[{"data":1,"prerenderedAt":73},["ShallowReactive",2],{"leap-week-ai-in-the-enterprise":3,"leap-week-ai-in-the-enterprise-next":59},{"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":43,"season":44,"seo":10},"7271f0be-33fd-4cea-b7bd-9c63e74969e1","ai-in-the-enterprise","1176284514","How are enterprises giving teams access to AI while managing governance, data privacy, and risk?","ccb9efaf-5355-49a8-a6b4-5088bd5200a7",45,null,3,"2026-03-27","AI in the Enterprise","\u003Cp>Speaker 0: Excellent. Today, we have another episode of Bridging Bytes. It's been a minute, since we did our last one, but I'm extremely excited, to have two folks here to run through some exciting, topics around AI in the enterprise. So today, I'm joined with, with Holger Hammel, who joins us from HelloFresh. I actually think the intro came from Emma, our VP of marketing, from your time at Ivan.\u003C\u002Fp>\u003Cp>But, really excited to have you here, Holger. And then also Peter Bell, CTO, founder, head, of AI over at Gather dot dev. And I think you're actually a little bit closer, Peter. You're in Absolutely. In New York, in the city?\u003C\u002Fp>\u003Cp>Speaker 1: New York City. Just north of the city in Westchester.\u003C\u002Fp>\u003Cp>Speaker 0: Nice. And, Holger, you're you're over in Berlin. Right?\u003C\u002Fp>\u003Cp>Speaker 2: Yeah. Correct.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. We have slightly different time zones and, and lighting probably in the background. But, Holger, do you wanna give a quick intro in terms of, like, what you're doing over there at HelloFresh?\u003C\u002Fp>\u003Cp>Speaker 2: Yeah. I like to. So first, thanks a lot for having me. It's a super interesting topic, of course. And so I'm VP engineering at HelloFresh.\u003C\u002Fp>\u003Cp>I'm leading there the consumer alliance. So it's, like, about three hundred three hundred engineers across different tribes and squads, and we're covering the, client side applications, right, the web and the app page, and first layer of back end, aggregation layer, and data science, data engineering a little bit, and customer care. And, basically, you know HelloFresh. Right? We have, we have, the meal kits in, The US and in European countries, 18 countries overall.\u003C\u002Fp>\u003Cp>And we deliver fresh meals to everybody at home, and we have as well ready to eat meals. And it's a it's an interesting, setup where we have, physical products delivered to to customers, right, really, like, food products with their own kind of challenges in a sense, right, how to do this. And then we have the digital product on top, that we work on. And my key part here is to really, solve for personalization, right, to really create a digital experience that brings the best customer experience. And, and as well, and of course, the topic for today is, like, how to make this very effectively, and kind of make best use of AI internally for engineering departments, but as well of of introducing it into our products, to, for\u003C\u002Fp>\u003Cp>Speaker 0: Love that. Very, very relevant. It's always good to hear, you know, anything on the hardware or the physical side, with what you're doing at HelloFresh. It's that that's the moat these days, in terms of, in terms of AI, and everything that that it's it's gobbling up. Thank you, Holger.\u003C\u002Fp>\u003Cp>Peter, could you maybe tell us a little bit about Gather Dev, and what and everything else that you're you're doing on your side?\u003C\u002Fp>\u003Cp>Speaker 1: Absolutely. I was an IC, wrote software for many years, then I became an engineering leader, CTO of a bunch of startups. I ran engineering at general assembly, built teams up to about 50, so fairly small scale. These days, I'm doing two things. I am writing the book, Scaling AI Adoption and Engineering.\u003C\u002Fp>\u003Cp>Basically, AI for CTOs. And it sure. We're gonna cover software factories and verifications and all the tech, but the hard part is how do you manage stakeholder expectations? How do you do the change management? How do you teach people when the curriculum changes every three days?\u003C\u002Fp>\u003Cp>Like, how do you do that at scale? And so we have these niche in person and online communities for founding CTOs. You're like an individual contributor or a team lead. You got a team of less than 10. Startup CTOs venture back to a scaling from 10 to about a 100, like, how do I hire my first DMs, and how do I do performance management, and how do I create a consistent hiring, process, things like that.\u003C\u002Fp>\u003Cp>And then VPs and CTOs at scale who are running also, like, a 100 up to a couple thousand where it's generally good news is you have the systems in place. Bad news is they are now what is stopping you from getting the work done, that and the people. And so I spend a lot of time talking with engineering leaders about how do you scale adoption, not only what are the good technical practices, but how do you actually do the change management, which will be the hard part.\u003C\u002Fp>\u003Cp>Speaker 0: Love that. Yeah. I mean, that is super relevant for some of the questions we'll get into today. I I, you know, built teams to 50 sounds small maybe, but, that's still where we are, and that's our whole team, over here at Directus. But, and I think you also do CTO hour, right, part of O'Reilly's Yep.\u003C\u002Fp>\u003Cp>Speaker 1: So couple of the things I'll do is I do a quarterly CTO hour for O'Reilly. The most recent one, I got to interview Camille Fournier, the author of Manager's Path, and Ali Adasan, who's the CTO at Dropbox. It was a lot of fun. And then twice a year, I get to, go to KubeCon, the Kubernetes conference. And for CNCF, I get to facilitate the executive summit along with Kelsey Hightower.\u003C\u002Fp>\u003Cp>So that's always a blast.\u003C\u002Fp>\u003Cp>Speaker 0: I love that. Is it KubeCon or KubeCon? Because it seems like Kubernetes, it'd be KubeCon.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. You know what? I don't even know how to pronounce it, but with my accent, people always think I'm saying Q con, which is great, but a whole other thing. I I was at QCon AI recently presented something there, but so I I just try to make it clear, which one I'm talking about because my my computer always gets it wrong for sure when I'm dictating.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Well, say it with confidence, and I guess it doesn't matter too much. So let's kick it off. I think sort of like just table stakes, you know, setting the foundation here. In terms of day to day AI, Holger, maybe you can kick this one off, with your thoughts.\u003C\u002Fp>\u003Cp>Where where is your organization today with AI? Like, what does it actually look like for your employees, for your teams? You know, how's how's that actually operate?\u003C\u002Fp>\u003Cp>Speaker 2: Yes. So I think we are in a crossroad. Right? In a we're in a in a in a in a situation where maybe some of the teams are as well, but, we are, very good, I think, already in adoption of AI on an individual level. So we have all engineers, most of, you know, using some form of AI.\u003C\u002Fp>\u003Cp>We have everybody using some form of AI to kind of improve the documentation, their workflows, and so on. What we recently, what we started, beginning of, the year or a little bit before as introducing our framework for the whole product development life cycle. And that's the interesting part. That's where we wanna go to, right, is, really understanding how we solve for not just cycle time, like, making one engineer faster, but, like, getting the flow for the whole squad to a complete new level. Right?\u003C\u002Fp>\u003Cp>It doesn't really make only sense to have one engineer being four times faster. Right? Not bad, but it's not enough. Right? And I think we're in the middle of we we as organization understood this.\u003C\u002Fp>\u003Cp>We are in the middle of transforming this. I really look at this as an AI transformation. We had DevOps transformation. We have AI transformation. Now we are in the in all of that on steroids, right, which is the AI transformation, and really trying to understand the cultural, the technical, aspects of this is something we're in the middle of right now.\u003C\u002Fp>\u003Cp>And it's something. Right? So you see engineers having, like, tons of agents running. You have designers creating pull requests. All of this is happening.\u003C\u002Fp>\u003Cp>Right? Just really bringing it together and super effective, on scale is where we are at and trying to solve for.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Is there is there a specific team or department that was know, doing something different than you expected in terms of, you know, that day to day? You know, may either moving faster, maybe they're moving slower, or just kind of a different different vibe altogether?\u003C\u002Fp>\u003Cp>Speaker 2: I think it was what what was striking was that we have a few, like, smaller businesses that, basically, we have, like, HelloFresh and and Factor, the established big businesses, if you want. And then we have smaller, separately managed, startups, kind of. Right? So GoodJob and Pets Table. And they are organized a bit more lean in a sense.\u003C\u002Fp>\u003Cp>Right? They kind of kept the startup vibe, and they you saw picking up. They saw we saw them picking up very fast. Right? Because it's just out of pure necessity in a sense.\u003C\u002Fp>\u003Cp>Right? So resources are scarce. So to get anything done, right, and they couldn't ask for, like, five more engineers, and they had to solve it. And they start solving it with AI, and they went very fast. Right?\u003C\u002Fp>\u003Cp>And they were the first one adopting it. And there was maybe less red band as well. Right? So they just, you know, did it. Right?\u003C\u002Fp>\u003Cp>And so they kind of act as the as the template or as the as the as the leaders down. Right? And where we got the inspire the whole organization are inspired by that speed and the that results.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. I mean, that's a huge network effect to be able to inspire other teams, you know, by seeing how fast that you can move and, like, maybe be more efficient. Obviously, that startup mentality is huge right now. We're seeing a lot of, you know, constriction across headcount in different orgs, you know, huge orgs that are saying like, oh, we can not just do things more efficiently, but, let's kind of adopt this more agile, you know, startup way of thinking. I think that's good across the board, you know, and, of course, sort of, sped up by by AI.\u003C\u002Fp>\u003Cp>Speaker 2: And the interesting thought I was just want to mention is really is the the constraints. Right? So the the the constraints of limited resources and still an ambitious motivated team that wants to run this, this was a key cut cut cultural aspect. Right? Because if you then have a big organization, very distributed, decentralized, like everybody does their part, this might not be there.\u003C\u002Fp>\u003Cp>Right? And so thinking about how to create this urgency, is a big part of that transformation, I feel.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Yeah. I mean, the only you have these, like, enormous organizations, and it's almost like you you can only shrink the overall headcount so much. Like, how do you actually get that startup speed when you have hundreds, thousands, tens of thousands plus of, employees? And I think, you you know, that sort of had you eat an elephant, you know, one bite at a time.\u003C\u002Fp>\u003Cp>If you kind of break up your teams and kind of make them very autonomous and give them that that skill individually, they can operate like many different, smaller agencies or smaller smaller, startups within a larger org. That's I love that. Peter, similar question to you. I guess, you know, what what does that AI look like day to day for you? And maybe more specifically, like, how has that changed, you know, over the past year or so?\u003C\u002Fp>\u003Cp>Speaker 1: So for me personally, it's a very different thing. As an individual, I'm a solopreneur, and it's great. I have a team of eight named agents running twenty four seven. I built my own little orchestrator. I'm creating my own custom memory system, which, puts skills.\u003C\u002Fp>\u003Cp>It puts, prompts and agents into a GitHub repo, and some high value shared informally structured context also goes there. Everything else goes into a Postgres database, which allows me to have things like I can have agents doing message passing through the database so that I can have multiple agents collaborating without having to go through a pull push cycle that I would have to if all of my tickets or work to be done was in some kind of Git based system. So personally, it's great. I'm just starting to dig into software generation. I'm looking to spin up in the next two weeks my own clone of monday.com so that I can basically have an interface for playbooks for running the business and projects for either running experiments, building new playbooks, or incrementally improving playbooks.\u003C\u002Fp>\u003Cp>And the idea is deterministic workflows with small model and human of the loop steps. So that way I can firstly reduce the cost latency of the models by the like, I just need a simple classifier. I don't need to be running Opus four six to do that. And then on the other side, for the human in the loop steps, every single interaction is captured in a fresh session. So we keep them in the smart zone, very small percentage of context used, and it means that then I run a compounding loop so that every night, every agent reviews what it did and updates its own instructions and skills in a way that it would be able to do it a little better the next day.\u003C\u002Fp>\u003Cp>And I'm already this system I'm building, I I started two weeks ago, and it's already running the entire business. And now I'm digging into software factories and the verification and looking at the Dan Shapiro, looking at what OpenAI did internally, looking at what Harper Reid is talking about about building rich software factories. When I see people who are blessed and cursed with 200 or 500 or 2,000 engineers, it's a slightly different story. The first thing I see is that, the widespread thing we're still seeing is augmented rather than agentic development, cursor in IDE mode or Copilot. It's fine.