[{"data":1,"prerenderedAt":61},["ShallowReactive",2],{"leap-week-ui-era-shifting":3,"leap-week-ui-era-shifting-next":46},{"id":4,"slug":5,"vimeo_id":6,"description":7,"tile":8,"length":9,"resources":10,"people":10,"episode_number":11,"published":12,"title":13,"video_transcript_html":14,"video_transcript_text":15,"content":10,"status":16,"episode_people":17,"recommendations":29,"season":30,"seo":45},"bbaa3063-6fbe-4d96-bbc7-e50672f9a308","ui-era-shifting","1176272144","Hear from Directus CEO & Co-Founder Ben Haynes on software is shifting from UI-first to data-first. When everyone can build an interface, what is actually valuable?\n\n","08f64249-c879-4294-92f1-b73a3ea82750",14,null,1,"2026-03-27","The UI Era is Shifting","\u003Cp>Speaker 0: So let's kick things off. This is Ben, founder and CEO at Directus. And today, I need to talk to you about the end of the UI era. A little sensational, but let's dive in so that I can explain. First off, you might be thinking, don't we have better tools, more people than ever building all these front ends, all these UIs?\u003C\u002Fp>\u003Cp>Yes, absolutely. But when everything is special, nothing is. The front ends have been, commoditized. So let's explore what actually makes software magic, and specifically where the value of software has shifted to and where it's heading next. So since software was created, the interface has been mistaken for the product itself.\u003C\u002Fp>\u003Cp>Users click through, buttons and interfaces and they said, Oh, this is the magic. They saw a dashboard and they said, Oh, that's the application. The Shiny part got the attention for most users. But engineers, you know, for those who are out there, you know better. The visible part of the software is the tip of the iceberg.\u003C\u002Fp>\u003Cp>What lives underneath is the expensive, the fragile part. The part that takes years, decades, to get right. AI has made this a really big deal. Anyone can spin up a decent looking UI, a form, with solid UX, dashboards, components, all of that. They can stitch it together and you can get something that looks just like software.\u003C\u002Fp>\u003Cp>That's amazing. But it's also misleading and maybe even a little dangerous. It creates the illusion that software got easy, but it did not. The front end got easier. The back end maybe even got harder.\u003C\u002Fp>\u003Cp>We have so many builders out there. The back end is the tricky part. And the reason for that is you don't know what you don't know. So while most non engineers can definitely dream up an interface, you are a non technical person. You probably don't know what is needed behind the scenes to actually build this out, make it scalable, make it resilient.\u003C\u002Fp>\u003Cp>So you can Vibe code a UI in a day, but you cannot Vibe code trust. So maybe we just delete the backend. Cursor a little while back actually deleted their headless CMS. As an example. They replaced it in a couple days for a couple $100 worth of tokens.\u003C\u002Fp>\u003Cp>And that really works for a company like theirs. A small team, highly technical, very capable. But that's not going to work for other organizations. You can delete the CMS, but you can't delete the problems that the CMS solves. They come back and when they do, you're either gonna be building from scratch or you're gonna be drowning.\u003C\u002Fp>\u003Cp>So switching gears, let's talk about the back end. You can't one shot your way into production resilience. You can't fake your way into scalable architecture, granular RBAC or access control, reliable APIs, multi tenancy observability, governance rollback strategy, operational maturity. All of this is earned the hard way. Cursor, Lovable, Replit, Bolt and others, they've all made it really easy to create a front end in minutes.\u003C\u002Fp>\u003Cp>And it works. But without a back end, really what you've created is a clickable prototype. The front end and back end distinction gets really blurry because of the excitement around, you know, the generation of this of this front end. And I I love metaphors, so I'll kinda take a a second here. You wouldn't build a house starting with the doorbell, or something like that.\u003C\u002Fp>\u003Cp>You know, the the fresh paint, the beautiful fixtures, the interior design, none of that matters if you've got plumbing, leaking behind the walls, if you've got electricals, you know, sparking fires behind the sheetrock, a foundation that's not even poured and the house is just sinking into the mud. That's just not how things work. Anyone can build these demos, these facades, but far fewer people can build something that actually survives success. You just you can't vibe code trust. So if the front end is becoming free, where does the value go?\u003C\u002Fp>\u003Cp>It goes deeper but not just deeper technically, deeper organizationally. The thing that makes software software wasn't really the pixels. It's not the front end. It's the layer underneath where your whole organization goes, to work together. It's the logic, the APIs, the permissions, the 99.999%, uptime.\u003C\u002Fp>\u003Cp>Your performance under load. You know, does that work? Auditability. Maintainability. The stability that people don't notice until things break.\u003C\u002Fp>\u003Cp>It's not what you see, it's what you feel. You feel when your app goes down. When the data is suddenly corrupted or truncated unexpectedly or deleted. Permissions break. Traffic spikes and things just fall apart.\u003C\u002Fp>\u003Cp>The front end may be what people admire, but the back end again is what people trust. That's what really matters here. So let's first clarify what do we mean by back end. There's the Supabase Lovable partnership and people would refer to Supabase as the backend. It's a database.\u003C\u002Fp>\u003Cp>It's got infrastructure, an API. But really the backend is more than that. The backend is a collaboration layer. It's where technical and non technical users can go to actually access, to browse, manage, and visualize data. It's where data is governed, and used.\u003C\u002Fp>\u003Cp>Not just by engineers, but by anybody in the team. AI accelerates what individuals can do. Our own engineering team uses Cloud Code every day and the velocity gains are very much real. But more complex use cases require humans working together across teams. Engineers with PMs and ops, marketing, legal.\u003C\u002Fp>\u003Cp>AI doesn't solve the collaboration problem. It actually makes it larger. The winners won't be those that can create the prettiest screen the fastest, but those that combine the speed of AI built interfaces with the true discipline of back end engineering. And then democratizing that across the entire organization, technical users and non technical. That's how a hobby project becomes an actual product, or in some cases, how a product becomes an actual company.\u003C\u002Fp>\u003Cp>Okay. So let's break down the back end into three pillars. The first one we'll call governance, how you collaborate safely. The data is there. It's in the database behind every application, that's that's out there.\u003C\u002Fp>\u003Cp>But it's locked behind the doors of IT, as it were, for every report, every change of the data that's needed. If you're non technical, you're probably submitting a support ticket. We need tools that unlock the database for the technical users. Technical users are about 3% of an org, typically. And so we need something like a database administration tool but that's simple, safe, and intuitive enough for anyone to use.\u003C\u002Fp>\u003Cp>It's not about locking things down. It's about opening them up, responsibly. How many one shot AI applications are actually considering RBAC and granular permissions and all that in their system? I would guess not many. So we need to browse, manage, and visualize our data, in an intuitive way.\u003C\u002Fp>\u003Cp>We also need docs, API references, specifications. Those are needed as your team grows, more people join, you need integrations with other services. Those are boilerplate. Let the subject matter experts do their work. We all have seen, you know, a game of telephone and how that can degrade things.\u003C\u002Fp>\u003Cp>You can't pass these requests along. You need people to do that work directly. And so we need interfaces that work for everyone. The second one is scale, or how we take that collaboration and we make it survive the complexity, of the real world. You could also consider that to be production efficiency, resilience.\u003C\u002Fp>\u003Cp>There are a lot of things to consider, consider here. Anyone can build a v one. The problem is when you get to v two, v three, v four, v n, when you go from beta testers out to actual users, in the real world, you bring more people onto your team, internal users. Eventually, you can't just throw hardware at the problem to get by. You actually need to have a proper back end that supports growing your application and whatever it is that you're building.\u003C\u002Fp>\u003Cp>And the third pillar is ownership, or taking all of this and making sure you own what you're building. Again, the value of what you're building is not the front end. That's been commoditized. It's not your cloud infrastructure, your hosting, that's a utility bill, and it's not even really your data that can be scraped, pretty easily these days. So the value is your back end.\u003C\u002Fp>\u003Cp>It's the architecture, the logic and your team, that's, you know, back there moving faster through this collaboration. It's the trust that you create, again, through scale and governance and all of that. It's the value that you unlock from your data with all of this. So you want to pull these systems closer. That's your fence.\u003C\u002Fp>\u003Cp>That's your moat. Self hosting, data sovereignty, open systems. This is where you take control back. So these are the three things that are needed for a proper Backend. But even for the modern organizations that check off all of these boxes, they're still seeing ten, twenty, 30 plus, of these back ends, you know, all over the place.\u003C\u002Fp>\u003Cp>Disparate systems, each with their own API. Maybe it's REST or GraphQL, webhooks. You know, they're all different integrations, different interfaces, different logins. They don't talk to each other. Data is siloed, and so is collaboration.\u003C\u002Fp>\u003Cp>So we're looking at something that has no federation, no unified layer, no consistency and no strategy. And for AI apps, that's even worse. They typically don't even have a backend at all, you know? So here we are. The backend era has begun.\u003C\u002Fp>\u003Cp>So what do we do now? What do we need? We need a unified layer, a single pane of glass that you can put across these disparate systems and bring them together, leaving the data in situ, leaving the data where it is, but federating it and bringing it together. We need to wrap all of that with the granular access control, the rule based permissions, that allow humans and other integrations and agents all alike to have CRUD, to have create, read, update, and delete control, so they're not deleting the data so that you gain more of that trust. You need to bake the tools into this platform that give it the scale and resilience that we were talking about.\u003C\u002Fp>\u003Cp>Not just the API throughput and the data pipelines, but also scaling your team. So again, you need to be able to have everyone working together. And you need to make this a model that is completely yours. So deploying where you need to deploy, whether that's cloud or on prem, you wanna make sure that you are working with an open core, that it's source available. This is the value.\u003C\u002Fp>\u003Cp>So you want to keep all of this really safe. This is actually what we've been building at Directus, and it works. We have 45,000,000 Docker downloads, 35,000 stars on GitHub, You know, over a thousand customers. It started back in 2004, twenty two years ago. My agency in Brooklyn.\u003C\u002Fp>\u003Cp>We were building for Google, Prada, Snapchat, SoulCycled US government, AT and T. And behind every project that we built, there was a database. And we made something for that database that was simple, safe, and intuitive enough for the business user, this collaborative back end. Back then, my team was human, obviously. And we were creating those innovative front ends and now here we are two decades later and it's even more important.