon. Yes. So excited about it. And then we have a question. What's the topic for today? So, we're actually about to start wrapping up now. Um, but today was general cloud office hours. So, we answered questions people had about Laravel Cloud as well as uh general questions about Laravel and about AI. But we also had Andy demo Arbback which is coming in the following days for Laravel Cloud. Um so he demoed Arbback and then he also showed off a secret feature which is um being able to scope your API tokens as well and have them tied to like specific roles. So if you don't if you want to have an API token but you don't want it to be able to touch the database, you just want to be able to create environments or stuff like that, being able to do that and scope the API tokens to specific things. So that was the topic. Um Andy, do you want to give like a summary again or any like closing words of what we talked about today too? Yeah, so rolebased access control coming in days. Um you'll see starter plan will keep your developer and admin roles. Our growth plan will see five predefined roles um that gives you various levels of control over how people can access your uh organization settings. And then for our business and enterprise customers, we'll have the advanced arbback which is gives you custom roles and then app and environment level permissions. And then yeah, scoped API tokens should be coming out uh shortly after the role-based access control feature. So, a lot of good ways to tighten up your org app and environment settings and empower your AI agents to uh help you do more deployments with a little bit more confidence. It's not going to take any negative action without your awareness. So, lots of good things coming for ROS, access, permissions, and agent enablement. I love that. And where should people go if they have any questions or um say if they even want to have a new region added to cloud, if they see a bug, any of that, where should people go? I think all of us are on the Discord. That's always a good spot. Um my Twitter handle is up here on the screen. you can send me a message or if you want to get in touch uh just ping support if you want to try our back out for example maybe a day ahead of time let me know uh we can get it attached to your account yep when in doubt open a support ticket there's a give feedback thing there uh that you can use for asking questions it's a great way for us to track those we can tie those requests directly to the projects themselves so we can uh use that as we're trying to prioritize what's the next features we're looking to roll out. So, definitely do that. Uh, and then anything else, yeah, reach out. Yeah. And I definitely say if you have a bug or anything to open a support ticket and then if like you're waiting and it's something urgent, you need someone to see it, like you can feel free to go in the Discord and like say the issue and link the ticket because then I can help with that as well. But definitely always go to opening a support ticket. That is the best way for our team to track things and keep, like I said, track of all of that. also just like see what team members have seen it, what's been tried. It's just easier if you go through as well. Then we can route it to the right person. Yep. And I have also pinned Andy's Twitter, so you can go there as well. And you can find our Discord through our links on larbell.com as well. And yeah, Devon, do you have any closing words and where people can find you online? Yeah, it's been great hanging out per usual. Can't wait for uh the new schedule here on Tuesdays. Lee and I will be out here with various guests or just ourselves sometimes. And uh if you want to find me on X, it is Dgarbs51_. I'm glad you said it because I was trying to find you and now if I just look up Devon, I cannot find you. Um let's see. So let me have you. Here is Devon's. Like you said, we will now be here on Tuesdays. So every other Tuesday, so you won't see us next Tuesday. you will the following one, which I believe is May 19th. Um, so you'll see us on May 19th at 12:00 p.m. to 1:00 p.m. Eastern time. Again, we'll actually have special guests Andy and Calvin for that one as well with more exciting news. So, make sure you stay tuned for that. And for my other streams this week, we have a stream this Thursday at 11:00 a.m. to 12:30 Mountain time. So that would be 1:00 p.m. to 2:30 Eastern time. That we will be on Thursday. It'll be me, Josh Siri, and Jake. And we will be reviewing Jake's code that he generated using AI. He is not a traditional developer, but he was using AI to create a Larva application. And we will be reviewing his code. So stay tuned to that uh which will be Thursday. And thank you all for being here. Thank you, Andy. And bye everyone.