\u003C\u002Fp>\u003Cp>And if you wanna be 30% quicker doing 10% of your job, which is actually writing the code, you're gonna get some incremental improvement there. I think the next step to go is agentic. So maybe you've got six or eight small agents doing that you trust to write code or review code. That's fine too, and that you know, maybe you can be one and a half, two x as fast depending upon the code base, how well you prompt it, how much context you provide it. It starts to get interesting then you need to compound.\u003C\u002Fp>\u003Cp>I see people saying, oh, you know, like, I keep trying to do this, and it creates bad code. I'm like, have you told it what good looks like? Have you told it to remember what good looks like? And then if there's too much context where you're blowing the entire context window with all the rules, have you decomposed it so that you have one agent writing the code, one reviewing it? Maybe you have one reviewer that's looking for cyclomatic complexity, quality of naming, and architecture, and another one that's looking for security because you don't want one agent to have to solve for seven different dimensions in a single pass.\u003C\u002Fp>\u003Cp>It starts to get dumb. And once you so that then becomes interesting both on an individual level, even I'm getting compounding out of my small, agentic army. But once you start to do interesting things, you generally have a DevEx or platform team that owns the repo with the skills, agents, and prompts, and the key context, what you then do is you have this mechanism where other people within the org can fork that, tweak it, try to make it better, and then if, say, three people in the org give it a thumbs up, you then say, okay. This now gets owned by the head of AI, the platform team, the DevX team, and then you can share good practices across the organization without a priority knowing what the good practices are gonna be. Because Hochul is absolutely right.\u003C\u002Fp>\u003Cp>This is exactly the same as a cloud migration. You don't just give people a book on Kubernetes and say, are we there yet? Right? You actually want to have a program manager, and you want to have dashboards, and you want to have KPIs, and you want to have lunch and learns, and you want to have training sessions and Slack channels. You want to elevate people in your in your all hands.\u003C\u002Fp>\u003Cp>But at the same time, while you're doing all of that stuff, the curriculum changes every day, so you actually need to have a bottom up mechanism to get the best ideas from whenever they come. The one other thing I would say is, create a little bit of space for solo players. You can run way faster with this stuff solo. And if you have a small number of fire breathers, especially if they're building something that's not it's not the main way you charge all your customers. Right?\u003C\u002Fp>\u003Cp>This is the admin dashboard or this is the internal dev tooling where it goes down for an hour. It's not the worst thing. Those people will learn the practices that can then broadly be shared to the majority of your team. There's three buckets, and I promise to stop talking. The first bucket is the people who can't wait to do this more.\u003C\u002Fp>\u003Cp>They play with Steve Yegi's Gastown on the weekend. They are killing it. They're like, if I don't have 60 agents running, if I'm not blowing through 12 max plans at $200 each, personally, I'm not doing it right. Then there's the vast then there's some people who are honestly like, this is destroying the planet. This is destroying my job.\u003C\u002Fp>\u003Cp>This is miserable. I shall never do this. AI doesn't know how to write code as well as I do because I'm a Java developer. And they're gonna have a hard time. Hopefully, they will get AI infected like we got test infected for TDD.\u003C\u002Fp>\u003Cp>And the vast majority of people are somewhere in the middle saying, dude, I would do this, but you're telling me to read 50 blogs. Don't I have, like, features to shit? Tell me how to do it, and I'll give it a shot. Yeah. And you need to figure out how you deal with each one of those populations.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. No. And that's that's huge. And it's interesting, you know, hearing about, you know, minimum spend, like, you know, on on AI. Like, is that a motivator?\u003C\u002Fp>\u003Cp>You know, you just kinda say, like, you're not spending enough. It's also interesting, you know, hearing the comparison of, like, a small team, a big org down to small teams or solopreneur. You know, at the end of the day, you can ramp up the the minions as it were, and get the the team that you need quickly. But you mentioned sort of like piping your data into a database, you know, into Postgres. Like, whether that's you doing that with with the minions or it's a bunch of teams, I think that kind of bridges into the next the next topic pretty well, which is either way, we're connecting data up to these these services, up to the tools that we're building, whether it's internal and maybe, like you said, it goes down for a second, Peter, and maybe that's not the end of the world.\u003C\u002Fp>\u003Cp>You know, maybe it is, depending on the usage. But in terms of governance, like, how how do we find that balance? Holger, I'm gonna throw this back over to you. When we think about people moving fast, you know, getting that agility, we want people building things and experimenting, optimizing their workflows. How do you think about guardrails?\u003C\u002Fp>\u003Cp>Because, obviously, you know, people are just kind of out there, you know, building, but, you know, data is is crucial. Data is the backbone of all of this. And if that gets leaked, if that, you know, isn't know, given given the proper RBAC or permissions, in these quick systems that are being built, you know, how do you how do you speak to that across your org?\u003C\u002Fp>\u003Cp>Speaker 2: That's a very interesting question. And there's probably no final answer, but I think one, of course, we we have a high responsibility for for the data of our customers and our employees, and that's not gonna be sacrificed or, like, a change. Right? So we need central governance for for data, and we have it. So it's basically centralized through the existing teams that we have, like security or data privacy teams, reliability teams, ops teams.\u003C\u002Fp>\u003Cp>And we have a central centralized Gen AI team, basically, infrastructure team that kind of, is owning owning those things. Having said this, traditionally, you have guard rates around cost or maybe, I don't know, who can who can access to this. And and this, we we deliberately said we wanna, while protecting all the all the, PII data and relevant data very clearly, we want to open up basically everything else and and make people just try it out. Like, remove every red tape that we can. Right?\u003C\u002Fp>\u003Cp>And I think, you know, we were not looking at, should we have Corsa and Cloud Code and Gemini and whatnot. Right? We just take it all, cut it out, and see what sticks. Right? And if there's a new kid on the blog, we probably take this on as well and then see later, how we decommission it.\u003C\u002Fp>\u003Cp>It's just more important to get people excited, to get people working, and then, there will be a time for consolidation. And, you know, it's it's kind of ongoing. It's ready. We know more about, like, how how to manage context, you know, how to have a memory that is not, tool dependent and agnostic and stuff like that. So that helps.\u003C\u002Fp>\u003Cp>But I think that's the key part here. Right? It's removing red tape where we can while, while protecting the the data that we have to.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. I mean and there's also I mean, shadow IT, obviously, and now we have, you know, shadow AI. Is that an issue for you? Like, people just bypassing, like, you know, the the research shows that, you know, when leadership is actually shaping all of this, AI governance, you're getting better value for your company. But you're still gonna have your ICs, your, you know, pretty much anybody go out and say, I'm gonna use my personal account, my personal service.\u003C\u002Fp>\u003Cp>You know, is what is your solution to that? You know, is that that that's, like, sort of a sidecar risk. You know, do you just lock that down? You just make it easier to use the approved, like, internal services? Or, you know, how do you avoid that sort of issue where people are using things that just aren't even approved and they're just kinda going rogue for better or worse?\u003C\u002Fp>\u003Cp>Speaker 2: Yeah. So the honest answer is you cannot fully fully, kind of mitigate this, I would say. Right? And I think, I think we have a very compelling offer for people. Right?\u003C\u002Fp>\u003Cp>The you know, you have, you have, plans for for, for Copilot, for Gemini. There are. Right? So I think, if people now still choose to copy paste something in their own, Chativity, you cannot really, do much about it, I guess. But I think it's it's about training.\u003C\u002Fp>\u003Cp>Right? Kind of, we have a policy for for JNI. Right? We very early on had a had a policy and a got got guidelines, of course. Right?\u003C\u002Fp>\u003Cp>So that that people understand this. And I think especially if you're not in engineering, we want everybody to become a builder. Right? But not everybody in outside of engineering might have the context and might be security, sensitive also. So I think it's a big part of, like, training and making people aware of the risks and, what IP means, and, you know, what are the difference between private and enterprise accounts and stuff like that.\u003C\u002Fp>\u003Cp>So that is very, very important. But I think we tend to be more encouraging of using the tools than kind of limiting for now.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Well, Well, I would hope that everybody's being security sensitive. You know, that's that's obviously the name of the game. It it only takes one incident, one issue, and everything goes down quickly.\u003C\u002Fp>\u003Cp>Speaker 2: Wonderful.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Peter, I guess, you know, different different sort of, route to, you know, think about this. But, you know, are you doing are you kind of baking that into your process and how you're thinking of you know, if if we're saving data into a database, are you just piping it straight in? Are you piping it straight out? Is there are there any sort of things that you're thinking of when you're you're building out these these systems?\u003C\u002Fp>\u003Cp>Speaker 1: So one of the things I like is that, in many ways, I can build, an engineering infrastructure that feels like a a bigger company. And in fact, I need to. I find that the guidance, the onboarding, the design systems, the less decisions an agent is allowed to make, the more processing it can bring to those decisions, and the less likely is it it is to make random bad ones. I I was just chatting with a a few CTOs running larger teams at breakfast yesterday, and one of the common threads was how can we go more heavily into design systems, into standard patterns, into processes, into minimizing the number of decent technical decisions that need to be made so that the agents don't get don't don't have too too much space to keep getting it wrong. So I I think, managing the number of decisions for the agents, I think being very thoughtful about data, you need to think through I think we're gonna see a lot more about agentic roles and permissions.\u003C\u002Fp>\u003Cp>It turns out that if you just have a prompt that says never, never, never merge your work domain, It works almost all the time. If, however, you literally, I know somebody who has 50 agents and each one has their own GitHub account. And the nice thing is that you just put branch based permissions on main, and they are unable to merge it in until either a human or some other agent has done it. So I think you need to be very thoughtful. Don't assume that the agents will do what you say or even something reasonably close to what you say and have the guardrails.\u003C\u002Fp>\u003Cp>And I think it also comes back to all the classic good engineering practices. You know, you should decouple release from production by using feature flags so that you can canary rollout. You can load test things. You can roll things back. All of these are important.\u003C\u002Fp>\u003Cp>And I think to the biggest story is, like, whether it's shadow IT or, like, how you use this, I recommend picking a lane. If you think about it, technology adoption life cycle, crossing the chasm is like a 30 year old book, and it's still true. You can have innovators, early adopters, early late majority, and laggard. And, you know, that's okay. If you run a bunch of gyms, if you run grocery stores like physical plant, if you run ski resorts, maybe you can just wait till Microsoft figures it out and just tell everyone to go use Copilot.\u003C\u002Fp>\u003Cp>And in a year or two, you'll be a little bit faster. That's actually perfectly okay. You're gonna lose anybody who wants to work with AI, so it's gonna be a negative selection in terms of the team you get, but you're gonna keep most of the people who know how your systems work, and it'll keep them happy. And shadow IT is gonna be a problem for a while, but, honestly, all the people who wanna use shadow AI are gonna leave your company anyway. So it just is what it is.\u003C\u002Fp>\u003Cp>And then you can go to the other extreme. You can be like a Toby at Shopify. Right? You can be like, hey. Not only do I want to say we're AI first very early on, there was one time where he got the head of his, AI team to say, I want the 20 people using the most tokens.\u003C\u002Fp>\u003Cp>Promote them. Is that a good business decision where those tokens being used usefully? Doesn't matter. That is a cultural concept that that creates a sense of we want to be innovators. We want the people who wanna be innovators, and that's what we wanna attract.