\u003C\u002Fp>\u003Cp>Everyone's a builder. Everyone has these interfaces that they're generating. Alright. So I will leave you with this one last bit, for today. Our Directus platform powers production applications for very large brands, around the world.\u003C\u002Fp>\u003Cp>One of these global pizza delivery app, brands is actually handling 9,000 requests per second, sustained. This is not something that you just one shot your way into. You're not going to vibe code your way, in, you know, the hour, the day, the week for these mission critical production scale applications. The front end may be what most people admire this exciting, you know, glamorous thing that we're seeing, very, very recently come out of all these AI app building tools. And it's phenomenal, but the backend is what they trust.\u003C\u002Fp>\u003Cp>So invest accordingly. From me and everyone at Directus, thank you so much and happy building.\u003C\u002Fp>","So let's kick things off. This is Ben, founder and CEO at Directus. And today, I need to talk to you about the end of the UI era. A little sensational, but let's dive in so that I can explain. First off, you might be thinking, don't we have better tools, more people than ever building all these front ends, all these UIs? Yes, absolutely. But when everything is special, nothing is. The front ends have been, commoditized. So let's explore what actually makes software magic, and specifically where the value of software has shifted to and where it's heading next. So since software was created, the interface has been mistaken for the product itself. Users click through, buttons and interfaces and they said, Oh, this is the magic. They saw a dashboard and they said, Oh, that's the application. The Shiny part got the attention for most users. But engineers, you know, for those who are out there, you know better. The visible part of the software is the tip of the iceberg. What lives underneath is the expensive, the fragile part. The part that takes years, decades, to get right. AI has made this a really big deal. Anyone can spin up a decent looking UI, a form, with solid UX, dashboards, components, all of that. They can stitch it together and you can get something that looks just like software. That's amazing. But it's also misleading and maybe even a little dangerous. It creates the illusion that software got easy, but it did not. The front end got easier. The back end maybe even got harder. We have so many builders out there. The back end is the tricky part. And the reason for that is you don't know what you don't know. So while most non engineers can definitely dream up an interface, you are a non technical person. You probably don't know what is needed behind the scenes to actually build this out, make it scalable, make it resilient. So you can Vibe code a UI in a day, but you cannot Vibe code trust. So maybe we just delete the backend. Cursor a little while back actually deleted their headless CMS. As an example. They replaced it in a couple days for a couple $100 worth of tokens. And that really works for a company like theirs. A small team, highly technical, very capable. But that's not going to work for other organizations. You can delete the CMS, but you can't delete the problems that the CMS solves. They come back and when they do, you're either gonna be building from scratch or you're gonna be drowning. So switching gears, let's talk about the back end. You can't one shot your way into production resilience. You can't fake your way into scalable architecture, granular RBAC or access control, reliable APIs, multi tenancy observability, governance rollback strategy, operational maturity. All of this is earned the hard way. Cursor, Lovable, Replit, Bolt and others, they've all made it really easy to create a front end in minutes. And it works. But without a back end, really what you've created is a clickable prototype. The front end and back end distinction gets really blurry because of the excitement around, you know, the generation of this of this front end. And I I love metaphors, so I'll kinda take a a second here. You wouldn't build a house starting with the doorbell, or something like that. You know, the the fresh paint, the beautiful fixtures, the interior design, none of that matters if you've got plumbing, leaking behind the walls, if you've got electricals, you know, sparking fires behind the sheetrock, a foundation that's not even poured and the house is just sinking into the mud. That's just not how things work. Anyone can build these demos, these facades, but far fewer people can build something that actually survives success. You just you can't vibe code trust. So if the front end is becoming free, where does the value go? It goes deeper but not just deeper technically, deeper organizationally. The thing that makes software software wasn't really the pixels. It's not the front end. It's the layer underneath where your whole organization goes, to work together. It's the logic, the APIs, the permissions, the 99.999%, uptime. Your performance under load. You know, does that work? Auditability. Maintainability. The stability that people don't notice until things break. It's not what you see, it's what you feel. You feel when your app goes down. When the data is suddenly corrupted or truncated unexpectedly or deleted. Permissions break. Traffic spikes and things just fall apart. The front end may be what people admire, but the back end again is what people trust. That's what really matters here. So let's first clarify what do we mean by back end. There's the Supabase Lovable partnership and people would refer to Supabase as the backend. It's a database. It's got infrastructure, an API. But really the backend is more than that. The backend is a collaboration layer. It's where technical and non technical users can go to actually access, to browse, manage, and visualize data. It's where data is governed, and used. Not just by engineers, but by anybody in the team. AI accelerates what individuals can do. Our own engineering team uses Cloud Code every day and the velocity gains are very much real. But more complex use cases require humans working together across teams. Engineers with PMs and ops, marketing, legal. AI doesn't solve the collaboration problem. It actually makes it larger. The winners won't be those that can create the prettiest screen the fastest, but those that combine the speed of AI built interfaces with the true discipline of back end engineering. And then democratizing that across the entire organization, technical users and non technical. That's how a hobby project becomes an actual product, or in some cases, how a product becomes an actual company. Okay. So let's break down the back end into three pillars. The first one we'll call governance, how you collaborate safely. The data is there. It's in the database behind every application, that's that's out there. But it's locked behind the doors of IT, as it were, for every report, every change of the data that's needed. If you're non technical, you're probably submitting a support ticket. We need tools that unlock the database for the technical users. Technical users are about 3% of an org, typically. And so we need something like a database administration tool but that's simple, safe, and intuitive enough for anyone to use. It's not about locking things down. It's about opening them up, responsibly. How many one shot AI applications are actually considering RBAC and granular permissions and all that in their system? I would guess not many. So we need to browse, manage, and visualize our data, in an intuitive way. We also need docs, API references, specifications. Those are needed as your team grows, more people join, you need integrations with other services. Those are boilerplate. Let the subject matter experts do their work. We all have seen, you know, a game of telephone and how that can degrade things. You can't pass these requests along. You need people to do that work directly. And so we need interfaces that work for everyone. The second one is scale, or how we take that collaboration and we make it survive the complexity, of the real world. You could also consider that to be production efficiency, resilience. There are a lot of things to consider, consider here. Anyone can build a v one. The problem is when you get to v two, v three, v four, v n, when you go from beta testers out to actual users, in the real world, you bring more people onto your team, internal users. Eventually, you can't just throw hardware at the problem to get by. You actually need to have a proper back end that supports growing your application and whatever it is that you're building. And the third pillar is ownership, or taking all of this and making sure you own what you're building. Again, the value of what you're building is not the front end. That's been commoditized. It's not your cloud infrastructure, your hosting, that's a utility bill, and it's not even really your data that can be scraped, pretty easily these days. So the value is your back end. It's the architecture, the logic and your team, that's, you know, back there moving faster through this collaboration. It's the trust that you create, again, through scale and governance and all of that. It's the value that you unlock from your data with all of this. So you want to pull these systems closer. That's your fence. That's your moat. Self hosting, data sovereignty, open systems. This is where you take control back. So these are the three things that are needed for a proper Backend. But even for the modern organizations that check off all of these boxes, they're still seeing ten, twenty, 30 plus, of these back ends, you know, all over the place. Disparate systems, each with their own API. Maybe it's REST or GraphQL, webhooks. You know, they're all different integrations, different interfaces, different logins. They don't talk to each other. Data is siloed, and so is collaboration. So we're looking at something that has no federation, no unified layer, no consistency and no strategy. And for AI apps, that's even worse. They typically don't even have a backend at all, you know? So here we are. The backend era has begun. So what do we do now? What do we need? We need a unified layer, a single pane of glass that you can put across these disparate systems and bring them together, leaving the data in situ, leaving the data where it is, but federating it and bringing it together. We need to wrap all of that with the granular access control, the rule based permissions, that allow humans and other integrations and agents all alike to have CRUD, to have create, read, update, and delete control, so they're not deleting the data so that you gain more of that trust. You need to bake the tools into this platform that give it the scale and resilience that we were talking about. Not just the API throughput and the data pipelines, but also scaling your team. So again, you need to be able to have everyone working together. And you need to make this a model that is completely yours. So deploying where you need to deploy, whether that's cloud or on prem, you wanna make sure that you are working with an open core, that it's source available. This is the value. So you want to keep all of this really safe. This is actually what we've been building at Directus, and it works. We have 45,000,000 Docker downloads, 35,000 stars on GitHub, You know, over a thousand customers. It started back in 2004, twenty two years ago. My agency in Brooklyn. We were building for Google, Prada, Snapchat, SoulCycled US government, AT and T. And behind every project that we built, there was a database. And we made something for that database that was simple, safe, and intuitive enough for the business user, this collaborative back end. Back then, my team was human, obviously. And we were creating those innovative front ends and now here we are two decades later and it's even more important. Everyone's a builder. Everyone has these interfaces that they're generating. Alright. So I will leave you with this one last bit, for today. Our Directus platform powers production applications for very large brands, around the world. One of these global pizza delivery app, brands is