\u003C\u002Fp>\u003Cp>And if that costs, you know, 500 k in bonuses, it was totally worth it. So I think you need to figure out where you play on that. But then even if you're an f aider or early adopter, you still need education and enablement. Make it as easy as possible, like Holger said, for them to use. If you're just like, what do you mean we do AI?\u003C\u002Fp>\u003Cp>We have Copilot licenses, and you can request one. You're gonna get a lot of shadow AI and lose a lot of great people.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah.\u003C\u002Fp>\u003Cp>Speaker 1: If you work hard but have reasonable red lines that explain why sending PII to service in China may not be in your customers or your business's best interest, That kinda makes sense. And providing you teach people what's going on and especially for the nontechnical folks in the org, that you give them an understanding of why the rules exist and good ways to be a good corporate citizen and still actually get stuff done.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Oh, absolutely. It's interesting thing. I've never even heard of, you know, just finding who's using the most tokens and promote them. Like, I can imagine people starting to use their agents just to run other agents just to ramp up the token spend.\u003C\u002Fp>\u003Cp>Over here, you know, building AI into our platform, it's all about, you know, how do we optimize this and get the tokens down. But I guess different different strokes for different folks. You had mentioned sort of nightly builds and sort of feature gating, you know, behind the flag. That's sort of kinda leads us into the next topic here, which is, you know, we we're seeing a million apps, you know, flood the market. We're seeing all these cool POCs and pilots and experimental, you know, things internal and external.\u003C\u002Fp>\u003Cp>Like, what what gets that into production? Like, what makes that a viable application or something that, you know, you kinda mentioned it works almost all the time, you know, and, you know, we're gonna wrap it in the secure prompt, like, never ever ever do this, and it still does it, you know, on occasion. That doesn't fly in production when you're dealing with mission critical systems, when you're, you know, really building for the customer, and externally. Beyond sort of, like, the the point that you mentioned, Peter, maybe you can kinda is there anything else that you think is is critical? Like, we talked on governance and sort of maybe permissions.\u003C\u002Fp>\u003Cp>But what else helps get you to that production scale and resilience?\u003C\u002Fp>\u003Cp>Speaker 1: So there's a lot of things. A really good starting point is to remember we already have nondeterministic systems building software. They're called humans. There's just slightly different parameters when you deal with these new nondeterministic systems that are building software. It's it's an intern, and by the time you get to the end of the context window, it's an intern who's been on Adderall for two nights and is starting to forget stuff.\u003C\u002Fp>\u003Cp>So you have to be thoughtful about how you engage with your guardrails. My assumption that to take the extreme example is that my only job is to provide training for my future robot overlords. Do I believe that's true? Do I believe there's no rule for humanity? Hopefully not.\u003C\u002Fp>\u003Cp>But if I use that as an operating assumption, then there is no part of there's no time that I interact with an agent that I'm not doing it in a thoughtful way to improve their ability to give me a better outcome next time. So the first thing is you just rinse and repeat. You, like, you wanna build an entire software factory, right, where it goes from you just say the vision and it's in production with no human no required human gates between the two. All you do is you tell an agent to write some code, you tell it what was wrong, you make sure that it captures that context, and you tell it to go again. And you keep repeating that loop time and time and time again until you build more and more validators, adversarial reviews, and other elements that will reduce the likelihood.\u003C\u002Fp>\u003Cp>And you know what? I've taken that in production database in my life. Like, humans get it wrong too, but it's much less likely the more revs you do around that cycle to figure out what does good mean, what does bad mean, and how do we put, both tooling and and also just prompts. You've got the prompts, so you've got the tooling, you've got the context, but also you do need, I think, permissioning systems. You need to think about the blast radius.\u003C\u002Fp>\u003Cp>And then all the good stuff we've always been doing, observability, monitoring, alerting, I think there's great opportunities now for or self healing software is a fancy word for incident response where the LLM knows enough about the system to propose a a move forward. I think that's going to become mainstream. And, also, just stuff that we're doing fifteen years ago, game daying, resilience engineering. How can you make it so if your cart goes down, at least you can still see the products? If your payment provider goes down, it can still save your cart and send you an email once it goes up.\u003C\u002Fp>\u003Cp>Assume that stuff's going to break more and create more resilient systems with less moving pieces so that even when things go wrong, the blast radius is constrained.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah.\u003C\u002Fp>\u003Cp>Speaker 1: One final thing I'd point is and the only other part is you can also we're not gonna review every line of code at some point that's going away. However, you can take a statistical process control approach to that, which is it's like ball bearings. You don't check everyone, but but if one of them is out of tolerance, you start checking them more frequently. And you do it risk adjusted, you're probably gonna test a smaller percentage of the PRs for your admin dashboard than you are for your main financial production flow.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. That's that's amazing. I I love that. It's interesting. You kinda bring up sort of like, oh, when this stops working, you know, you have this, like, graceful degradation, process.\u003C\u002Fp>\u003Cp>I my first thought is, you know, you don't know what you don't know. And you have these sort of non engineers building out systems, and the system might work. But if they don't know to prompt in, you know, hey, let's let's make sure that we have this, you know, progressive enhancement, graceful degradation, then that just won't be included there, which is which is interesting to think of. Like, the creator of, you know, the software still needs to be architecturally aware of, you know, software engine, design and architecture.\u003C\u002Fp>\u003Cp>Speaker 1: Two possible wonders I've seen for that. One model which I think is is good for now, which is I think we're going to see, you know, the two pizza team is not always going to be six to eight people now. I think you're gonna find triads in groups of four being able to get a lot of move a very long way and very fast when you have teams of twenty, fifty, 60 agents working with them. But there's going to be a combination of product and engineering skills. Somebody needs to know what we're building and why.\u003C\u002Fp>\u003Cp>Somebody needs to be thoughtful about, wait a second, latency, queuing, retime, split brain problems, all this stuff we deal with with large distributed systems. Those awarenesses need to exist. And then you might have some design skills or some data skills depending upon where the group is in your org and what it's trying to do. So I think in the early days, you make sure there's an engineer in the room before something you care about goes to production. I think, eventually, if you build a sufficiently rich software factory, you build those architectural concepts into the review process.\u003C\u002Fp>\u003Cp>Whether we get there and how close we get there, I don't know. So for now, I'm gonna keep a staff plus engineer having to look over the code just to be sure.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. How and how quickly we get there. Exactly. Holger, I'm I'm really curious to hear your sort of, like, take on this. You know?\u003C\u002Fp>\u003Cp>Obviously, Peter mentioned a lot of things that do get sort of these pilots up into production. But, you know, our platform direct us, for instance, you know, we power, you know, mission critical software where we're pumping out, you know, bandwidth at, like, nine nine thousand pizzas per second, you know, for for certain customers or whatever applications they might be, building. And it's funny to me thinking of, like, that that is, like, production grade, like, at scale. You can't just one shot, you know, or Vibe code your way to that. Like, there's there's a process.\u003C\u002Fp>\u003Cp>So I'm curious, you know, when you take sort of that equivalent at, you know, HelloFresh, like, what are you what are you doing? What do you layer on top to make sure, that your applications, you know, if you're if you're building in that way, you know, through AI, that it can handle that.\u003C\u002Fp>\u003Cp>Speaker 2: Yeah. That's a fantastic question. I think we, so so some some of the stuff, you mentioned, we do have to be one of the things, like, with the higher throughput, right, of of agentic AI or just, you know, teams speeding up. We see more, pull request reviews coming up with a new problem. Right?\u003C\u002Fp>\u003Cp>So so we need to find more, ways of, I think it's a really good idea, right, to kind of, find another abstraction level, basically, to look at, you know, risk management and and reviews and and and in a sense. Right? So that's quite interesting. So I think there we need to invest more. Right?\u003C\u002Fp>\u003Cp>This is something we need to do. What we have been doing is we have brought a few things to production, like AI products to production, which is kind of interesting. So we've been, maybe the most obvious case. We did go first as well. This is like customer care and having chatbots right there and experimenting there, a lot.\u003C\u002Fp>\u003Cp>And there, it's it's really it was we were kind of mid mid class last year or so. Right? So we were learning how to do this as we go. Right? And then managing, basically, the technology was not so much of the problem.\u003C\u002Fp>\u003Cp>It's more like managing the, really the the quality of it. Right? So all the different ways of of of how how a conversation with a human, the one deterministic, entity, right, can go, is, is, you know, not to underestimate. Right? And there's a lot of things that can go go wrong and how to how to design communication, the getting the intent right, and and keeping it safe, right, and and correct, is an is an ongoing task.\u003C\u002Fp>\u003Cp>And we, as we invested into, you know, the teams, that build up this messaging and then as well, the the automatic test. We experiment with AI as judged. Right? So, basically, having LLMs checking those those those flows as well. I think this this is very, very promising.\u003C\u002Fp>\u003Cp>Right? So this is one pattern that we see more is that you'll introduce, AI, maybe even different models, either for pull request reviews, having iterations on them, basically watching the manual testing themselves. Right? Like, having an eye look at, at how how it looks like and then doing pull requests by themselves. This actually works pretty, pretty well and has a very has a very high showed very high impact.\u003C\u002Fp>\u003Cp>So these are a few examples of where I can see. Right? So I think my my main concern right now is really getting the the the review part out of the way and and and on the other side, getting more, I don't know, engaging more UX and so on to create more experiments. So that's the other thing. I would like to have more, more variants basically tested earlier with the idea of having just already a more winning variant, basically, created into production, have higher higher signal, higher, success with more expensive AB testing in production.\u003C\u002Fp>\u003Cp>But why not having, like, 10 we we see this already, but I think we can do more 10 variations, test the prototypes, high high fidelity, bring it into the building, having tested before or with synthetic audiences. Right? Checking out quantitative analysis even on on scale if a business model works with a new hypothesis. So these are the things we're currently on and want to invest more into. So to get it on the front.\u003C\u002Fp>\u003Cp>Right? So getting better better signals earlier in the process, having the winning variance going to production faster and solving the PR revenue bottleneck problem. There are a few topics that we're currently working on. And then a few other things about personalization, improving the models, and as well creation from automatic creation of videos and and images, and we're creating a lot of, like, meals. And we have some AI support tooling that helps menu chefs, right, the creators of recipes to improve the process of getting this from idea to production, significantly faster than before.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Yeah. Well, it it that's I I love that. How how big is your org, Holger?\u003C\u002Fp>\u003Cp>Speaker 2: So my org is about 300, people right now. The overall HelloTech org. So how this is how we call this either, combination of all, like, tech technical staff, if you want, is, is about 1,000 people.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. And I I it's interesting. Like, all this only happens if we actually you know, the governance, you know, getting to production, you still have to have the people to make this happen. You know, going back to early, like, we can't just make the agents have the other agents work. Like, there's still human in a loop as Peter keeps putting it.\u003C\u002Fp>\u003Cp>It. You know, we think of, you know, Jack Dorsey and and kinda what happened over at Square. Like, switching to, like, the people side of this, you know, with an org of that size, like, how do you actually get the workforce to adopt, to use, to, you know, to actually have this happen in the first place. You know? Is I'm I'm assuming there's a fear piece.\u003C\u002Fp>\u003Cp>You know, we've talked a little bit about, I think, Peter, you had mentioned sort of building your for the future AI overlords. You know? There's a perception, I think, across some people that, like, oh, am I just digging my own grave and sort of, like, going out and and sort of building these these systems? But, you know, we have to lean in. You know, we can't just ignore it.\u003C\u002Fp>\u003Cp>So, you know, Holger, within your org, like, how does that work? Like, how do you make that happen?\u003C\u002Fp>\u003Cp>Speaker 2: Yeah. So I like to to frame it in a way that, I think HelloFresh is a really great way to experiment with this, learn this, apply it on scale, and just make yourself, stay relevant if you want. Right? And and be on on on top of the game there. Right?\u003C\u002Fp>\u003Cp>So we we really try we we try to make this work. We support people. We support the teams. And that's kind of the positive way of framing it on if you wanna play it a bit differently. If you don't do this, right, the gap between, what the what the what the industry demands or what where the where the top level is and and everybody else, it's getting bigger very fast.\u003C\u002Fp>\u003Cp>Right? So you you you risk of falling behind in a sense, and that's not limited to engineering at all at all. Right? It's like every function and software engineering in and around it at least is, has the same kind of, opportunities and and challenges in a sense. But I don't, like, underestimate the, those factors.\u003C\u002Fp>\u003Cp>They are they are real. Right? So and I think there's a system of incentivizing, helping people, making it, you know, as we talked about it, easy to figure to try out things. Right? But then as well, set up clear expectations, right, that, that we expect every team to onboard into the new process and to to adopt to this agentic AI, for instance, use cases or to to have more variance in production now that we can do it, right, fairly easily.\u003C\u002Fp>\u003Cp>So there will be, the expectations grow steadily as well, to keep up with this. So that's the both of the side. Right? Enabling people, but as well-being very clear about expectations.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Oh, 100%. And it did like, Deloitte's latest report, I think, said something like only 20% of companies of orgs, actually, their their teams are actually ready, and prepared for AI. So, you know, getting that number up, hopefully, HelloFresh, you know, is in that 20%. But, but we're come we're coming up on time here.\u003C\u002Fp>\u003Cp>So I wanna kind of at least be a little bit more forward looking, with with sort of wrapping up here in terms of what's next. Peter, maybe I'll send it over to you. If you're looking out the next, like, say, one, one and a half years, like, where do you think the most change is gonna happen within AI? With an enterprise AI? I I think you go anything beyond that.\u003C\u002Fp>\u003Cp>I mean, even if you go out six months, the question mark start piling up so high. But if you were just kinda, you know, behind the sky, where do you think things are are heading?\u003C\u002Fp>\u003Cp>Speaker 1: Within the engineering org, I think it's pretty straight this will be I'm not gonna put a timeline on it. I've got a very good friend, senior CTO. He's like, you can make whatever prognostications prognostications, but don't put a timeline on it. You you'll thank me later. And I I think he's right.\u003C\u002Fp>\u003Cp>I have no idea. It will be much faster at smaller companies in greenfield environments, in places where you happen to have, an engineer that's just super passionate about this and pulls the team forward. It should happen more quickly in SaaS companies and in companies that are potentially have an existential threat from AI. As I said, you're on a chain of gyms, maybe it doesn't matter. So there's going to be uneven, speed of change, but I think, eventually, we're going to see a flattening.\u003C\u002Fp>\u003Cp>What the year of efficiency continues to resonate many years after. You could imagine a world where you have triads or maybe teams of three to four doing something similar to what a two pizza team does now so you can split them out and have more of them capturing more initiatives. I I think it's possible you might have five to 15 of those reporting straight up to a senior director or a VP where you have a team lead in each. I think we're gonna continue to see the collapsing of that, and I think the VP is gonna be spending a lot of time on the keyboard reviewing detailed specifications, verifications, harness improvements, as well as research and product elements. So I think we're going to see leaner engineering orgs.\u003C\u002Fp>\u003Cp>The good news is we need a 100 x the software, and, I think there are lots of opportunities for anybody who is passionate about either. And I think we need to be happy with this. Right now, we're all excited about AI engineers. And for me, that's basically harness engineers, people who are thinking about prompts and verifications and steps and gates. We need 8% of those.\u003C\u002Fp>\u003Cp>That's our platform or DevX team. Everyone else is going to become a product engineer who gets better at proposing experiments, writing specifications, and creating rich verifications while also being thoughtful architecturally. And I think those are the two roles that are gonna continue to grow and be incredibly impactful in the future.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Oh, I love that. Slightly different lens, hold on. I wanna hear sort of your your sort of take on where things are heading. But, on the enterprise, something that's very sort of near and dear to my, my heart, my job is sort of the commodification of the front end.\u003C\u002Fp>\u003Cp>Like, we're we're seeing that happen actively with the AI app builders. The value of software that's being built, where where does that value shift? As as the front end, as the facade becomes commoditized and and everybody can build that, where does it shift? I mean, down the stack to, you know, all the way back to the database or somewhere in between. Where do you feel, that's that's heading over the next year or so?\u003C\u002Fp>\u003Cp>Speaker 2: Yeah. It's a it's a it's a very good question. So basically, it's, I I still think, a a creativity and and being able to create a mode, the connection to your customers, building up this relationship is probably more important than ever. So you can be if you double speed on the wrong side of the highway, right, it it will not help you necessarily. So, so and I think that the the productivity gains that we have, that we're seeing now and that are super exciting, they will become a commodity at some point.\u003C\u002Fp>\u003Cp>I'm not putting a timeline on it either. But, you know, you you will have your great engineering teams, and product engineering teams. They will they will work on a completely different level. You still need to figure out what really works for a customer. And now the levels the field is, we're closer.\u003C\u002Fp>\u003Cp>Right? So I cannot rely on, say, a company like HelloFresh being faster, to production with some features than a small start up. So I need to be really, really smart about this. Right? I need to be have the most creative people writing the best specs.\u003C\u002Fp>\u003Cp>Right? So and, we have a lot of customers. We have a lot of customer relationship that is that is super valuable. That is not easy to copy, right, even with AI. So building on top of this and making sure that it stays this way and we stay on top of the innovation, that is what, I think, what still matters the most.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. I mean, bring it back to the people, you know, the as a designer, you know, focusing on the creative. You know, you can always it's really exciting right now because AI is able to take, you know, this massive context and, like, pipe it in and make everybody creative, make everyone a developer. But at the end of the day, you know, you have to be able to break outside of, like, what's already there and sort of just rearranging those things. So I love hearing the answer kind of always coming back to the human, to the core of, you know, creativity and, you know, even going back to earlier.\u003C\u002Fp>\u003Cp>You know, you can throw the the kill switch in there. You can do all the right things and still, AI will kinda find its way to work around it. So, it's really important to have people that understand the the proper architecture of of what we're building and and why those things are important. So, hopefully, we don't stray too far from from those first principles. Holger, Peter, it was really, really exciting, getting to just chat through, you know, and the enterprise with both of you.\u003C\u002Fp>\u003Cp>You know, hopefully, we'll get to chat again soon, maybe again on a bridging bites episode. But for now, anything that you guys wanted to sign off with before we wrap this episode up?\u003C\u002Fp>\u003Cp>Speaker 2: No. Just thanks a lot for for for having me. And, yeah. It was a pleasure talking to you. And, it's just exciting times, I must say.\u003C\u002Fp>\u003Cp>Speaker 1: Likewise, Ben. Thank you so much for the invite, Holger. It was wonderful to hear your insights. So much fun and great conversations. Lots for all of us to learn.\u003C\u002Fp>\u003Cp>Speaker 0: Awesome. Yeah. Well, absolutely. Thank you both for being here, and we'll see everybody, soon on the next episode. Alright.\u003C\u002Fp>\u003Cp>Thanks, guys.\u003C\u002Fp>","Excellent. Today, we have another episode of Bridging Bytes. It's been a minute, since we did our last one, but I'm extremely excited, to have two folks here to run through some exciting, topics around AI in the enterprise. So today, I'm joined with, with Holger Hammel, who joins us from HelloFresh. I actually think the intro came from Emma, our VP of marketing, from your time at Ivan. But, really excited to have you here, Holger. And then also Peter Bell, CTO, founder, head, of AI over at Gather dot dev. And I think you're actually a little bit closer, Peter. You're in Absolutely. In New York, in the city? New York City. Just north of the city in Westchester. Nice. And, Holger, you're you're over in Berlin. Right? Yeah. Correct. Yeah. We have slightly different time zones and, and lighting probably in the background. But, Holger, do you wanna give a quick intro in terms of, like, what you're doing over there at HelloFresh? Yeah. I like to. So first, thanks a lot for having me. It's a super interesting topic, of course. And so I'm VP engineering at HelloFresh. I'm leading there the consumer alliance. So it's, like, about three hundred three hundred engineers across different tribes and squads, and we're covering the, client side applications, right, the web and the app page, and first layer of back end, aggregation layer, and data science, data engineering a little bit, and customer care. And, basically, you know HelloFresh. Right? We have, we have, the meal kits in, The US and in European countries, 18 countries overall. And we deliver fresh meals to everybody at home, and we have as well ready to eat meals. And it's a it's an interesting, setup where we have, physical products delivered to to customers, right, really, like, food products with their own kind of challenges in a sense, right, how to do this. And then we have the digital product on top, that we work on. And my key part here is to really, solve for personalization, right, to really create a digital experience that brings the best customer experience. And, and as well, and of course, the topic for today is, like, how to make this very effectively, and kind of make best use of AI internally for engineering departments, but as well of of introducing it into our products, to, for Love that. Very, very relevant. It's always good to hear, you know, anything on the hardware or the physical side, with what you're doing at HelloFresh. It's that that's the moat these days, in terms of, in terms of AI, and everything that that it's it's gobbling up. Thank you, Holger. Peter, could you maybe tell us a little bit about Gather Dev, and what and everything else that you're you're doing on your side? Absolutely. I was an IC, wrote software for many years, then I became an engineering leader, CTO of a bunch of startups. I ran engineering at general assembly, built teams up to about 50, so fairly small scale. These days, I'm doing two things. I am writing the book, Scaling AI Adoption and Engineering. Basically, AI for CTOs. And it sure. We're gonna cover software factories and verifications and all the tech, but the hard part is how do you manage stakeholder expectations? How do you do the change management? How do you teach people when the curriculum changes every three days? Like, how do you do that at scale? And so we have these niche in person and online communities for founding CTOs. You're like an individual contributor or a team lead. You got a team of less than 10. Startup CTOs venture back to a scaling from 10 to about a 100, like, how do I hire my first DMs, and how do I do performance management, and how do I create a consistent hiring, process, things like that. And then VPs and CTOs at scale who are running also, like, a 100 up to a couple thousand where it's generally good news is you have the systems in place. Bad news is they are now what is stopping you from getting the work done, that and the people. And so I spend a lot of time talking with engineering leaders about how do you scale adoption, not only what are the good technical practices, but how do you actually do the change management, which will be the hard part. Love that. Yeah. I mean, that is super relevant for some of the questions we'll get into today. I I, you know, built teams to 50 sounds small maybe, but, that's still where we are, and that's our whole team, over here at