actually handling 9,000 requests per second, sustained. This is not something that you just one shot your way into. You're not going to vibe code your way, in, you know, the hour, the day, the week for these mission critical production scale applications. The front end may be what most people admire this exciting, you know, glamorous thing that we're seeing, very, very recently come out of all these AI app building tools. And it's phenomenal, but the backend is what they trust. So invest accordingly. From me and everyone at Directus, thank you so much and happy building.","published",[18],{"people_id":19},{"id":20,"first_name":21,"last_name":22,"avatar":23,"bio":24,"links":25},"4f76549a-cebc-439b-98da-cc0568e708a9","Ben","Haynes","8d3aa9f0-f76f-4c77-82fe-0bab6d40ef41","CEO at Directus",[26],{"url":27,"service":28},"https:\u002F\u002Fdirectus.io\u002Fteam\u002Fben-haynes","website",[],{"id":31,"number":32,"year":33,"episodes":34,"show":42},"289f6534-7fdd-46df-8c00-89a75469fe41",4,"2026",[4,35,36,37,38,39,40,41],"f885409e-0ace-41e5-aca3-faf4dcd7659b","7271f0be-33fd-4cea-b7bd-9c63e74969e1","0baede33-974c-4343-abad-3cea928c8112","a10f99c3-6b45-46e6-b703-64366f150c57","1310befc-e361-4e19-848f-d685c19dddef","37e28ea2-bec3-40bd-8b09-b9fbb47c1759","68536266-9502-4df8-a295-ef082dfe6fd0",{"title":43,"tile":44},"Leap Week","62816023-fa7e-4a76-b9a1-2733ee2093a6",{"title":10,"meta_description":10},{"id":35,"slug":47,"season":31,"vimeo_id":48,"description":49,"tile":50,"length":51,"resources":10,"people":10,"episode_number":52,"published":12,"title":53,"video_transcript_html":54,"video_transcript_text":55,"content":10,"seo":10,"status":16,"episode_people":56,"recommendations":60},"the-build-off","1176279072","Three builders. Three AI tools. One backend. Who can ship the best frontend in 30 minutes?","4059a3df-317a-4665-9a4a-9f5f2548ca07",47,2,"The Build-Off","\u003Cp>Speaker 0: What's up, everybody? It is great to see you, to meet you virtually here. My name is Matt. I'm on the marketing team here at Directus. And if you're watching this, you're in for a treat because we have quite the repertoire of guests here today, also from the Directus team, and we're gonna be doing something pretty fun.\u003C\u002Fp>\u003Cp>But first, let me introduce each guest. You know him. You love him. You've seen his face all over the place. It's John Daniels.\u003C\u002Fp>\u003Cp>Speaker 1: John Daniels.\u003C\u002Fp>\u003Cp>Speaker 2: What's up? Glad that you\u003C\u002Fp>\u003Cp>Speaker 0: started with me.\u003C\u002Fp>\u003Cp>Speaker 2: I was worried you were gonna say Brian. So hello.\u003C\u002Fp>\u003Cp>Speaker 1: Your face is all over the place right here.\u003C\u002Fp>\u003Cp>Speaker 0: I don't think anybody really knows who Bryant is. So, no. No. No. No.\u003C\u002Fp>\u003Cp>Of course, we have the man himself, mister Bry. Bryant Gillespie, who's one of our esteemed developer advocate type folks, I guess, on the team. You know? He kinda does everything.\u003C\u002Fp>\u003Cp>Speaker 1: Type folks? Yes. Happy to be here. Very excited for this. Look.\u003C\u002Fp>\u003Cp>I'm just gonna start this off with saying whatever we're doing in this competition, I'm gonna mop the floor with you guys. That's it.\u003C\u002Fp>\u003Cp>Speaker 0: Strong start. I was gonna do this not in the spirit of competition, but in Oh. In in friendliness, but I guess that's what we're gonna do now. Alright. And if anybody doesn't know, John is one of our AEs here at Directus, so the most competitive of of everyone here.\u003C\u002Fp>\u003Cp>So I'm sure pretty sure that'll awaken a fire in a moment. So, yeah. Brian, what are we doing today? I\u003C\u002Fp>\u003Cp>Speaker 1: this is your show, my friend. You tell me what we're doing. Clearly, it's something to do with an AI skills directory because that is what's on the screen.\u003C\u002Fp>\u003Cp>Speaker 0: That is what's on the screen. Alright. I'll give you the brief rundown. So today, we are gonna have a build off. Everybody's been talking about front ends and UIs, and the errors are shifting.\u003C\u002Fp>\u003Cp>And there's so many tools out there. I'm a user of Cloud Code for a lot of the marketing stuff that we do here at Directus, including building the Leap Week landing page. Our team uses codex. Our team uses a lot of the out of the box, like, harnesses, like cursor. And we also use things like Lovable.\u003C\u002Fp>\u003Cp>So it's a good mix of things, but that's a lot of subscriptions add up. So we thought, why not put the skills to the test and see which we can use best for building the front end on a Directus back end, which would be a skills directory. And one of the reasons we wanna do a skills directory is because there's a lot of shady skills directories out there and there's a lot of mistrust in a lot of them. So our developer team has a lot of, skills that they share, that we thought would be cool to build something along with that, as well as our go to market team uses a lot of things too. So, yeah, that's really the gist.\u003C\u002Fp>\u003Cp>Did I\u003C\u002Fp>\u003Cp>Speaker 1: What are the what are the stipulations for this, man? We're building a skills directory front end. We're gonna need a back end for that, which we'll do, but what are the rules?\u003C\u002Fp>\u003Cp>Speaker 0: There are no rules. There's only one rule. And that is\u003C\u002Fp>\u003Cp>Speaker 1: that's details?\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. There's gonna be one rule where we have to create the back end and direct us. And then whatever tool we're gonna use, we are going to use the same back end for that front end. So running down the list of tools that we will use, I will be using Lovable. It's something I've used in the past.\u003C\u002Fp>\u003Cp>I've I've not actually used it in quite a bit, because I've been using, Cloud Code. But, Brian, what are you gonna be using?\u003C\u002Fp>\u003Cp>Speaker 1: I'm gonna be using Cloud Code.\u003C\u002Fp>\u003Cp>Speaker 0: Alright. John, what you got?\u003C\u002Fp>\u003Cp>Speaker 2: I use my old dependable cursor.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Is cursor what you built? The, garage inventory?\u003C\u002Fp>\u003Cp>Speaker 2: It is. Yeah. Yeah. And I already have a I already have an idea.\u003C\u002Fp>\u003Cp>Speaker 0: Oh, boy. Oh, no.\u003C\u002Fp>\u003Cp>Speaker 2: Retro. It's a lot yeah.\u003C\u002Fp>\u003Cp>Speaker 1: Front end directory.\u003C\u002Fp>\u003Cp>Speaker 2: It's a long\u003C\u002Fp>\u003Cp>Speaker 1: Here we go.\u003C\u002Fp>\u003Cp>Speaker 0: I'm doing that maybe more than, like, moving forward in my chair.\u003C\u002Fp>\u003Cp>Speaker 1: Cool.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Well So Right. To kick things off, I'm assuming we'll kick we've started.\u003C\u002Fp>\u003Cp>Speaker 1: We we got a totally blank data model here. We'll just call it skills directory. Right? John is just banging on the keyboard, man. Alright.\u003C\u002Fp>\u003Cp>Totally empty. It's like you know, we'll we'll do this 100 app style. Like, what what what do we even need on a skills directory? Right? You're gonna need some skills, but what what is the what are the other stuff that you're planning on tracking here?\u003C\u002Fp>\u003Cp>What do you need for the back end?\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. That's a good question. When you say skills, what does that encompass? Is that like the title and the description and the actual I mean, skills are basically just a fancy word for markdown files. Right?\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. I mean, that's, yeah, that's the the funny thing about all of this stuff to me is it's all just markdown. Like, been been dealing with markdown since I was 12. A week ago. Yeah.\u003C\u002Fp>\u003Cp>So we got a name, a description for the skill. You've got the, what, the actual skill content. I don't I don't\u003C\u002Fp>\u003Cp>Speaker 0: know.\u003C\u002Fp>\u003Cp>Speaker 1: Are you I'm assuming you're gonna have, like, users who submit the skills?\u003C\u002Fp>\u003Cp>Speaker 0: Mhmm. We'll have I think, at some point, if we wanna\u003C\u002Fp>\u003Cp>Speaker 1: Submit a bug.\u003C\u002Fp>\u003Cp>Speaker 0: Public facing, but for our internal team makes sense. You probably want, some sort of, like, tagging system. So if it's engineering skill or if it's\u003C\u002Fp>\u003Cp>Speaker 1: So the tag name. Okay. What else you got?\u003C\u002Fp>\u003Cp>Speaker 0: Option for an image, if you wanna add that to the front end potentially.\u003C\u002Fp>\u003Cp>Speaker 1: Image? Okay. You're getting away from the markdown stuff now. I'm gonna just draw the arrows for fun here\u003C\u002Fp>\u003Cp>Speaker 0: at this point. Here we go.\u003C\u002Fp>\u003Cp>Speaker 1: Boom. Alright. So you got skills. You got tags. What else do you need?\u003C\u002Fp>\u003Cp>Speaker 0: I mean,\u003C\u002Fp>\u003Cp>Speaker 1: I Repository. GitHub repository. Is that where they're coming from?\u003C\u002Fp>\u003Cp>Speaker 0: Could you just host it within Directus itself, like, as a file?\u003C\u002Fp>\u003Cp>Speaker 1: I mean, we can. Yeah. Like, do you want it linked to somewhere?\u003C\u002Fp>\u003Cp>Speaker 2: I think\u003C\u002Fp>\u003Cp>Speaker 0: I wanna keep it, like, self like, contained within the one\u003C\u002Fp>\u003Cp>Speaker 1: thing. Okay. Alright. Cool. Is this anything else?\u003C\u002Fp>\u003Cp>Or is it we're just gonna roll with this.\u003C\u002Fp>\u003Cp>Speaker 0: Just keep it simple. Unless you got anything to do. Wait.\u003C\u002Fp>\u003Cp>Speaker 2: No. Keep it simple, stupid.\u003C\u002Fp>\u003Cp>Speaker 1: Which one is stupid, man?\u003C\u002Fp>\u003Cp>Speaker 2: You probably said it.\u003C\u002Fp>\u003Cp>Speaker 0: Alright.\u003C\u002Fp>\u003Cp>Speaker 1: I've got this. So, hopefully, I've got some secret sauce here. Been working on a CLI tool. So I'm gonna tell Claude code here. Can you guys see this?\u003C\u002Fp>\u003Cp>Should I zoom way in? Alright. Run d six s agents. It's our CLI for connecting to direct us.\u003C\u002Fp>\u003Cp>Speaker 0: D six s. It's the first time seeing the CLI in action right here.\u003C\u002Fp>\u003Cp>Speaker 1: Very nerdy. You know, the Kubernetes deal. What is this thing doing? Why is it not running this command? It's reading four files, which d success.\u003C\u002Fp>\u003Cp>Just run the command, dude. What are you doing?\u003C\u002Fp>\u003Cp>Speaker 2: Chose the wrong tool. Maybe cursor's been\u003C\u002Fp>\u003Cp>Speaker 1: Maybe you chose the wrong tool. Yeah. There it is. Okay. So, basically, that command gives it a little context about this.\u003C\u002Fp>\u003Cp>Cool. And now this thing should be primed to go. Agents. Right. Cool.\u003C\u002Fp>\u003Cp>Alright. So now we're gonna kick it over into plan mode. I'm gonna paste that image that I had. Right? And we're just gonna say, hey.\u003C\u002Fp>\u003Cp>We're building an AI skills directory. Here's a screenshot of a data model I'm competing with John and Matt. We need the back end in Directus first before we build the front end. Can you plan out the Directus back end?\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Alright.\u003C\u002Fp>\u003Cp>Speaker 1: I've been using Let's see what happens here. Okay. Right here. Alright. It is going to alright.\u003C\u002Fp>\u003Cp>This is where I tell you to do as I say, not as I do, because I'm just gonna turn this thing loose. Right? I should have actually set dangerously set permissions here. But\u003C\u002Fp>\u003Cp>Speaker 2: I don't even have it ask for permission anymore. I just let it go.\u003C\u002Fp>\u003Cp>Speaker 1: Go, baby. Go.\u003C\u002Fp>\u003Cp>Speaker 0: I have to leave this a lot. That's probably why.\u003C\u002Fp>\u003Cp>Speaker 1: If you can't leave that out, security infraction.\u003C\u002Fp>\u003Cp>Speaker 2: Security. Spent hundreds of dollars in tokens last month.\u003C\u002Fp>\u003Cp>Speaker 1: Alright. It's going to write this plan. Are you so are are you not using cloud code at all? You're still using cursor every day?\u003C\u002Fp>\u003Cp>Speaker 2: I use both. Depends on what I'm doing.\u003C\u002Fp>\u003Cp>Speaker 0: So what so what you're you're basically just doing what the MCP used to do just from the terminal\u003C\u002Fp>\u003Cp>Speaker 1: So is the why the CLI? Right? Well, the reason why is because context. The context window. So MCP uses up a lot of context, for comparison.