Directus. But, and I think you also do CTO hour, right, part of O'Reilly's Yep. So couple of the things I'll do is I do a quarterly CTO hour for O'Reilly. The most recent one, I got to interview Camille Fournier, the author of Manager's Path, and Ali Adasan, who's the CTO at Dropbox. It was a lot of fun. And then twice a year, I get to, go to KubeCon, the Kubernetes conference. And for CNCF, I get to facilitate the executive summit along with Kelsey Hightower. So that's always a blast. I love that. Is it KubeCon or KubeCon? Because it seems like Kubernetes, it'd be KubeCon. Yeah. You know what? I don't even know how to pronounce it, but with my accent, people always think I'm saying Q con, which is great, but a whole other thing. I I was at QCon AI recently presented something there, but so I I just try to make it clear, which one I'm talking about because my my computer always gets it wrong for sure when I'm dictating. Yeah. Well, say it with confidence, and I guess it doesn't matter too much. So let's kick it off. I think sort of like just table stakes, you know, setting the foundation here. In terms of day to day AI, Holger, maybe you can kick this one off, with your thoughts. Where where is your organization today with AI? Like, what does it actually look like for your employees, for your teams? You know, how's how's that actually operate? Yes. So I think we are in a crossroad. Right? In a we're in a in a in a in a situation where maybe some of the teams are as well, but, we are, very good, I think, already in adoption of AI on an individual level. So we have all engineers, most of, you know, using some form of AI. We have everybody using some form of AI to kind of improve the documentation, their workflows, and so on. What we recently, what we started, beginning of, the year or a little bit before as introducing our framework for the whole product development life cycle. And that's the interesting part. That's where we wanna go to, right, is, really understanding how we solve for not just cycle time, like, making one engineer faster, but, like, getting the flow for the whole squad to a complete new level. Right? It doesn't really make only sense to have one engineer being four times faster. Right? Not bad, but it's not enough. Right? And I think we're in the middle of we we as organization understood this. We are in the middle of transforming this. I really look at this as an AI transformation. We had DevOps transformation. We have AI transformation. Now we are in the in all of that on steroids, right, which is the AI transformation, and really trying to understand the cultural, the technical, aspects of this is something we're in the middle of right now. And it's something. Right? So you see engineers having, like, tons of agents running. You have designers creating pull requests. All of this is happening. Right? Just really bringing it together and super effective, on scale is where we are at and trying to solve for. Yeah. Is there is there a specific team or department that was know, doing something different than you expected in terms of, you know, that day to day? You know, may either moving faster, maybe they're moving slower, or just kind of a different different vibe altogether? I think it was what what was striking was that we have a few, like, smaller businesses that, basically, we have, like, HelloFresh and and Factor, the established big businesses, if you want. And then we have smaller, separately managed, startups, kind of. Right? So GoodJob and Pets Table. And they are organized a bit more lean in a sense. Right? They kind of kept the startup vibe, and they you saw picking up. They saw we saw them picking up very fast. Right? Because it's just out of pure necessity in a sense. Right? So resources are scarce. So to get anything done, right, and they couldn't ask for, like, five more engineers, and they had to solve it. And they start solving it with AI, and they went very fast. Right? And they were the first one adopting it. And there was maybe less red band as well. Right? So they just, you know, did it. Right? And so they kind of act as the as the template or as the as the as the leaders down. Right? And where we got the inspire the whole organization are inspired by that speed and the that results. Yeah. I mean, that's a huge network effect to be able to inspire other teams, you know, by seeing how fast that you can move and, like, maybe be more efficient. Obviously, that startup mentality is huge right now. We're seeing a lot of, you know, constriction across headcount in different orgs, you know, huge orgs that are saying like, oh, we can not just do things more efficiently, but, let's kind of adopt this more agile, you know, startup way of thinking. I think that's good across the board, you know, and, of course, sort of, sped up by by AI. And the interesting thought I was just want to mention is really is the the constraints. Right? So the the the constraints of limited resources and still an ambitious motivated team that wants to run this, this was a key cut cut cultural aspect. Right? Because if you then have a big organization, very distributed, decentralized, like everybody does their part, this might not be there. Right? And so thinking about how to create this urgency, is a big part of that transformation, I feel. Yeah. Yeah. I mean, the only you have these, like, enormous organizations, and it's almost like you you can only shrink the overall headcount so much. Like, how do you actually get that startup speed when you have hundreds, thousands, tens of thousands plus of, employees? And I think, you you know, that sort of had you eat an elephant, you know, one bite at a time. If you kind of break up your teams and kind of make them very autonomous and give them that that skill individually, they can operate like many different, smaller agencies or smaller smaller, startups within a larger org. That's I love that. Peter, similar question to you. I guess, you know, what what does that AI look like day to day for you? And maybe more specifically, like, how has that changed, you know, over the past year or so? So for me personally, it's a very different thing. As an individual, I'm a solopreneur, and it's great. I have a team of eight named agents running twenty four seven. I built my own little orchestrator. I'm creating my own custom memory system, which, puts skills. It puts, prompts and agents into a GitHub repo, and some high value shared informally structured context also goes there. Everything else goes into a Postgres database, which allows me to have things like I can have agents doing message passing through the database so that I can have multiple agents collaborating without having to go through a pull push cycle that I would have to if all of my tickets or work to be done was in some kind of Git based system. So personally, it's great. I'm just starting to dig into software generation. I'm looking to spin up in the next two weeks my own clone of monday.com so that I can basically have an interface for playbooks for running the business and projects for either running experiments, building new playbooks, or incrementally improving playbooks. And the idea is deterministic workflows with small model and human of the loop steps. So that way I can firstly reduce the cost latency of the models by the like, I just need a simple classifier. I don't need to be running Opus four six to do that. And then on the other side, for the human in the loop steps, every single interaction is captured in a fresh session. So we keep them in the smart zone, very small percentage of context used, and it means that then I run a compounding loop so that every night, every agent reviews what it did and updates its own instructions and skills in a way that it would be able to do it a little better the next day. And I'm already this system I'm building, I I started two weeks ago, and it's already running the entire business. And now I'm digging into software factories and the verification and looking at the Dan Shapiro, looking at what OpenAI did internally, looking at what Harper Reid is talking about about building rich software factories. When I see people who are blessed and cursed with 200 or 500 or 2,000 engineers, it's a slightly different story. The first thing I see is that, the widespread thing we're still seeing is augmented rather than agentic development, cursor in IDE mode or Copilot. It's fine. And if you wanna be 30% quicker doing 10% of your job, which is actually writing the code, you're gonna get some incremental improvement there. I think the next step to go is agentic. So maybe you've got six or eight small agents doing that you trust to write code or review code. That's fine too, and that you know, maybe you can be one and a half, two x as fast depending upon the code base, how well you prompt it, how much context you provide it. It starts to get interesting then you need to compound. I see people saying, oh, you know, like, I keep trying to do this, and it creates bad code. I'm like, have you told it what good looks like? Have you told it to remember what good looks like? And then if there's too much context where you're blowing the entire context window with all the rules, have you decomposed it so that you have one agent writing the code, one reviewing it? Maybe you have one reviewer that's looking for cyclomatic complexity, quality of naming, and architecture, and another one that's looking for security because you don't want one agent to have to solve for seven different dimensions in a single pass. It starts to get dumb. And once you so that then becomes interesting both on an individual level, even I'm getting compounding out of my small, agentic army. But once you start to do interesting things, you generally have a DevEx or platform team that owns the repo with the skills, agents, and prompts, and the key context, what you then do is you have this mechanism where other people within the org can fork that, tweak it, try to make it better, and then if, say, three people in the org give it a thumbs up, you then say, okay. This now gets owned by the head of AI, the platform team, the DevX team, and then you can share good practices across the organization without a priority knowing what the good practices are gonna be. Because Hochul is absolutely right. This is exactly the same as a cloud migration. You don't just give people a book on Kubernetes and say, are we there yet? Right? You actually want to have a program manager, and you want to have dashboards, and you want to have KPIs, and you want to have lunch and learns, and you want to have training sessions and Slack channels. You want to elevate people in your in your all hands. But at the same time, while you're doing all of that stuff, the curriculum changes every day, so you actually need to have a bottom up mechanism to get the best ideas from whenever they come. The one other thing I would say is, create a little bit of space for solo players. You can run way faster with this stuff solo. And if you have a small number of fire breathers, especially if they're building something that's not it's not the main way you charge all your customers. Right? This is the admin dashboard or this is the internal dev tooling where it goes down for an hour. It's not the worst thing. Those people will learn the practices that can then broadly be shared to the majority of your team. There's three buckets, and I promise to stop talking. The first bucket is the people who can't wait to do this more. They play with Steve Yegi's Gastown on the weekend. They are killing it. They're like, if I don't have 60 agents running, if I'm not blowing through 12 max plans at $200 each, personally, I'm not doing it right. Then there's the vast then there's some people who are honestly like, this is destroying the planet. This is destroying my job. This is miserable. I shall never do this. AI doesn't know how to write code as well as I do because I'm a Java developer. And they're gonna have a hard time. Hopefully, they will get AI infected like we got test infected for TDD. And the vast majority of people are somewhere in the middle saying, dude, I would do this, but you're telling me to read 50 blogs. Don't I have, like, features to shit? Tell me how to do it, and I'll give it a shot. Yeah. And you need to figure out how you deal with each one of those populations. Yeah. No. And that's that's huge. And it's interesting, you know, hearing about, you know, minimum spend, like, you know, on on AI. Like, is that a motivator? You know, you just kinda say, like, you're not spending enough. It's also interesting, you know, hearing the comparison of, like, a small team, a big org down to small teams or solopreneur. You know, at the end of the day, you can ramp up the the minions as it were, and get the the team that you need quickly. But you mentioned sort of like piping your data into a database, you know, into Postgres. Like, whether that's you doing that with with the minions or it's a bunch of teams, I think that kind of bridges into the next the next topic pretty well, which is either way, we're connecting data up to these these services, up to the tools that we're building, whether it's internal and maybe, like you said, it goes down for a second, Peter, and maybe that's not the end of the world. You know, maybe it is, depending on the usage. But in terms of governance, like, how how do we find that balance? Holger, I'm gonna throw this back over to you. When we think about people moving fast, you know, getting that agility, we want people building things and experimenting, optimizing their workflows. How do you think about guardrails? Because, obviously, you know, people are just kind of out there, you know, building, but, you know, data is is crucial. Data is the backbone of all of this. And if that gets leaked, if that, you know, isn't know, given given the proper RBAC or permissions, in these quick systems that are being built, you know, how do you how do you speak to that across your org? That's a very interesting question. And there's probably no final answer, but I think one, of course, we we have a high responsibility for for the data of our customers and our employees, and