\u003C\u002Fp>\u003Cp>Right? Like, if you have all the tools enabled, MCP or our AI assistant right now could use up to thirty, forty thousand tokens to have all the tools enabled. The CLI, by default, doesn't load anything into the context, and the agent is or like CloudCo, these agent harnesses are pretty good at navigating, like, CLIs just because they're they're trained on a lot of bash scripts and stuff like that. So, you know, it basically you keep the context window clean until you need it, and, yeah, you follow this progressive discovery pattern. I'm yeah.\u003C\u002Fp>\u003Cp>I've been talking. I'm not even looking at this plan. Hopefully, this thing actually does what we wanted to do. We shall see. Right?\u003C\u002Fp>\u003Cp>Actually, use recipes. D success recipes. This is should be a shorthand. Again, this is all highly experimental, so failed to log the plug in, so it's not gonna use that at all. Plug ins.\u003C\u002Fp>\u003Cp>Index. Oh, good. Alright. Well, just push forward. We may be kicking back to the MCP for this.\u003C\u002Fp>\u003Cp>Speaker 0: Hey.\u003C\u002Fp>\u003Cp>Speaker 2: Alright. Is what I when I type in the when it when MCP goes wrong for me, I say, just do it, and it'll it'll figure it out.\u003C\u002Fp>\u003Cp>Speaker 1: Just do it. Figure it out funny, man.\u003C\u002Fp>\u003Cp>Speaker 0: Turn off permissions and just go.\u003C\u002Fp>\u003Cp>Speaker 1: What is this thing doing? Your token doesn't have admin access. I do have admin access.\u003C\u002Fp>\u003Cp>Speaker 0: No. You're wrong. You said, highly experimental. Correct?\u003C\u002Fp>\u003Cp>Speaker 1: I mean, it it was working well. Works on my machine until you start recording. Right? Right. So we'll just go back to the old tried and true here.\u003C\u002Fp>\u003Cp>Both collections exist, but I have no schema. Oh my gosh. This thing is hopelessly lost.\u003C\u002Fp>\u003Cp>Speaker 2: Lost. Hey. There's two collections in there now.\u003C\u002Fp>\u003Cp>Speaker 1: Yep. They're actually just folder collections because they don't exist. Alright. Skills. When I say tried and true, I mean me.\u003C\u002Fp>\u003Cp>Alright. What do we have there? We have name, description, content. Name.\u003C\u002Fp>\u003Cp>Speaker 0: I've never seen anyone move this fast inside of the.\u003C\u002Fp>\u003Cp>Speaker 1: Content. Just markdown. It's just markdown. There you go. Alright.\u003C\u002Fp>\u003Cp>You got tags. Yeah?\u003C\u002Fp>\u003Cp>Speaker 0: Mhmm.\u003C\u002Fp>\u003Cp>Speaker 1: Cool. Tags. I have a name for the tags. Tag name. Cool.\u003C\u002Fp>\u003Cp>And then we need a skill tags. We need a junction table. Cool. I'm taking the long way around here. Skills.\u003C\u002Fp>\u003Cp>Now I'm gonna go to the mini mini relationship inside directions. It's been a minute since I've done this, man. Don't you miss it? Always I do miss it. Yeah.\u003C\u002Fp>\u003Cp>It took me an hour to do half of this stuff that I could now do in, like, five minutes using the AI tools when they work, mind you. Skills ID, tags ID. Let's just call it skill. Let's call it tag. Skills.\u003C\u002Fp>\u003Cp>Cool. Give it a sort field. Boom. Boom. Boom.\u003C\u002Fp>\u003Cp>Speaker 0: Boom. You know, it's worth that messing up to watch you do this. It's like watching Boom.\u003C\u002Fp>\u003Cp>Speaker 1: That's it. There's the back end. You're good. Golden. Very simple.\u003C\u002Fp>\u003Cp>Very easy. Now what?\u003C\u002Fp>\u003Cp>Speaker 2: Now we need some data.\u003C\u002Fp>\u003Cp>Speaker 1: Mhmm. Maybe it can do that. Look, I I set it all up. Nice. Thanks for nothing.\u003C\u002Fp>\u003Cp>Now I need you to create some sample data for us, though. I mean, I'm a make sure you rub it in the face of the thing. Right? Talk to me well, I this might take a minute, but,\u003C\u002Fp>\u003Cp>Speaker 2: like, what are you guys Like, when you\u003C\u002Fp>\u003Cp>Speaker 1: are are we supposed to talk about what we're gonna build? Like, how the front end is gonna look, or is this gonna be, like, a secret thing?\u003C\u002Fp>\u003Cp>Speaker 0: We'll talk about, you know, how things are going. I'm still I'm between a couple ideas on what I want this front end to look like. I I want it to not look like a generic directory because there's, like, directories of directories out there. I already\u003C\u002Fp>\u003Cp>Speaker 2: have an idea, so don't even don't even worry about Wait.\u003C\u002Fp>\u003Cp>Speaker 0: Alright. Was it one that we talked about earlier, or is this one this is completely new?\u003C\u002Fp>\u003Cp>Speaker 2: I got a new idea.\u003C\u002Fp>\u003Cp>Speaker 0: Interesting. Okay. I think I'll have an idea.\u003C\u002Fp>\u003Cp>Speaker 1: You think you'll have an idea? Alright. Well, it's got there's eight tags. There's six skills. Let's add 10 more skills.\u003C\u002Fp>\u003Cp>Create a few sample users as well, give them a different role. Just the ability to view skills and update the ones they've uploaded. Alright. So now we're gonna move to the the front end stuff. Yeah?\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. I think so. So we're all operating off of the same direct us back end, and you just created a few users. I'm assuming that's to give us access to that data.\u003C\u002Fp>\u003Cp>Speaker 1: Yep. So you guys have your own users in there. Looks like it's created\u003C\u002Fp>\u003Cp>Speaker 2: Only one guy cares about a photo.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Oh my gosh. Alright, dude. Fine. Fine.\u003C\u002Fp>\u003Cp>Fine. Fine.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Boom. I'll update my You\u003C\u002Fp>\u003Cp>Speaker 1: look you look rather young in that one,\u003C\u002Fp>\u003Cp>Speaker 2: John. Thanks.\u003C\u002Fp>\u003Cp>Speaker 1: Alright. So what do I dig at all.\u003C\u002Fp>\u003Cp>Speaker 0: Thank you. Alright. So each of us is gonna have our own role, and you will provide that role\u003C\u002Fp>\u003Cp>Speaker 1: to us.\u003C\u002Fp>\u003Cp>Speaker 2: If you refresh real quick. Sorry. Refresh bright. There.\u003C\u002Fp>\u003Cp>Speaker 0: Of course.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. We're reading we're watching this. You're certainly not watching this on Saint Patrick's Day. We're recording it on Saint Patrick's Day, though.\u003C\u002Fp>\u003Cp>Speaker 2: Yeah. Yeah. Yeah. Well done.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. That one.\u003C\u002Fp>\u003Cp>Speaker 2: Sorry. Continue, Matt.\u003C\u002Fp>\u003Cp>Speaker 0: How do you have access to this already? I\u003C\u002Fp>\u003Cp>Speaker 1: I already gave him access. I put it in the chat. I sent you guys up already. Your direct us email and your password. What's the password?\u003C\u002Fp>\u003Cp>The password is Password. Now, hopefully, one of us will remember to have removed that by the time this recording goes live.\u003C\u002Fp>\u003Cp>Speaker 2: I'll tell him how to change\u003C\u002Fp>\u003Cp>Speaker 0: it. In the chat. Must be what is this\u003C\u002Fp>\u003Cp>Speaker 1: thing doing. Okay. I was creating, like, 35. Why can't I not just accept all here? Good lord.\u003C\u002Fp>\u003Cp>Speaker 0: Shoot me the shoot me the URL in the chat. I don't see it in there.\u003C\u002Fp>\u003Cp>Speaker 1: It's a imigos.directus.app.\u003C\u002Fp>\u003Cp>Speaker 0: Got it now. Alright. Alright. I'm in.\u003C\u002Fp>\u003Cp>Speaker 1: Alright. Alright. Alright. Alright. Invalid input.\u003C\u002Fp>\u003Cp>Sarah Chen. Why does it keep trying to I don't know. It really likes the name Sarah for whatever reason.\u003C\u002Fp>\u003Cp>Speaker 0: Hey. I got a crush.\u003C\u002Fp>\u003Cp>Speaker 1: Marcus, Priya, Sarah. Those are the folks who are uploading the other skills. Alright. So we're good then. Yeah.\u003C\u002Fp>\u003Cp>There's gonna be some skills that populate in there, and then everybody is jamming on the back end. How long are we do we have on the back or on the front end? Right?\u003C\u002Fp>\u003Cp>Speaker 0: I set a timer. We have, like, two minutes now, but, if we wanna take ten minutes, you know, and\u003C\u002Fp>\u003Cp>Speaker 1: do what we Ten minutes to do a whole back end or a front end?\u003C\u002Fp>\u003Cp>Speaker 0: Yeah, man. I mean, it's the UI error is changing, so they say.\u003C\u002Fp>\u003Cp>Speaker 1: I at least, like, 15. Yeah?\u003C\u002Fp>\u003Cp>Speaker 0: 20 to 30.\u003C\u002Fp>\u003Cp>Speaker 1: We could do, like, a supercut.\u003C\u002Fp>\u003Cp>Speaker 0: We'll extend this session a little bit. We've got some time.\u003C\u002Fp>\u003Cp>Speaker 1: Do a supercut. Yeah. Alright. Twenty minutes then. That's it.\u003C\u002Fp>\u003Cp>Twenty minutes. No more. No less.\u003C\u002Fp>\u003Cp>Speaker 0: I think You\u003C\u002Fp>\u003Cp>Speaker 1: guys better not be starting with the back end already or front end already.\u003C\u002Fp>\u003Cp>Speaker 0: Sure John's already done. I've been Yeah.\u003C\u002Fp>\u003Cp>Speaker 1: He's already done. Hearing him. Alright.\u003C\u002Fp>\u003Cp>Speaker 2: I'm just right now, I'm just setting up the m I'm setting up my MCP and cursor.\u003C\u002Fp>\u003Cp>Speaker 1: Alright. Well, you guys just tell me when you're ready to go.\u003C\u002Fp>\u003Cp>Speaker 0: I'm I'm pretty confident\u003C\u002Fp>\u003Cp>Speaker 1: in my skills.\u003C\u002Fp>\u003Cp>Speaker 0: Let's say we're using, like, you know, lovable or something. I've got my user here. I create It's got admin policy, create a new token, save said token, and should be good to pull data from that directly.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. And just lovable have FCP?\u003C\u002Fp>\u003Cp>Speaker 0: I don't know. It's a good question.\u003C\u002Fp>\u003Cp>Speaker 1: Personal connectors. There you go, dude.\u003C\u002Fp>\u003Cp>Speaker 0: I see it. I sees it.\u003C\u002Fp>\u003Cp>Speaker 1: Here's the directisio docs. I'm gonna just point you in the right direction there. Connect AI, AI, MCP server. This is the page that you need from the documentation. I'm gonna paste that in the chat.\u003C\u002Fp>\u003Cp>Good luck to you, sir. You guys just all the skill the back end is done. The back end is locked. Don't even think about changing the back end.\u003C\u002Fp>\u003Cp>Speaker 2: Alright.\u003C\u002Fp>\u003Cp>Speaker 1: When you guys say go, I'm gonna hit go. Hit go. There it is. Alright. Twenty minutes, boys.\u003C\u002Fp>\u003Cp>Speaker 0: I'm I've gotta upgrade my lovable account because, apparently, Lovable doesn't allow custom MCPs unless it's It\u003C\u002Fp>\u003Cp>Speaker 2: sounds like an excuse, bro.\u003C\u002Fp>\u003Cp>Speaker 1: Already an excuse.\u003C\u002Fp>\u003Cp>Speaker 0: Alright. One pro. I'm gonna spend this whole time setting up my custom MCP.\u003C\u002Fp>\u003Cp>Speaker 1: Oh, no. Why can't I see the text here? Oh, that's not good. I'm going react on this one. Why can't I see this, though?\u003C\u002Fp>\u003Cp>I just got this new Mac, man. Still, if you guys beat me, that will be b y since we're just throwing excuses out there.\u003C\u002Fp>\u003Cp>Speaker 0: I am still signing a still I'm still going there by a sign up for a year.\u003C\u002Fp>\u003Cp>Speaker 1: I can't even get the I can't get the tan stack start script here from the their website.\u003C\u002Fp>\u003Cp>Speaker 2: Start.\u003C\u002Fp>\u003Cp>Speaker 0: Connection failed. Gotta love it. Alright. Well, mobile mobile is not really communicating with me right now. I'm having trouble.\u003C\u002Fp>\u003Cp>I hope you guys are having fun. I'm having a blast.\u003C\u002Fp>\u003Cp>Speaker 2: Matt, is your keyboard connected yet?\u003C\u002Fp>\u003Cp>Speaker 0: Yes, John.\u003C\u002Fp>\u003Cp>Speaker 2: Alright. Let's see. I got my build plan ready. Let's see if this looks good.\u003C\u002Fp>\u003Cp>Speaker 0: No. I don't think Lovable's gonna be the the move. Not great, Lovable. Not great.\u003C\u002Fp>\u003Cp>Speaker 2: Alright, guys. I've clicked build. We're gonna we're gonna get v one running. What if we finish in ten minutes? Does that give us bonus points?\u003C\u002Fp>\u003Cp>Speaker 0: I should be done setting up my MCB by then. So\u003C\u002Fp>\u003Cp>Speaker 2: Matt, why don't you have a rabbit in your background? Brian and I both have a rabbit in the background.\u003C\u002Fp>\u003Cp>Speaker 1: I have several.\u003C\u002Fp>\u003Cp>Speaker 0: Because I like to have my rabbits in my point of view. So not in the background. Alright. Well, local's not gonna work with me on the MCP, so I'm gonna have to roll.\u003C\u002Fp>\u003Cp>Speaker 1: You just just gonna have to roll. Can I just copy, Jace? You could just tell it to use the SDK. I think you\u003C\u002Fp>\u003Cp>Speaker 2: can copy.\u003C\u002Fp>\u003Cp>Speaker 1: I think\u003C\u002Fp>\u003Cp>Speaker 2: you can also just give it your API token and just give it access. I think I've done that before. You'd be like, hey. I'm an API token. Didn't know we'd get excuses so quick out the gate, though.\u003C\u002Fp>\u003Cp>Speaker 0: You know, usually, I'm using, like, whisper for this. Go for it. Okay. Then we\u003C\u002Fp>\u003Cp>Speaker 1: know what you're gonna build. Right?\u003C\u002Fp>\u003Cp>Speaker 2: Just meet your just meet yourself while you mouth it and then unmute.\u003C\u002Fp>\u003Cp>Speaker 1: I need a front end that looks great. Make no mistakes.\u003C\u002Fp>\u003Cp>Speaker 2: Make it look beautiful.\u003C\u002Fp>\u003Cp>Speaker 1: I I hope John's idea is something that looked like a pot of gold and a leprechaun.\u003C\u002Fp>\u003Cp>Speaker 2: No. But that would have been good. Dang. I think mine'll be good, though.\u003C\u002Fp>\u003Cp>Speaker 0: I I have no idea what this output of this prompt is gonna mean. I'm actually kind of terrified. I\u003C\u002Fp>\u003Cp>Speaker 1: know. I'm just reading my typing, and it's freaking terrible.\u003C\u002Fp>\u003Cp>Speaker 0: That's why he's was\u003C\u002Fp>\u003Cp>Speaker 2: The good news is AI doesn't care about typos.\u003C\u002Fp>\u003Cp>Speaker 0: What were the tags we put in here, by the way?\u003C\u002Fp>\u003Cp>Speaker 1: Research writing, design, automation, productivity, coding, marketing, and data analysis.\u003C\u002Fp>\u003Cp>Speaker 2: We should upload these to to, like, Vercel or Netlify or something afterwards so people can play with them, the front ends.\u003C\u002Fp>\u003Cp>Speaker 1: That should be that should be solid.\u003C\u002Fp>\u003Cp>Speaker 2: Here's the file being created right now. It's called bang effect dot JSX.\u003C\u002Fp>\u003Cp>Speaker 1: Bang effect. You may win this, man. I think you guys had more time to think about the ideas here.\u003C\u002Fp>\u003Cp>Speaker 2: Don't\u003C\u002Fp>\u003Cp>Speaker 0: yeah. I\u003C\u002Fp>\u003Cp>Speaker 1: don't have a solid idea.\u003C\u002Fp>\u003Cp>Speaker 0: Lost trying to set up the MCP server in Lovable.\u003C\u002Fp>\u003Cp>Speaker 1: Oh, that's a risky run. You've got what what did we got on the clock? Thirteen twenty one.\u003C\u002Fp>\u003Cp>Speaker 0: Plenty of time.\u003C\u002Fp>\u003Cp>Speaker 1: Just just come on, Claude. Don't film me now. I like you, dawg. We got this. We got this.\u003C\u002Fp>\u003Cp>Speaker 2: Oh, first preview. It looks good.\u003C\u002Fp>\u003Cp>Speaker 0: Did you one shot it? Oh,\u003C\u002Fp>\u003Cp>Speaker 2: damn. This is okay. You guys are gonna lose.\u003C\u002Fp>\u003Cp>Speaker 1: He's blowing smoke. I know.\u003C\u002Fp>\u003Cp>Speaker 0: He's he doesn't have anything. It's just a blank screen.\u003C\u002Fp>\u003Cp>Speaker 2: I wish you guys could see this. No.\u003C\u002Fp>\u003Cp>Speaker 0: I'm trusting lovable with my lunch.\u003C\u002Fp>\u003Cp>Speaker 2: The sound effects are amazing.\u003C\u002Fp>\u003Cp>Speaker 1: You have sound effects?\u003C\u002Fp>\u003Cp>Speaker 2: The only problem is I can tell the words are backwards. I'm like, why are these words backwards?\u003C\u002Fp>\u003Cp>Speaker 0: What, wait. Tags. One, two, three.\u003C\u002Fp>\u003Cp>Speaker 2: I like how there's, like, duplicate tags.\u003C\u002Fp>\u003Cp>Speaker 0: Lovable has gotten a lot better on the user interface since all of us use it. Like, it quite looks nice, except for the whole connecting and MCP thing.\u003C\u002Fp>\u003Cp>Speaker 2: I wish you guys could hear these sound effects.\u003C\u002Fp>\u003Cp>Speaker 1: I think I can hear a few.\u003C\u002Fp>\u003Cp>Speaker 0: Did you do freaking lasers?\u003C\u002Fp>\u003Cp>Speaker 2: Oh, it's looking so much better now. Oh my gosh. This looks so good.\u003C\u002Fp>\u003Cp>Speaker 0: It's a risk. It's poker face.\u003C\u002Fp>\u003Cp>Speaker 2: No. You know it's a good a good project when you are playing the game that you've built.\u003C\u002Fp>\u003Cp>Speaker 1: I'm still waiting for dependencies to resign. Alright. You know me, Ben. If you enjoy watching people slide by on the skin of their teeth, please tune in to Directus TV for where yours truly embarrasses himself twenty four seven by trying to build apps. Oh god.\u003C\u002Fp>\u003Cp>This is going horribly wrong. I should not have should not have done React that is using way outdated stuff. We're just I I don't know. Tam, if you watch this, man, I don't know what happened to the website today when I was recording this, but\u003C\u002Fp>\u003Cp>Speaker 0: I think this is gonna be\u003C\u002Fp>\u003Cp>Speaker 1: Oh, there it is. Finally. It's okay.\u003C\u002Fp>\u003Cp>Speaker 0: You got something?\u003C\u002Fp>\u003Cp>Speaker 1: Must have been I the site must have been broken, man. I don't I don't know.\u003C\u002Fp>\u003Cp>Speaker 2: Okay. Now now it's playing its playing its game on its own. This is where where it's fun to watch Cursor play its own game.\u003C\u002Fp>\u003Cp>Speaker 0: You got your AI playing an AI? K. What are you doing over there?\u003C\u002Fp>\u003Cp>Speaker 2: Who? Me? Just kicking ass?\u003C\u002Fp>\u003Cp>Speaker 0: I'm gonna have to rely on a one shot here.\u003C\u002Fp>\u003Cp>Speaker 2: I think I might have this with a one shot. We'll see. It's still it's finishing up the build.\u003C\u002Fp>\u003Cp>Speaker 0: This is building out the comp I'm building the character select and move list UI right now, so we're gonna have to wait a a second. Did you create an RPG, John?\u003C\u002Fp>\u003Cp>Speaker 2: No. I created a a video game. Do you wanna know what its theme theme is?\u003C\u002Fp>\u003Cp>Speaker 0: I wanna wait and see.\u003C\u002Fp>\u003Cp>Speaker 2: Okay.\u003C\u002Fp>\u003Cp>Speaker 1: Leprechauns?\u003C\u002Fp>\u003Cp>Speaker 2: No.\u003C\u002Fp>\u003Cp>Speaker 0: Pecauns. I think I you know, if I had more time, it would be cool to do, like, an RPG style directory where you have to, like, level up to get the better prompts. I wonder if anybody would spend time or not. Probably not.\u003C\u002Fp>\u003Cp>Speaker 1: I've been looking at this thing, and I'm just like, I don't even know.\u003C\u002Fp>\u003Cp>Speaker 0: Do you have anything? Do you have a UI at all, Brian?\u003C\u002Fp>\u003Cp>Speaker 1: I I have nothing on the screen.\u003C\u002Fp>\u003Cp>Speaker 0: That's not good. How much time do we have left?\u003C\u002Fp>\u003Cp>Speaker 1: Six minutes. Oh, man.\u003C\u002Fp>\u003Cp>Speaker 2: Guys, this you might as well just stop. I'm telling you, you're gonna lose. I know Matt basically has nothing. He's gonna show a white screen. It's gonna be like index.html.\u003C\u002Fp>\u003Cp>Speaker 0: We're gonna wait and ask I\u003C\u002Fp>\u003Cp>Speaker 1: have a header.\u003C\u002Fp>\u003Cp>Speaker 0: You have a header?\u003C\u002Fp>\u003Cp>Speaker 1: I have a header.\u003C\u002Fp>\u003Cp>Speaker 2: Match this line to us.\u003C\u002Fp>\u003Cp>Speaker 0: Have have I been in the roost this whole time? Who knows?\u003C\u002Fp>\u003Cp>Speaker 1: I feel like I do miss the old days where I just good old fashioned handwriting code. I probably yeah. On a 100 apps, hundred hours, like, the days of all, I woulda had something already that looked nice. I'm not super happy with what I've got here.\u003C\u002Fp>\u003Cp>Speaker 0: I do only have a white screen.\u003C\u002Fp>\u003Cp>Speaker 2: K. I put in one more prompt. I I'm three shotting it, and I think three shots might be it.\u003C\u002Fp>\u003Cp>Speaker 0: I think my idea was here, but the execution is lacking.\u003C\u002Fp>\u003Cp>Speaker 2: Oh, it even gave it a little tab icon. Nice.\u003C\u002Fp>\u003Cp>Speaker 0: Alright. I think we got something.\u003C\u002Fp>\u003Cp>Speaker 2: Matt, are you past the home page yet? Did you get logged in yet?\u003C\u002Fp>\u003Cp>Speaker 0: Been spending all this time on the initial\u003C\u002Fp>\u003Cp>Speaker 2: You spent fifteen minutes of the twenty minutes just upgrading to pro. I did.\u003C\u002Fp>\u003Cp>Speaker 0: I might have credit card information floating out there, so remind me to cancel my cards. Alright. We got something working. Yeah. Without the MCP, I'm I'm not really able to pull the.\u003C\u002Fp>\u003Cp>Speaker 2: Here's what here's what my cursor just said so you know what it's doing. The buttons are noticeably bigger now. Let me shoot a couple skills first, then test the collected view and detail pop up. That's a little hit into what I've built.\u003C\u002Fp>\u003Cp>Speaker 0: Are we talking legend of Zelda or something? What? Dangerous to go\u003C\u002Fp>\u003Cp>Speaker 1: alone. Sweet mother.\u003C\u002Fp>\u003Cp>Speaker 0: He's still in there, or is that going too?\u003C\u002Fp>\u003Cp>Speaker 1: I still got it. Started off on the wrong foot, man. Should not have, should not have tried to go React. Should just you gotta stick with what you know, man.\u003C\u002Fp>\u003Cp>Speaker 0: Talk with what you know. Why did you go with React? You were trying to get fancy?\u003C\u002Fp>\u003Cp>Speaker 1: I I've been using, I I got a project, a personal project where I've used, TanStack start lately. I'm trying to learn a little more about React so that I'm not one dimensional. Mhmm. Did not pay off today. Did not pay off today.\u003C\u002Fp>\u003Cp>Mhmm. Thirty seconds. Yeah.\u003C\u002Fp>\u003Cp>Speaker 0: This is I've just spent all the time troubleshooting. Lovable doesn't seem to really work well with custom MCPs, so if anybody's curious.\u003C\u002Fp>\u003Cp>Speaker 1: Fifteen, fourteen, 13. Who are you who's who gets to go first?\u003C\u002Fp>\u003Cp>Speaker 0: Well, I\u003C\u002Fp>\u003Cp>Speaker 1: mean, clearly, it sounds like John.\u003C\u002Fp>\u003Cp>Speaker 0: Like, John's face means that\u003C\u002Fp>\u003Cp>Speaker 1: he may have something really good. He might have. Yeah. Time's up.\u003C\u002Fp>\u003Cp>Speaker 2: Yeah. Time's up. It's just finishing it's just finishing its build, but I've just been I only needed three prompts to get it to some pretty amazing.\u003C\u002Fp>\u003Cp>Speaker 0: Okay. I mean, I've got something. It's not,\u003C\u002Fp>\u003Cp>Speaker 2: Sounds like you have nothing.\u003C\u002Fp>\u003Cp>Speaker 1: It's basically You go then. Me?\u003C\u002Fp>\u003Cp>Speaker 0: Let me it's still finishing this moment. The junction table I faced. Alright. I just refreshed. We'll see.\u003C\u002Fp>\u003Cp>Speaker 1: I I think that is I'll go then. I'll go.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Let's see what you got, Bryant. I'm excited to see this\u003C\u002Fp>\u003Cp>Speaker 1: header. Screen. I didn't have time to work on the design. We have a skills directory. I was going for, like, the brutalist theme.\u003C\u002Fp>\u003Cp>It didn't really get. I I had literally no time to worry on the design. So the the tags do work, which is nice. We can search. So we search for SEO.\u003C\u002Fp>\u003Cp>That works. We can load up the skill. We see it being loaded here. You know, you got the nice little subtle detail, the little arrows instead of the standard bullet points. You got the the fake install.\u003C\u002Fp>\u003Cp>And then the coup d'etat. Is that what the I don't know. I don't know. Matt has surely lost.\u003C\u002Fp>\u003Cp>Speaker 2: Nice.\u003C\u002Fp>\u003Cp>Speaker 1: Stakes. Pretty badly.\u003C\u002Fp>\u003Cp>Speaker 0: Oh, so you can submit skills from the front. That's pretty good. Pretty good.\u003C\u002Fp>\u003Cp>Speaker 1: Oh. No.\u003C\u002Fp>\u003Cp>Speaker 0: Failure.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. I did get a chance to to fix that. If I had, like, two more minutes, like, whatever it is that you're doing right now while I'm demoing mine.\u003C\u002Fp>\u003Cp>Speaker 2: I'm not doing anything.\u003C\u002Fp>\u003Cp>Speaker 1: This looks good. I mean talking to Matt.\u003C\u002Fp>\u003Cp>Speaker 0: Design wise? What like, I mean, I think it looks solid. I would probably call it\u003C\u002Fp>\u003Cp>Speaker 1: I I wouldn't have gone, like, the dark mode design. Right? I would have gone light mode, and I would've, like, just had it be very in your face. Would have been nice to have, like, the login to be able to submit this, but, again,\u003C\u002Fp>\u003Cp>Speaker 0: just good\u003C\u002Fp>\u003Cp>Speaker 1: to me. Wrestle with it too much.