that's not gonna be sacrificed or, like, a change. Right? So we need central governance for for data, and we have it. So it's basically centralized through the existing teams that we have, like security or data privacy teams, reliability teams, ops teams. And we have a central centralized Gen AI team, basically, infrastructure team that kind of, is owning owning those things. Having said this, traditionally, you have guard rates around cost or maybe, I don't know, who can who can access to this. And and this, we we deliberately said we wanna, while protecting all the all the, PII data and relevant data very clearly, we want to open up basically everything else and and make people just try it out. Like, remove every red tape that we can. Right? And I think, you know, we were not looking at, should we have Corsa and Cloud Code and Gemini and whatnot. Right? We just take it all, cut it out, and see what sticks. Right? And if there's a new kid on the blog, we probably take this on as well and then see later, how we decommission it. It's just more important to get people excited, to get people working, and then, there will be a time for consolidation. And, you know, it's it's kind of ongoing. It's ready. We know more about, like, how how to manage context, you know, how to have a memory that is not, tool dependent and agnostic and stuff like that. So that helps. But I think that's the key part here. Right? It's removing red tape where we can while, while protecting the the data that we have to. Yeah. I mean and there's also I mean, shadow IT, obviously, and now we have, you know, shadow AI. Is that an issue for you? Like, people just bypassing, like, you know, the the research shows that, you know, when leadership is actually shaping all of this, AI governance, you're getting better value for your company. But you're still gonna have your ICs, your, you know, pretty much anybody go out and say, I'm gonna use my personal account, my personal service. You know, is what is your solution to that? You know, is that that that's, like, sort of a sidecar risk. You know, do you just lock that down? You just make it easier to use the approved, like, internal services? Or, you know, how do you avoid that sort of issue where people are using things that just aren't even approved and they're just kinda going rogue for better or worse? Yeah. So the honest answer is you cannot fully fully, kind of mitigate this, I would say. Right? And I think, I think we have a very compelling offer for people. Right? The you know, you have, you have, plans for for, for Copilot, for Gemini. There are. Right? So I think, if people now still choose to copy paste something in their own, Chativity, you cannot really, do much about it, I guess. But I think it's it's about training. Right? Kind of, we have a policy for for JNI. Right? We very early on had a had a policy and a got got guidelines, of course. Right? So that that people understand this. And I think especially if you're not in engineering, we want everybody to become a builder. Right? But not everybody in outside of engineering might have the context and might be security, sensitive also. So I think it's a big part of, like, training and making people aware of the risks and, what IP means, and, you know, what are the difference between private and enterprise accounts and stuff like that. So that is very, very important. But I think we tend to be more encouraging of using the tools than kind of limiting for now. Yeah. Well, Well, I would hope that everybody's being security sensitive. You know, that's that's obviously the name of the game. It it only takes one incident, one issue, and everything goes down quickly. Wonderful. Yeah. Peter, I guess, you know, different different sort of, route to, you know, think about this. But, you know, are you doing are you kind of baking that into your process and how you're thinking of you know, if if we're saving data into a database, are you just piping it straight in? Are you piping it straight out? Is there are there any sort of things that you're thinking of when you're you're building out these these systems? So one of the things I like is that, in many ways, I can build, an engineering infrastructure that feels like a a bigger company. And in fact, I need to. I find that the guidance, the onboarding, the design systems, the less decisions an agent is allowed to make, the more processing it can bring to those decisions, and the less likely is it it is to make random bad ones. I I was just chatting with a a few CTOs running larger teams at breakfast yesterday, and one of the common threads was how can we go more heavily into design systems, into standard patterns, into processes, into minimizing the number of decent technical decisions that need to be made so that the agents don't get don't don't have too too much space to keep getting it wrong. So I I think, managing the number of decisions for the agents, I think being very thoughtful about data, you need to think through I think we're gonna see a lot more about agentic roles and permissions. It turns out that if you just have a prompt that says never, never, never merge your work domain, It works almost all the time. If, however, you literally, I know somebody who has 50 agents and each one has their own GitHub account. And the nice thing is that you just put branch based permissions on main, and they are unable to merge it in until either a human or some other agent has done it. So I think you need to be very thoughtful. Don't assume that the agents will do what you say or even something reasonably close to what you say and have the guardrails. And I think it also comes back to all the classic good engineering practices. You know, you should decouple release from production by using feature flags so that you can canary rollout. You can load test things. You can roll things back. All of these are important. And I think to the biggest story is, like, whether it's shadow IT or, like, how you use this, I recommend picking a lane. If you think about it, technology adoption life cycle, crossing the chasm is like a 30 year old book, and it's still true. You can have innovators, early adopters, early late majority, and laggard. And, you know, that's okay. If you run a bunch of gyms, if you run grocery stores like physical plant, if you run ski resorts, maybe you can just wait till Microsoft figures it out and just tell everyone to go use Copilot. And in a year or two, you'll be a little bit faster. That's actually perfectly okay. You're gonna lose anybody who wants to work with AI, so it's gonna be a negative selection in terms of the team you get, but you're gonna keep most of the people who know how your systems work, and it'll keep them happy. And shadow IT is gonna be a problem for a while, but, honestly, all the people who wanna use shadow AI are gonna leave your company anyway. So it just is what it is. And then you can go to the other extreme. You can be like a Toby at Shopify. Right? You can be like, hey. Not only do I want to say we're AI first very early on, there was one time where he got the head of his, AI team to say, I want the 20 people using the most tokens. Promote them. Is that a good business decision where those tokens being used usefully? Doesn't matter. That is a cultural concept that that creates a sense of we want to be innovators. We want the people who wanna be innovators, and that's what we wanna attract. And if that costs, you know, 500 k in bonuses, it was totally worth it. So I think you need to figure out where you play on that. But then even if you're an f aider or early adopter, you still need education and enablement. Make it as easy as possible, like Holger said, for them to use. If you're just like, what do you mean we do AI? We have Copilot licenses, and you can request one. You're gonna get a lot of shadow AI and lose a lot of great people. Yeah. If you work hard but have reasonable red lines that explain why sending PII to service in China may not be in your customers or your business's best interest, That kinda makes sense. And providing you teach people what's going on and especially for the nontechnical folks in the org, that you give them an understanding of why the rules exist and good ways to be a good corporate citizen and still actually get stuff done. Yeah. Oh, absolutely. It's interesting thing. I've never even heard of, you know, just finding who's using the most tokens and promote them. Like, I can imagine people starting to use their agents just to run other agents just to ramp up the token spend. Over here, you know, building AI into our platform, it's all about, you know, how do we optimize this and get the tokens down. But I guess different different strokes for different folks. You had mentioned sort of nightly builds and sort of feature gating, you know, behind the flag. That's sort of kinda leads us into the next topic here, which is, you know, we we're seeing a million apps, you know, flood the market. We're seeing all these cool POCs and pilots and experimental, you know, things internal and external. Like, what what gets that into production? Like, what makes that a viable application or something that, you know, you kinda mentioned it works almost all the time, you know, and, you know, we're gonna wrap it in the secure prompt, like, never ever ever do this, and it still does it, you know, on occasion. That doesn't fly in production when you're dealing with mission critical systems, when you're, you know, really building for the customer, and externally. Beyond sort of, like, the the point that you mentioned, Peter, maybe you can kinda is there anything else that you think is is critical? Like, we talked on governance and sort of maybe permissions. But what else helps get you to that production scale and resilience? So there's a lot of things. A really good starting point is to remember we already have nondeterministic systems building software. They're called humans. There's just slightly different parameters when you deal with these new nondeterministic systems that are building software. It's it's an intern, and by the time you get to the end of the context window, it's an intern who's been on Adderall for two nights and is starting to forget stuff. So you have to be thoughtful about how you engage with your guardrails. My assumption that to take the extreme example is that my only job is to provide training for my future robot overlords. Do I believe that's true? Do I believe there's no rule for humanity? Hopefully not. But if I use that as an operating assumption, then there is no part of there's no time that I interact with an agent that I'm not doing it in a thoughtful way to improve their ability to give me a better outcome next time. So the first thing is you just rinse and repeat. You, like, you wanna build an entire software factory, right, where it goes from you just say the vision and it's in production with no human no required human gates between the two. All you do is you tell an agent to write some code, you tell it what was wrong, you make sure that it captures that context, and you tell it to go again. And you keep repeating that loop time and time and time again until you build more and more validators, adversarial reviews, and other elements that will reduce the likelihood. And you know what? I've taken that in production database in my life. Like, humans get it wrong too, but it's much less likely the more revs you do around that cycle to figure out what does good mean, what does bad mean, and how do we put, both tooling and and also just prompts. You've got the prompts, so you've got the tooling, you've got the context, but also you do need, I think, permissioning systems. You need to think about the blast radius. And then all the good stuff we've always been doing, observability, monitoring, alerting, I think there's great opportunities now for or self healing software is a fancy word for incident response where the LLM knows enough about the system to propose a a move forward. I think that's going to become mainstream. And, also, just stuff that we're doing fifteen years ago, game daying, resilience engineering. How can you make it so if your cart goes down, at least you can still see the products? If your payment provider goes down, it can still save your cart and send you an email once it goes up. Assume that stuff's going to break more and create more resilient systems with less moving pieces so that even when things go wrong, the blast radius is constrained. Yeah. One final thing I'd point is and the only other part is you can also we're not gonna review every line of code at some point that's going away. However, you can take a statistical process control approach to that, which is it's like ball bearings. You don't check everyone, but but if one of them is out of tolerance, you start checking them more frequently. And you do it risk adjusted, you're probably gonna test a smaller percentage of the PRs for your admin dashboard than you are for your main financial production flow. Yeah. That's that's amazing. I I love that. It's interesting. You kinda bring up sort of like, oh, when this stops working, you know, you have this, like, graceful degradation, process. I my first thought is, you know, you don't know what you don't know. And you have these sort of non engineers building out systems, and the system might work. But if they don't know to prompt in, you know, hey, let's let's make sure that we have this, you know, progressive enhancement, graceful degradation, then that just won't be included there, which is which is interesting to think of. Like, the creator of, you know, the software still needs to be architecturally aware of, you know, software engine, design and architecture. Two possible wonders I've seen for that. One model which I think is is good for now, which is I think we're going to see, you know, the two pizza team is not always going to be six to eight people now. I think you're gonna find triads in groups of four being able to get a lot of move a very long way and very fast when you have teams of twenty, fifty, 60 agents working with them. But there's going to be a combination of product and engineering skills. Somebody needs to know what we're building and why. Somebody needs to be thoughtful about, wait a second, latency, queuing, retime, split brain problems, all this stuff we deal with with large distributed systems. Those awarenesses need to exist. And then you might have some design skills or some data skills depending upon where the group is in your org and what it's trying to do. So I think in the early days, you make sure there's an engineer in the room before something you care about goes to production. I think, eventually, if you build a sufficiently rich software factory, you build those architectural concepts into the review process. Whether we get there and how close we get there, I don't know. So for now, I'm gonna keep a staff plus engineer having to look over the code just to be sure. Yeah. How and how quickly we get there. Exactly. Holger, I'm I'm really curious to hear your sort of, like, take on this. You know? Obviously, Peter mentioned a lot of things that do get sort of these pilots up into production. But, you know, our platform direct us, for instance, you know, we power, you know, mission critical software where we're pumping out, you know, bandwidth at, like, nine nine thousand pizzas per second, you know, for for certain customers or whatever applications they might be, building. And it's funny to me thinking of, like, that that is, like, production grade, like, at scale. You can't just one shot, you know, or Vibe code your way to that. Like, there's there's a process. So I'm curious, you know, when you take sort of that equivalent at, you know, HelloFresh, like, what are you what are you doing? What do you layer on top to make sure, that your applications, you know, if you're if you're building in that way, you know, through AI, that it can handle that. Yeah. That's a fantastic question. I think we, so so some some of the stuff, you mentioned, we do have to be one of the things, like, with the higher throughput, right, of of agentic AI or just, you know, teams speeding up. We see more, pull request reviews coming up with a new problem. Right? So so we need to find more, ways of, I think it's a really good idea, right, to kind of, find another abstraction level, basically, to look at, you know, risk management and and reviews and and and in a sense. Right? So that's quite interesting. So I think there we need to invest more. Right? This is something we need to do. What we have been doing is we have brought a few things to production, like AI products to production, which is kind of interesting. So we've been, maybe the most obvious case. We did go first as well. This is like customer care and having chatbots right there and experimenting there, a lot. And there, it's it's really it was we were kind of mid mid class last year or so. Right? So we were learning how to do this as we go. Right? And then managing, basically, the technology was not so much of the problem. It's more like managing the, really the the quality of it. Right? So all the different ways of of of how how a conversation with a human, the one deterministic, entity, right, can go, is, is, you know, not to underestimate. Right? And there's a lot of things that can go go wrong and how to how to design communication, the getting the intent right, and and keeping it safe, right, and and correct, is an is an ongoing task. And we, as we invested into, you know, the teams, that build up this messaging and then as well, the the automatic test. We experiment with AI as judged. Right? So, basically, having LLMs checking those those those flows as well. I think this this is very, very promising. Right? So this is one pattern that we see more is that you'll introduce, AI, maybe even different models, either for pull request reviews, having iterations on them, basically watching the manual testing themselves. Right? Like, having an eye look at, at how how it looks like and then doing pull requests by themselves. This actually works pretty, pretty well and has a very has a very high showed very high impact. So these are a few examples of where I can see. Right? So I think my my main concern right now is really getting the the the review part out of the way and and and on the other side, getting more, I don't know, engaging more UX and so on to create more experiments. So that's the other thing. I would like to have more, more variants basically tested earlier with the idea of having just already a more winning variant, basically, created into production, have higher higher signal, higher, success with more expensive AB testing in production. But why not having, like, 10 we we see this already, but I think we can do more 10 variations, test the prototypes, high high fidelity, bring it into the building, having tested before or with synthetic audiences. Right? Checking out quantitative analysis even on on scale if a business model works with a new hypothesis. So these are the things we're currently on and want to invest more into. So to get it on the front. Right? So getting better better signals earlier in the process, having the winning variance going to production faster and solving the PR revenue bottleneck problem. There are a few topics that we're currently working on. And then a few other things about personalization, improving the models, and as well creation from automatic creation of videos and and images, and we're creating a lot of, like, meals. And we have some AI support tooling that helps menu chefs, right, the creators of recipes to improve the process of getting this from idea to production, significantly faster than before. Yeah. Yeah. Well, it it that's I I love that. How how big is your org, Holger? So my org is about 300, people right now. The overall HelloTech org. So how this is how we call this either, combination of all, like, tech technical staff, if you want, is, is about 1,000 people. Yeah. And I I it's interesting. Like, all this only happens if we actually you know, the governance, you know, getting to production, you still have to have the people to make this happen. You know, going back to early, like, we can't just make the agents have the other agents work. Like, there's still human in a loop as Peter keeps putting it. It. You know, we think of, you know, Jack Dorsey and and kinda what happened over at Square. Like, switching to, like, the people side of this, you know, with an org of that size, like, how do you actually get the workforce to adopt, to use, to, you know, to actually have this happen in the first place. You know? Is I'm I'm assuming there's a fear piece. You know, we've talked a little bit about, I think, Peter, you had mentioned sort of building your for the future AI overlords. You know? There's a perception, I think, across some people that, like, oh, am I just digging my own grave and sort of, like, going out and and sort of building these these systems? But, you know, we have to lean in. You know, we can't just ignore it. So, you know, Holger, within your org, like, how does that work? Like, how do you make that happen? Yeah. So I like to to frame it in a way that, I think HelloFresh is a really great way to experiment with this, learn this, apply it on scale, and just make yourself, stay relevant if you want. Right? And and be on on on top of the game there. Right? So we we really try we we try to make this work. We support people. We support the teams. And that's kind of the positive way of framing it on if you wanna play it a bit differently. If you don't do this, right, the gap between, what the what the what the industry demands or what where the where the top level is and and everybody else, it's getting bigger very fast. Right? So you you you risk of falling behind in a sense, and that's not limited to engineering at all at all. Right? It's like every function and software engineering in and around it at least is, has the same kind of, opportunities and and challenges in a sense. But I don't, like, underestimate the, those factors. They are they are real. Right? So and I think there's a system of incentivizing, helping people, making it, you know, as we talked about it, easy to figure to try out things. Right? But then as well, set up clear expectations, right, that, that we expect every team to onboard into the new process and to to adopt to this agentic AI, for instance, use cases or to to have more variance in production now that we can do it, right, fairly easily. So there will be, the expectations grow steadily as well, to keep up with this. So that's the both of the side. Right? Enabling people, but as well-being very clear about expectations. Yeah. Oh, 100%. And it did like, Deloitte's latest report, I think, said something like only 20% of companies of orgs, actually, their their teams are actually ready, and prepared for AI. So, you know, getting that number up, hopefully, HelloFresh, you know, is in that 20%. But, but we're come we're coming up on time here. So I wanna kind of at least be a little bit more forward looking, with with sort of wrapping up here in terms of what's next. Peter, maybe I'll send it over to you. If you're looking out the next, like, say, one, one and a half years, like, where do you think the most change is gonna happen within AI? With an enterprise AI? I I think you go anything beyond that. I mean, even if you go out six months, the question mark start piling up so high. But if you were just kinda, you know, behind the sky, where do you think things are are heading? Within the engineering org, I think it's pretty straight this will be I'm not gonna put a timeline on it. I've got a very good friend, senior CTO. He's like, you can make whatever prognostications prognostications, but don't put a timeline on it. You you'll thank me later. And I I think he's right. I have no idea. It will be much faster at smaller companies in greenfield environments, in places where you happen to have, an engineer that's just super passionate about this and pulls the team forward. It should happen more quickly in SaaS companies and in companies that are potentially have an existential threat from AI. As I said, you're on a chain of gyms, maybe it doesn't matter. So there's going to be uneven, speed of change, but I think, eventually, we're going to see a flattening. What the year of efficiency continues to resonate many years after. You could imagine a world where you have triads or maybe teams of three to four doing something similar to what a two pizza team does now so you can split them out and have more of them capturing more initiatives. I I think it's possible you might have five to 15 of those reporting straight up to a senior director or a VP where you have a team lead in each. I think we're gonna continue to see the collapsing of that, and I think the VP is gonna be spending a lot of time on the keyboard reviewing detailed specifications, verifications, harness improvements, as well as research and product elements. So I think we're going to see leaner engineering orgs. The good news is we need a 100 x the software, and, I think there are lots of opportunities for anybody who is passionate about either. And I think we need to be happy with this. Right now, we're all excited about AI engineers. And for me, that's basically harness engineers, people who are thinking about prompts and verifications and steps and gates. We need 8% of those. That's our platform or DevX team. Everyone else is going to become a product engineer who gets better at proposing experiments, writing specifications, and creating rich verifications while also being thoughtful architecturally. And I think those are the two roles that are gonna continue to grow and be incredibly impactful in the future. Yeah. Oh, I love that. Slightly different lens, hold on. I wanna hear sort of your your sort of take on where things are heading. But, on the enterprise, something that's very sort of near and dear to my, my heart, my job is sort of the commodification of the front end. Like, we're we're seeing that happen actively with the AI app builders. The value of software that's being built, where where does that value shift? As as the front end, as the facade becomes commoditized and and everybody can build that, where does it shift? I mean, down the stack to, you know, all the way back to the database or somewhere in between. Where do you feel, that's that's heading over the next year or so? Yeah. It's a it's a it's a very good question. So basically, it's, I I still think, a a creativity and and being able to create a mode, the connection to your customers, building up this relationship is probably more important than ever. So you can be if you double speed on the wrong side of the highway, right, it it will not help you necessarily. So, so and I think that the the productivity gains that we have, that we're seeing now and that are super exciting, they will become a commodity at some point. I'm not putting a timeline on it either. But, you know, you you will have your great engineering teams, and product engineering teams. They will they will work on a completely different level. You still need to figure out what really works for a customer. And now the levels the field is, we're closer. Right? So I cannot rely on, say, a company like HelloFresh being faster, to production with some features than a small start up. So I need to be really, really smart about this. Right? I need to be have the most creative people writing the best specs. Right? So and, we have a lot of customers. We have a lot of customer relationship that is that is super valuable. That is not easy to copy, right, even with AI. So building on top of this and making sure that it stays this way and we stay on top of the innovation, that is what, I think, what still matters the most. Yeah. I mean, bring it back to the people, you know, the as a designer, you know, focusing on the creative. You know, you can always it's really exciting right now because AI is able to take, you know, this massive context and, like, pipe it in and make everybody creative, make everyone a developer. But at the end of the day, you know, you have to be able to break outside of, like, what's already there and sort of just rearranging those things. So I love hearing the answer kind of always coming back to