\u003C\u002Fp>\u003Cp>Speaker 2: It's more than I was expecting. It's it looks good.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Call it Hill Skills. And\u003C\u002Fp>\u003Cp>Speaker 1: it's got,\u003C\u002Fp>\u003Cp>Speaker 0: like, kind of a nineties vibe to it. I dig it. Alright. I guess I'll go next. We'll see what happens here.\u003C\u002Fp>\u003Cp>But, alright. This is called, AI skills pro skater presented by AI Amigos. So the concept\u003C\u002Fp>\u003Cp>Speaker 2: I wanna say I'm impressed that you have something. I thought you were gonna show us nothing.\u003C\u002Fp>\u003Cp>Speaker 0: Maybe maybe I was in Parker Face the whole time. But, anyways, the concept here is you start by tag. So you pick your, it's not really skateboarding, but the concept was Tony Hawk's Pro Skater. But, it was supposed to be whoever like, I wanted to do it by people in an organization. So engineers would wanna probably see the engineering skills, marketers, that sort of thing.\u003C\u002Fp>\u003Cp>So starting with tags, which is something I haven't seen before. And then in true Tony Hawk fashion, ideally, what I was trying to get was to unlock the skill, you have to do the, like, combo thing with the keyboard. And, it would do a little trick sort of thing. If I had more time, this would be like a full Tony Hawk style game where you can unlock skills that way. But, have all of the ones here.\u003C\u002Fp>\u003Cp>It just randomly put some, like, easy hard sick, grade on here, but I still pulled in all the overview platform tips and stuff. Not as nice as I was, like, hoping to get, but\u003C\u002Fp>\u003Cp>Speaker 1: Where's the skateboards?\u003C\u002Fp>\u003Cp>Speaker 0: I know. There is no skateboards.\u003C\u002Fp>\u003Cp>Speaker 2: I was I I\u003C\u002Fp>\u003Cp>Speaker 0: I was disappointed.\u003C\u002Fp>\u003Cp>Speaker 2: I think you've got the beginnings of something. I think this is cool. I think this is a good idea. I think it's cool.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. I was trying to go for the most non directory directory type, but, yeah. That's what that's what I\u003C\u002Fp>\u003Cp>Speaker 1: That's cool. Yeah. This part, I like, man. This looks this looks super nice.\u003C\u002Fp>\u003Cp>Speaker 2: It's totally\u003C\u002Fp>\u003Cp>Speaker 1: It's even got the\u003C\u002Fp>\u003Cp>Speaker 2: Like, I like that they're even floating.\u003C\u002Fp>\u003Cp>Speaker 1: It's like the the CRT Yeah. Tube TV effect to it. Mhmm. Yeah. I've been lately.\u003C\u002Fp>\u003Cp>Speaker 0: But yeah. So that's the end. That's AI skills for a skater.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. I think it's for a skater.\u003C\u002Fp>\u003Cp>Speaker 0: Alright, John.\u003C\u002Fp>\u003Cp>Speaker 1: Alright. Let's see what you got, John boy.\u003C\u002Fp>\u003Cp>Speaker 2: Alright. Let's do this. I went you know, if you've if anyone's followed me on LinkedIn, I did the whole Nintendo theme garage inventory. So I kinda kept it Nintendo themed, and I went with Duck Hunt, although Skills Hunt. So you come here and you you shoot the little you you can't hear it or maybe you can't.\u003C\u002Fp>\u003Cp>There's little noises. There's a little creature down there, and then you can you can view the collections that you've got. You've got a little the only thing I would do is make the fonts bigger. You can see all of these things here. But you can go back to the game and you shoot the little you shoot the skills that you want.\u003C\u002Fp>\u003Cp>Speaker 1: The dog is So there there\u003C\u002Fp>\u003Cp>Speaker 2: is noises. I don't think you can hear them, but maybe the recording will show them. But there is noises when you're shooting things, and that's what I did. You got your skills there on the right that you've collected. You've got your skills here, and then you can view what those skills are.\u003C\u002Fp>\u003Cp>Speaker 0: Alright.\u003C\u002Fp>\u003Cp>Speaker 2: So, yeah, shoot for\u003C\u002Fp>\u003Cp>Speaker 1: ask questions later.\u003C\u002Fp>\u003Cp>Speaker 2: Actual game I already got that one. Damn. I need a new skill. There we go. Bang.\u003C\u002Fp>\u003Cp>Speaker 0: I love gamifying the skills. Yeah. You can't just get the skills. You gotta work.\u003C\u002Fp>\u003Cp>Speaker 2: You gotta work for it, man. Anyways, that's what I\u003C\u002Fp>\u003Cp>Speaker 1: know. Alright.\u003C\u002Fp>\u003Cp>Speaker 0: I mean, all skill one. Honestly, I think way better than all of us anticipated, at least for for me.\u003C\u002Fp>\u003Cp>Speaker 2: I mean, we did all of those Friday It's a working game. Twenty minutes.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Very nice.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. It was twenty minutes on the clock and mostly fighting with my, connection and react, and John was his master of vibes over there. So\u003C\u002Fp>\u003Cp>Speaker 2: Master of vibes.\u003C\u002Fp>\u003Cp>Speaker 0: I mean, if\u003C\u002Fp>\u003Cp>Speaker 1: I had to speak Does this go to a vote between us,\u003C\u002Fp>\u003Cp>Speaker 0: or does it does it go to the audience? I mean, I I think it's a good question. Right.\u003C\u002Fp>\u003Cp>Speaker 2: I think we all know who won this one here internally.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Yeah. Yeah. Yeah.\u003C\u002Fp>\u003Cp>Speaker 1: As much as it pains me to\u003C\u002Fp>\u003Cp>Speaker 0: say it, very creative. Very creative. Mhmm. I like shoes.\u003C\u002Fp>\u003Cp>Speaker 2: I think they're all good. I think they're\u003C\u002Fp>\u003Cp>Speaker 1: all\u003C\u002Fp>\u003Cp>Speaker 2: great, dare I say. They need all they all need adjustments, obviously, if you were to move this to production, but I think they were all great ideas in different manners.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. And then you combine them all. And that's I mean, I think the best thing is that, like, I'm imagining you have, like, a full team of folks operating off the same like, you like, at companies I've worked at, they've they've always been hackathons. But, like, opening up a hackathon to the entire team to be able to work off the same base set of data, I think, is just the really cool thing here. It's, like, everybody can bring ideas to the table now.\u003C\u002Fp>\u003Cp>But, yeah, overall, pretty cool.\u003C\u002Fp>\u003Cp>Speaker 1: Well, cool, guys. I think that's\u003C\u002Fp>\u003Cp>Speaker 0: about it for this session. I appreciate everybody hanging in there with us. It sounded like if you're just listening to this, it was awful, but, you know, the end result has been solid.\u003C\u002Fp>","What's up, everybody? It is great to see you, to meet you virtually here. My name is Matt. I'm on the marketing team here at Directus. And if you're watching this, you're in for a treat because we have quite the repertoire of guests here today, also from the Directus team, and we're gonna be doing something pretty fun. But first, let me introduce each guest. You know him. You love him. You've seen his face all over the place. It's John Daniels. John Daniels. What's up? Glad that you started with me. I was worried you were gonna say Brian. So hello. Your face is all over the place right here. I don't think anybody really knows who Bryant is. So, no. No. No. No. Of course, we have the man himself, mister Bry. Bryant Gillespie, who's one of our esteemed developer advocate type folks, I guess, on the team. You know? He kinda does everything. Type folks? Yes. Happy to be here. Very excited for this. Look. I'm just gonna start this off with saying whatever we're doing in this competition, I'm gonna mop the floor with you guys. That's it. Strong start. I was gonna do this not in the spirit of competition, but in Oh. In in friendliness, but I guess that's what we're gonna do now. Alright. And if anybody doesn't know, John is one of our AEs here at Directus, so the most competitive of of everyone here. So I'm sure pretty sure that'll awaken a fire in a moment. So, yeah. Brian, what are we doing today? I this is your show, my friend. You tell me what we're doing. Clearly, it's something to do with an AI skills directory because that is what's on the screen. That is what's on the screen. Alright. I'll give you the brief rundown. So today, we are gonna have a build off. Everybody's been talking about front ends and UIs, and the errors are shifting. And there's so many tools out there. I'm a user of Cloud Code for a lot of the marketing stuff that we do here at Directus, including building the Leap Week landing page. Our team uses codex. Our team uses a lot of the out of the box, like, harnesses, like cursor. And we also use things like Lovable. So it's a good mix of things, but that's a lot of subscriptions add up. So we thought, why not put the skills to the test and see which we can use best for building the front end on a Directus back end, which would be a skills directory. And one of the reasons we wanna do a skills directory is because there's a lot of shady skills directories out there and there's a lot of mistrust in a lot of them. So our developer team has a lot of, skills that they share, that we thought would be cool to build something along with that, as well as our go to market team uses a lot of things too. So, yeah, that's really the gist. Did I What are the what are the stipulations for this, man? We're building a skills directory front end. We're gonna need a back end for that, which we'll do, but what are the rules? There are no rules. There's only one rule. And that is that's details? Yeah. There's gonna be one rule where we have to create the back end and direct us. And then whatever tool we're gonna use, we are going to use the same back end for that front end. So running down the list of tools that we will use, I will be using Lovable. It's something I've used in the past. I've I've not actually used it in quite a bit, because I've been using, Cloud Code. But, Brian, what are you gonna be using? I'm gonna be using Cloud Code. Alright. John, what you got? I use my old dependable cursor. Yeah. Is cursor what you built? The, garage inventory? It is. Yeah. Yeah. And I already have a I already have an idea. Oh, boy. Oh, no. Retro. It's a lot yeah. Front end directory. It's a long Here we go. I'm doing that maybe more than, like, moving forward in my chair. Cool. Yeah. Well So Right. To kick things off, I'm assuming we'll kick we've started. We we got a totally blank data model here. We'll just call it skills directory. Right? John is just banging on the keyboard, man. Alright. Totally empty. It's like you know, we'll we'll do this 100 app style. Like, what what what do we even need on a skills directory? Right? You're gonna need some skills, but what what is the what are the other stuff that you're planning on tracking here? What do you need for the back end? Yeah. That's a good question. When you say skills, what does that encompass? Is that like the title and the description and the actual I mean, skills are basically just a fancy word for markdown files. Right? Yeah. I mean, that's, yeah, that's the the funny thing about all of this stuff to me is it's all just markdown. Like, been been dealing with markdown since I was 12. A week ago. Yeah. So we got a name, a description for the skill. You've got the, what, the actual skill content. I don't I don't know. Are you I'm assuming you're gonna have, like, users who submit the skills? Mhmm. We'll have I think, at some point, if we wanna Submit a bug. Public facing, but for our internal team makes sense. You probably want, some sort of, like, tagging system. So if it's engineering skill or if it's So the tag name. Okay. What else you got? Option for an image, if you wanna add that to the front end potentially. Image? Okay. You're getting away from the markdown stuff now. I'm gonna just draw the arrows for fun here at this point. Here we go. Boom. Alright. So you got skills. You got tags. What else do you need? I mean, I Repository. GitHub repository. Is that where they're coming from? Could you just host it within Directus itself, like, as a file? I mean, we can. Yeah. Like, do you want it linked to somewhere? I think I wanna keep it, like, self like, contained within the one thing. Okay. Alright. Cool. Is this anything else? Or is it we're just gonna roll with this. Just keep it simple. Unless you got anything to do. Wait. No. Keep it simple, stupid. Which one is stupid, man? You probably said it. Alright. I've got this. So, hopefully, I've got some secret sauce here. Been working on a CLI tool. So I'm gonna tell Claude code here. Can you guys see this? Should I zoom way in? Alright. Run d six s agents. It's our CLI for connecting to direct us. D six s. It's the first time seeing the CLI in action right here. Very nerdy. You know, the Kubernetes deal. What is this thing doing? Why is it not running this command? It's reading four files, which d success. Just run the command, dude. What are you doing? Chose the wrong tool. Maybe cursor's been Maybe you chose the wrong tool. Yeah. There it is. Okay. So, basically, that command gives it a little context about this. Cool. And now this thing should be primed to go. Agents. Right. Cool. Alright. So now we're gonna kick it over into plan mode. I'm gonna paste that image that I had. Right? And we're just gonna say, hey. We're building an AI skills directory. Here's a screenshot of a data model I'm competing with John and Matt. We need the back end in Directus first before we build the front end. Can you plan out the Directus back end? Yeah. Alright. I've been using Let's see what happens here. Okay. Right here. Alright. It is going to alright. This is where I tell you to do as I say, not as I do, because I'm just gonna turn this thing loose. Right? I should have actually set dangerously set permissions here. But I don't even have it ask for permission anymore. I just let it go. Go, baby. Go. I have to leave this a lot. That's probably why. If you can't leave that out, security infraction. Security. Spent hundreds of dollars in tokens last month. Alright. It's going to write this plan. Are you so are are you not using cloud code at all? You're still using cursor every day? I use both. Depends on what I'm doing. So what so what you're you're basically just doing what the MCP used to do just from the terminal So is the why the CLI? Right? Well, the reason why is because context. The context window. So MCP uses up a lot of context, for comparison. Right? Like, if you have all the tools enabled, MCP or our AI assistant right now could use up to thirty, forty thousand tokens to have all the tools enabled. The CLI, by default, doesn't load anything into the context, and the agent is or like CloudCo, these agent harnesses are pretty good at navigating, like, CLIs just because they're they're trained on a lot of bash scripts and stuff like that. So, you know, it basically you keep the context window clean until you need it, and, yeah, you follow this progressive discovery pattern. I'm yeah. I've been talking. I'm not even looking at this plan. Hopefully, this thing actually does what we wanted to do. We shall see. Right? Actually, use recipes. D success recipes. This is should be a shorthand. Again, this is all highly experimental, so failed to log the plug in, so it's not gonna use that at all. Plug ins. Index. Oh, good. Alright. Well, just push forward. We may be kicking back to the MCP for this. Hey. Alright. Is what I when I type in the when it when MCP goes wrong for me, I say, just do it, and it'll it'll figure it out. Just do it. Figure it out funny, man. Turn off permissions and just go. What is this thing doing? Your token doesn't have admin access. I do have admin access. No. You're wrong. You said, highly experimental. Correct? I mean, it it was working well. Works on my machine until you start recording. Right? Right. So we'll just go back to the old tried and true here. Both collections exist, but I have no schema. Oh my gosh. This thing is hopelessly lost. Lost. Hey. There's two collections in there now. Yep. They're actually just folder collections because they don't exist. Alright. Skills. When I say tried and true, I mean me. Alright. What do we have there? We have name, description, content. Name. I've never seen anyone move this fast inside of the. Content. Just markdown. It's just markdown. There you go. Alright. You got tags. Yeah? Mhmm. Cool. Tags. I have a name for the tags. Tag name. Cool. And then we need a skill tags. We need a junction table. Cool. I'm taking the long way around here. Skills. Now I'm gonna go to the mini mini relationship inside directions. It's been a minute since I've done this, man. Don't you miss it? Always I do miss it. Yeah. It took me an hour to do half of this stuff that I could now do in, like, five minutes using the AI tools when they work, mind you. Skills ID, tags ID. Let's just call it skill. Let's call it tag. Skills. Cool. Give it a sort field. Boom. Boom. Boom. Boom. You know, it's worth that messing up to watch you do this. It's like watching Boom. That's it. There's the back end. You're good. Golden. Very simple. Very easy. Now what? Now we need some data. Mhmm. Maybe it can do that. Look, I I set it all up. Nice. Thanks for nothing. Now I need you to create some sample data for us, though. I mean, I'm a make sure you rub it in the face of the thing. Right? Talk to me well, I this might take a minute, but, like, what are you guys Like, when you are are we supposed to talk about what we're gonna build? Like, how the front end is gonna look, or is this gonna be, like, a secret thing? We'll talk about, you know, how things are going. I'm still I'm between a couple ideas on what I want this front end to look like. I I want it to not look like a generic directory because there's, like, directories of directories out there. I already have an idea, so don't even don't even worry about Wait. Alright. Was it one that we talked about earlier, or is this one this is completely new? I got a new idea. Interesting. Okay. I think I'll have an idea. You think you'll have an idea? Alright. Well, it's got there's eight tags. There's six skills. Let's add 10 more skills. Create a few sample users as well, give them a different role. Just the ability to view skills and update the ones they've uploaded. Alright. So now we're gonna move to the the front end stuff. Yeah? Yeah. I think so. So we're all operating off of the same direct us back end, and you just created a few users. I'm assuming that's to give us access to that data. Yep. So you guys have your own users in there. Looks like it's created Only one guy cares about a photo. Yeah. Oh my gosh. Alright, dude. Fine. Fine. Fine. Fine. Yeah. Boom. I'll update my You look you look rather young in that one, John. Thanks. Alright. So what do I dig at all. Thank you. Alright. So each of us is gonna have our own role, and you will provide that role to us. If you refresh real quick. Sorry. Refresh bright. There. Of course. Yeah. We're reading we're watching this. You're certainly not watching this on Saint Patrick's Day. We're recording it on Saint Patrick's Day, though. Yeah. Yeah. Yeah. Well done. Yeah. That one. Sorry. Continue, Matt. How do you have access to this already? I I already gave him access. I put it in the chat. I sent you guys up already. Your direct us email and your password. What's the password? The password is Password. Now, hopefully, one of us will remember to have removed that by the time this recording goes live. I'll tell him how to change it. In the chat. Must be what is this thing doing. Okay. I was creating, like, 35. Why can't I not just accept all here? Good lord. Shoot me the shoot me the URL in the chat. I don't see it in there. It's a imigos.directus.app. Got it now. Alright. Alright. I'm in. Alright. Alright. Alright. Alright. Invalid input. Sarah Chen. Why does it keep trying to I don't know. It really likes the name Sarah for whatever reason. Hey. I got a crush. Marcus, Priya, Sarah. Those are the folks who are uploading the other skills. Alright. So we're good then. Yeah. There's gonna be some skills that populate in there, and then everybody is jamming on the back end. How long are we do we have on the back or on the front end? Right? I set a timer. We have, like, two minutes now, but, if we wanna take ten minutes, you know, and do what we Ten minutes to do a whole back end or a front end? Yeah, man. I mean, it's the UI error is changing, so they say. I at least, like, 15. Yeah? 20 to 30. We could do, like, a supercut. We'll extend this session a little bit. We've got some time. Do a supercut. Yeah. Alright. Twenty minutes then. That's it. Twenty minutes. No more. No less. I think You guys better not be starting with the back end already or front end already. Sure John's already done. I've been Yeah. He's already done. Hearing him. Alright. I'm just right now, I'm just setting up the m I'm setting up my MCP and cursor. Alright. Well, you guys just tell me when you're ready to go. I'm I'm pretty confident in my skills. Let's say we're using, like, you know, lovable or something. I've got my user here. I create It's got admin policy, create a new token, save said token, and should be good to pull data from that directly. Yeah. And just lovable have FCP? I don't know. It's a good question. Personal connectors. There you go, dude. I see it. I sees it. Here's the directisio docs. I'm gonna just point you in the right direction there. Connect AI, AI, MCP server. This is the page that you need from the documentation. I'm gonna paste that in the chat. Good luck to you, sir. You guys just all the skill the back end is done. The back end is locked. Don't even think about changing the back end. Alright. When you guys say go, I'm gonna hit go. Hit go. There it is. Alright. Twenty minutes, boys. I'm I've gotta upgrade my lovable account because, apparently, Lovable doesn't allow custom MCPs unless it's It sounds like an excuse, bro. Already an excuse. Alright. One pro. I'm gonna spend this whole time setting up my custom MCP. Oh, no. Why can't I see the text here? Oh, that's not good. I'm going react on this one. Why can't I see this, though? I just got this new Mac, man. Still, if you guys beat me, that will be b y since we're just throwing excuses out there. I am still signing a still I'm still going there by a sign up for a year. I can't even get the I can't get the tan stack start script here from the their website. Start. Connection failed. Gotta love it. Alright. Well, mobile mobile is not really communicating with me right now. I'm having trouble. I hope you guys are having fun. I'm having a blast. Matt, is your keyboard connected yet? Yes, John. Alright. Let's see. I got my build plan ready. Let's see if this looks good. No. I don't think Lovable's gonna be the the move. Not great, Lovable. Not great. Alright, guys. I've clicked build. We're gonna we're gonna get v one running. What if we finish in ten minutes? Does that give us bonus points? I should be done setting up my MCB by then. So Matt, why don't you have a rabbit in your background? Brian and I both have a rabbit in the background. I have several. Because I like to have my rabbits in my point of view. So not in the background. Alright. Well, local's not gonna work with me on the MCP, so I'm gonna have to roll. You just just gonna have to roll. Can I just copy, Jace? You could just tell it to use the SDK. I think you can copy. I think you can also just give it your API token and just give it access. I think I've done that before. You'd be like, hey. I'm an API token. Didn't know we'd get excuses so quick out the gate, though. You know, usually, I'm using, like, whisper for this. Go for it. Okay. Then we know what you're gonna build. Right? Just meet your just meet yourself while you mouth it and then unmute. I need a front end that looks great. Make no mistakes. Make it look beautiful. I I hope John's idea is something that looked like a pot of gold and a leprechaun. No. But that would have been good. Dang. I think mine'll be good, though. I I have no idea what this output of this prompt is gonna mean. I'm actually kind of terrified. I know. I'm just reading my typing, and it's freaking terrible. That's why he's was The good news is AI doesn't care about typos. What were the tags we put in here, by the way? Research writing, design, automation, productivity, coding, marketing, and data analysis. We should upload these to to, like, Vercel or Netlify or something afterwards so people can play with them, the front ends. That should be that should be solid. Here's the file being created right now. It's called bang effect dot JSX. Bang effect. You may win this, man. I think you guys had more time to think about the ideas here. Don't yeah. I don't have a solid idea. Lost trying to set up the MCP server in Lovable. Oh, that's a risky run. You've got what what did we got on the clock? Thirteen twenty one. Plenty of time. Just just come on, Claude. Don't film me now. I like you, dawg. We got this. We got this. Oh, first preview. It looks good. Did you one shot it? Oh, damn. This is okay. You guys are gonna lose. He's blowing smoke. I know. He's he doesn't have anything. It's just a blank screen. I wish you guys could see this. No. I'm trusting lovable with my lunch. The sound effects are amazing. You have sound effects? The only problem is I can tell the words are backwards. I'm like, why are these words backwards? What, wait. Tags. One, two, three. I like how there's, like, duplicate tags. Lovable has gotten a lot better on the user interface since all of us use it. Like, it quite looks nice, except for the whole connecting and MCP thing. I wish you guys could hear these sound effects. I think I can hear a few. Did you do freaking lasers? Oh, it's looking so much better now. Oh my gosh. This looks so good. It's a risk. It's poker face. No. You know it's a good a good project when you are playing the game that you've built. I'm still waiting for dependencies to resign. Alright. You know me, Ben. If you enjoy watching people slide by on the skin of their teeth, please tune in to Directus TV for where yours truly embarrasses himself twenty four seven by trying to build apps. Oh god. This is going horribly wrong. I should not have should not have done React that is using way outdated stuff. We're just I I don't know. Tam, if you watch this, man, I don't know what happened to the website today when I was recording this, but I think this is gonna be Oh, there it is. Finally. It's okay. You got something? Must have been I the site must have been broken, man. I don't I don't know. Okay. Now now it's playing its playing its game on its own. This is where where it's fun to watch Cursor play its own game. You got your AI playing an AI? K. What are you doing over there? Who? Me? Just kicking ass? I'm gonna have to rely on a one shot here. I think I might have this with a one shot. We'll see. It's still it's finishing up the build. This is building out the comp I'm building the character select and move list UI right now, so we're gonna have to wait a a second. Did you create an RPG, John? No. I created a a video game. Do you wanna know what its theme theme is? I wanna wait and see. Okay. Leprechauns? No. Pecauns. I think I you know, if I had more time, it would be cool to do, like, an RPG style directory where you have to, like, level up to get the better prompts. I wonder if anybody would spend time or not. Probably not. I've been looking at this thing, and I'm just like, I don't even know. Do you have anything? Do you have a UI at all, Brian? I I have nothing on the screen. That's not good. How much time do we have left? Six minutes. Oh, man. Guys, this you might as well just stop. I'm telling you, you're gonna lose. I know Matt basically has nothing. He's gonna show a white screen. It's gonna be like index.html. We're gonna wait and ask I have a header. You have a header? I have a header. Match this line to us. Have have I been in the roost this whole time? Who knows? I feel like I do miss the old days where I just good old fashioned handwriting code. I probably yeah. On a 100 apps, hundred hours, like, the days of all, I woulda had something already that looked nice. I'm not super happy with what I've got here. I do only have a white screen. K. I put in one more prompt. I I'm three shotting it, and I think three shots might be it. I think my idea was here, but the execution is lacking. Oh, it even gave it a little tab icon. Nice. Alright. I think we got something. Matt, are you past the home page yet? Did you get logged in yet? Been spending all this time on the initial You spent fifteen minutes of the twenty minutes just upgrading to pro. I did. I might have credit card information floating out there, so remind me to cancel my cards. Alright. We got something working. Yeah. Without the MCP, I'm I'm not really able to pull the. Here's what here's what my cursor just said so you know what it's doing. The buttons are noticeably bigger now. Let me shoot a couple skills first, then test the collected view and detail pop up. That's a little hit into what I've built. Are we talking legend of Zelda or something? What? Dangerous to go alone. Sweet mother. He's still in there, or is that going too? I still got it. Started off on the wrong foot, man. Should not have, should not have tried to go React. Should just you gotta stick with what you know, man. Talk with what you know. Why did you go with React? You were trying to get fancy? I I've been using, I I got a project, a personal project where I've used, TanStack start lately. I'm trying to learn a little more about React so that I'm not one dimensional. Mhmm. Did not pay off today. Did not pay off today. Mhmm. Thirty seconds. Yeah. This is I've just spent all the time troubleshooting. Lovable doesn't seem to really work well with custom MCPs, so if anybody's curious. Fifteen, fourteen, 13. Who are you who's who gets to go first? Well, I mean, clearly, it sounds like John. Like, John's face means that he may have something really good. He might have. Yeah. Time's up. Yeah. Time's up. It's just finishing it's just finishing its build, but I've just been I only needed three prompts to get it to some pretty amazing. Okay. I mean, I've got something. It's not, Sounds like you have nothing. It's basically You go then. Me? Let me it's still finishing this moment. The junction table I faced. Alright. I just refreshed. We'll see. I I think that is I'll go then. I'll go. Yeah. Let's see what you got, Bryant. I'm excited to see this header. Screen. I didn't have time to work on the design. We have a skills directory. I was going for, like, the brutalist theme. It didn't really get. I I had literally no time to worry on the design. So the the tags do work, which is nice. We can search. So we search for SEO. That works. We can load up the skill. We see it being loaded here. You know, you got the nice little subtle detail, the little arrows instead of the standard bullet points. You got the the fake install. And then the coup d'etat. Is that what the I don't know. I don't know. Matt has surely lost. Nice. Stakes. Pretty badly. Oh, so you can submit skills from the front. That's pretty good. Pretty good. Oh. No. Failure. Yeah. I did get a chance to to fix that. If I had, like, two more minutes, like, whatever it is that you're doing right now while I'm demoing mine. I'm not doing anything. This looks good. I mean talking to Matt. Design wise? What like, I mean, I think it looks solid. I would probably call it I I wouldn't have gone, like, the dark mode design. Right? I would have gone light mode, and I would've, like, just had it be very in your face. Would have been nice to have, like, the login to be able to submit this, but, again, just good to me. Wrestle with it too much. It's more than I was expecting. It's it looks good. Yeah. Call it Hill Skills. And it's got, like, kind of a nineties vibe to it. I dig it. Alright. I guess I'll go next. We'll see what happens here. But, alright. This is called, AI skills pro skater presented by AI Amigos. So the concept I wanna say I'm impressed that you have something. I thought you were gonna show us nothing. Maybe maybe I was in Parker Face the whole time. But, anyways, the concept here is you start by tag. So you pick your, it's not really skateboarding, but the concept was Tony Hawk's Pro Skater. But, it was supposed to be whoever like, I wanted to do it by people in an organization. So engineers would wanna probably see the engineering skills, marketers, that sort of thing. So starting with tags, which is something I haven't seen before. And then in true Tony Hawk fashion, ideally, what I was trying to get was to unlock the skill, you have to do the, like, combo thing with the keyboard. And, it would do a little trick sort of thing. If I had more time, this would be like a full Tony Hawk style game where you can unlock skills that way. But, have all of the ones here. It just randomly put some, like, easy hard sick, grade on here, but I still pulled in all the overview platform tips and stuff. Not as nice as I was, like, hoping to get, but Where's the skateboards? I know. There is no skateboards. I was I I I was disappointed. I think you've got the beginnings of something. I think this is cool. I think this is a good idea. I think it's cool. Yeah. I was trying to go for the most non directory directory type, but, yeah. That's what that's what I That's cool. Yeah. This part, I like, man. This looks this looks super nice. It's totally It's even got the Like, I like that they're even floating. It's like the the CRT Yeah. Tube TV effect to it. Mhmm. Yeah. I've been lately. But yeah. So that's the end. That's AI skills for a skater. Yeah. I think it's for a skater. Alright, John. Alright. Let's see what you got, John boy. Alright. Let's do this. I went you know, if you've if anyone's followed me on LinkedIn, I did the whole Nintendo theme garage inventory. So I kinda kept it Nintendo themed, and I went with Duck Hunt, although Skills Hunt. So you come here and you you shoot the little you you can't hear it or maybe you can't. There's little noises. There's a little creature down there, and then you can you can view the collections that you've got. You've got a little the only thing I would do is make the fonts bigger. You can see all of these things here. But you can go back to the game and you shoot the little you shoot the skills that you want. The dog is So there there is noises. I don't think you can hear them, but maybe the recording will show them. But there is noises when you're shooting things, and that's what I did. You got your skills there on the right that you've collected. You've got your skills here, and then you can view what those skills are. Alright. So, yeah, shoot for ask questions later. Actual game I already got that one. Damn. I need a new skill. There we go. Bang. I love gamifying the skills. Yeah. You can't just get the skills. You gotta work. You gotta work for it, man. Anyways, that's what I know. Alright. I mean, all skill one. Honestly, I think way better than all of us anticipated, at least for for me. I mean, we did all of those Friday It's a working game. Twenty minutes. Yeah. Very nice. Yeah. It was twenty minutes on the clock and mostly fighting with my, connection and react, and John was his master of vibes over there. So Master of vibes. I mean, if I had to speak Does this go to a vote between us, or does it does it go to the audience? I mean, I I think it's a good question. Right. I think we all know who won this one here internally. Yeah. Yeah. Yeah. Yeah. As much as it pains me to say it, very creative. Very creative. Mhmm. I like shoes. I think they're all good. I think they're all great, dare I say. They need all they all need adjustments, obviously, if you were to move this to production, but I think they were all great ideas in different manners. Yeah. And then you combine them all. And that's I mean, I think the best thing is that, like, I'm imagining you have, like, a full team of folks operating off the same like, you like, at companies I've worked at, they've they've always been hackathons. But, like, opening up a hackathon to the entire team to be able to work off the same base set of data, I think, is just the really cool thing here. It's, like, everybody can bring ideas to the table now. But, yeah, overall, pretty cool. Well, cool, guys. I think that's about it for this session. I appreciate everybody hanging in there with us. It sounded like if you're just listening to this, it was awful, but, you know, the end result has been solid.",[57,58,59],"d7e6adc3-560c-4e06-8e32-ae37949193c0","5abcac60-2f3d-472c-9976-e013163d7a80","5576f5a9-383a-4f43-80c0-6356e6aeaf79",[],1781213196152]