the human, to the core of, you know, creativity and, you know, even going back to earlier. You know, you can throw the the kill switch in there. You can do all the right things and still, AI will kinda find its way to work around it. So, it's really important to have people that understand the the proper architecture of of what we're building and and why those things are important. So, hopefully, we don't stray too far from from those first principles. Holger, Peter, it was really, really exciting, getting to just chat through, you know, and the enterprise with both of you. You know, hopefully, we'll get to chat again soon, maybe again on a bridging bites episode. But for now, anything that you guys wanted to sign off with before we wrap this episode up? No. Just thanks a lot for for for having me. And, yeah. It was a pleasure talking to you. And, it's just exciting times, I must say. Likewise, Ben. Thank you so much for the invite, Holger. It was wonderful to hear your insights. So much fun and great conversations. Lots for all of us to learn. Awesome. Yeah. Well, absolutely. Thank you both for being here, and we'll see everybody, soon on the next episode. Alright. Thanks, guys.","published",[18,29,36],{"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",{"people_id":30},{"id":31,"first_name":32,"last_name":33,"avatar":34,"bio":35,"links":10},"ac8e8e7c-390a-4589-ac64-bceeca8d2a32","Peter","Bell","d29dd6a7-efcb-4ec2-998f-311a70552287","Founder, CTO, Head of AI at Gather.dev",{"people_id":37},{"id":38,"first_name":39,"last_name":40,"avatar":41,"bio":42,"links":10},"654f5c29-b0fe-494c-969a-4a9e433fdad5","Holger","Hammel","0a59ff9e-160f-4862-a8f5-f420249ba565","VP Engineering at HelloFresh",[],{"id":45,"number":46,"year":47,"episodes":48,"show":56},"289f6534-7fdd-46df-8c00-89a75469fe41",4,"2026",[49,50,4,51,52,53,54,55],"bbaa3063-6fbe-4d96-bbc7-e50672f9a308","f885409e-0ace-41e5-aca3-faf4dcd7659b","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":57,"tile":58},"Leap Week","62816023-fa7e-4a76-b9a1-2733ee2093a6",{"id":51,"slug":60,"season":45,"vimeo_id":61,"description":62,"tile":63,"length":64,"resources":10,"people":10,"episode_number":46,"published":12,"title":65,"video_transcript_html":66,"video_transcript_text":67,"content":10,"seo":68,"status":16,"episode_people":69,"recommendations":72},"from-zero-to-production","1176293308","Spin up a backend, connect a frontend, and deploy to production - all in one session. Featuring Railway.\n\n","14919273-9a84-4b2a-a651-20a11144d00d",38,"From Zero to Production","\u003Cp>Speaker 0: Hi, everyone, and welcome to Leap Week. My name is Lindsay, and I'm an engineer on the direct us team where I mostly work on integrations and developer experience. For day two of Leap Week, the theme is build it, run it, sell it. The idea is to show how quickly you can go from an idea to a real production application using modern tools. We'll be focusing on the first part of that story, build it.\u003C\u002Fp>\u003Cp>For this session, we're going to build a fictional AI automation agency. The idea is that this agency helps companies implement AI workflows for things like marketing, automation, research, sales outreach, and internal knowledge assistance. So by the end of the session, we'll have a working back end for this agency, generate a front end website for it, and deploy the whole thing live. But before we can build anything, we need somewhere to run our back end and infrastructure. I'm joined today by someone from the Railway team.\u003C\u002Fp>\u003Cp>I'll let them introduce themselves and tell us a little bit about Railway before we start building.\u003C\u002Fp>\u003Cp>Speaker 1: Awesome. So, thank you so much for having me. My name is Mohammed. I'm a general engineer at Railway. And at Railway is I would say in a nutshell, it's a all in one intelligent cloud provider.\u003C\u002Fp>\u003Cp>You can deploy pretty much anything, databases, back ends, front ends, queues, whatever. Like, you can just deploy it. You can just run it. And, yeah, essentially, we allow you to deploy anything, and we really focus on making it easy for you to go from zero to deploy it as easy as possible. Yeah.\u003C\u002Fp>\u003Cp>So excited about what we'll build today today.\u003C\u002Fp>\u003Cp>Speaker 0: Awesome. So the first step in getting any project off the ground is gonna be your infrastructure. And instead of spending hours setting up servers and databases, Railways gonna let us start from a working template. For this demo, we have a template we've already created. It's a pretty simple one.\u003C\u002Fp>\u003Cp>It's gonna have everything you need to get started. We do actually have another version as well that launches with CMS collections already started if you want a starting point. But for this, we are going to use our blank one. So I'll go ahead and launch that template here.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. So a little bit of context. Templates essentially encapsulate a set of infrastructure components, all the configuration. And here, we essentially have, like, a database set up, post press red as a bucket for object storage, as well as directives, which runs our back end. So, like, directives here would be, a server that we actually use.\u003C\u002Fp>\u003Cp>And now, like, everything here is neatly on the railway canvas. You can see everything grouped under, like, the directors group. And, essentially, like, here, you have a project, and a project is where you can have essentially everything deployed. So, like, in this example here, you have practice running. Maybe you will have, like, the front end, so that will be a separate service.\u003C\u002Fp>\u003Cp>And that's pretty much all we need to get started. So, like, all you need to do is you can sign up for free, and then you can go to the link to deploy the template. You click deploy now. You can either deploy to a new project or an existing project. If you choose, like, the path we're going through right now is deploying to a new project, this is the experience you'll get.\u003C\u002Fp>\u003Cp>And now you can see, actually, it's like we have the difference for versus, they're being deployed. So, like, Post JS, that's the database. And then we have Redis and as well, we have, DirectThis, which is also just cute. So, like, one thing after another will be deployed. So, like, in a, hopefully, a minute or two, everything should be ready, configured.\u003C\u002Fp>\u003Cp>You don't even need to, like, touch an editor, and it just works. So that's pretty much yeah. That's pretty much the overview for templates.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. So it looks like\u003C\u002Fp>\u003Cp>Speaker 1: Like, this is also deploying I see it's so this one is actually deploying from a Docker image, and this is the official one, for directors. I assume this is the latest version. Right?\u003C\u002Fp>\u003Cp>Speaker 0: Yep. This is our latest version.\u003C\u002Fp>\u003Cp>Speaker 1: Yep. We make sure\u003C\u002Fp>\u003Cp>Speaker 0: to keep this updated.\u003C\u002Fp>\u003Cp>Speaker 1: That's awesome. But, yeah, if, let's say, for example, you deploy and you want to get the latest version, of, let's say, Directus. If you actually go to settings, you'll be able to see the source image. There's like automatic updates. You'll actually be able to see if there's an automatic update.\u003C\u002Fp>\u003Cp>You just click apply. You deploy it. It just works if you want like the latest version. So it should be very easy to set up, but so far we can now we see, we have the database ready. Red is ready.\u003C\u002Fp>\u003Cp>The bucket is created, and now the last component direct list is being deployed, which hopefully should be ready Yep. Very soon. And we're live.\u003C\u002Fp>\u003Cp>Speaker 0: We'll click that. So in just a couple minutes, we've gone from nothing to a running back end with a database, API, and admin interface. So as part of our setup process, we'll create our first admin user here. Give it a password.\u003C\u002Fp>\u003Cp>Speaker 1: So I assume this happens as the first initial step. Is this correct?\u003C\u002Fp>\u003Cp>Speaker 0: Yes. Yep. Okay. When you first set up, it'll ask you to create your admin user. And here we go.\u003C\u002Fp>\u003Cp>We're in our, UI here. And to get started, we are gonna do a little bit of, setup for our AI assistant because we are going to be using AI to create most of this. So Directus has a built in AI assistant that can help you. It can generate schema content, and it has knowledge of what's going on in your instance. So if you go into settings here, we'll give it a key.\u003C\u002Fp>\u003Cp>Oops. Let me grab my key.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. I think you just said\u003C\u002Fp>\u003Cp>Speaker 0: One password. Gotta have that security.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah.\u003C\u002Fp>\u003Cp>Speaker 0: Alright. And then we are also going to set up our MCP for later because we'll be using that when we do our front end.\u003C\u002Fp>\u003Cp>Speaker 1: So here, this is, like, the Directus MCP or, like, this is an MCP server for our deployed direct the Directus instance?\u003C\u002Fp>\u003Cp>Speaker 0: Yes. It is. Yep. So our cursor, which we'll use to do the front end, will be able to access all of our data here. And it'll depend on the permissions we give it, what it will be able to do.\u003C\u002Fp>\u003Cp>But we also have this in app AI assistant here. So we're gonna use that to generate our schema. So I already have some\u003C\u002Fp>\u003Cp>Speaker 1: So quick question. For someone who's not a 100 to 100% familiar, let's say, with all of the features, essentially, what we're doing is now we have a like, our complete back end deployed running. And part of this back end, you can actually expose an MTP server, and then I can connect to this MTP server via call code, code x cursor, whatever coding agent of my choice, and then just tell it to do stuff for me. That'll be reflected in my live deployed instance, and now it's just that's it.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Yeah. We'll see that a little bit later, but we're gonna do something similar here. We also have our AI assistant that's in the UI here, and we are gonna use that to build our back end, our collections, our schema. So right now, we have no collections created.\u003C\u002Fp>\u003Cp>We don't have any data tables, basically. So I'm gonna go ahead and prompt it with a prompt that I created earlier, and it's going to create the structure of our back end.\u003C\u002Fp>\u003Cp>Speaker 1: And the idea is, like, a collection would be, like, an object, right, for something.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. So it's gonna be, like, your data tables. So it'll have, like, an overlying day data table where we'll have one called site settings, and that's gonna be the settings for the website. We'll have one for case studies.\u003C\u002Fp>\u003Cp>Speaker 1: Mhmm.\u003C\u002Fp>\u003Cp>Speaker 0: That's gonna have case studies for our AI agency. Actually, I can show you the prompt here a little closer. So we've we're creating our collections, and we're giving it different fields. So we're giving it, like, a tagline, a headline. We're gonna give it different services that this AI agency does, give it some fields there, and we're gonna create playbooks and case studies out of those, so things you might need in an AI agency marketing website.\u003C\u002Fp>\u003Cp>And the AI on our end is going to do that work for us. So we've told it in our prompt what we want. We didn't really tell it what type of fields they need to be, if they're strings or numbs or any of that. It'll infer that if we want. And, actually, one nice thing is if you don't know what schema you want, you can talk it through with the AI assistant.\u003C\u002Fp>\u003Cp>It can give you suggestions of things that you might wanna add for a data table. So it's gonna ask questions too back to the user. So it's saying it's ready to create these, and it's saying, do we wanna proceed with its defaults? Do we wanna review each one and decide? Or do we wanna have, like, it be super minimal?\u003C\u002Fp>\u003Cp>I'm gonna say just create it. Let's see what it comes up with. It's asking how we should store the icon. So let's just say a single file upload. How do we wanna store the pricing start?\u003C\u002Fp>\u003Cp>We don't really care about that. We'll just pick something.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Yeah.\u003C\u002Fp>\u003Cp>Speaker 0: There's a\u003C\u002Fp>\u003Cp>Speaker 1: there's a recommendation. So I think we can\u003C\u002Fp>\u003Cp>Speaker 0: go with it. Yeah. It's collaborative. It's gonna ask you how you wanna do things. It's not gonna just bowl over your user and, like, do everything for you.\u003C\u002Fp>\u003Cp>It's it's not just an AI tool that's gonna do it for you. It is a human in the loop type of experience where you can make sure it's doing something you actually want it to do. And it's gonna tell you everything it's doing while it's doing it. So you can see here. It's planning it out.\u003C\u002Fp>\u003Cp>You can see the thinking if you wanna know, like, what it's thinking while it does it. And you can do the different models if you're you have a different one. We're currently just doing, GPT five. In those settings we set up earlier, you can change that out. Like, if you want something different going on, you can say which ones are allowed, because this UI is actually not just for your admin.\u003C\u002Fp>\u003Cp>You can allow other people here. So you can have other people with different access levels able to come in and modify your content once it gets created or your data tables. So you can let them use the AI assistant, and you can set what permission levels they have. So it's gonna ask for different tools like creating the collections if it can do it before it does it. You'll see here's our data models now.\u003C\u002Fp>\u003Cp>It's created a couple of them. We've got our case studies one. It's starting to add the fields into these. In the past, you would have to do this all manually and create collections. Now you can just use the AI to do it without having to worry too much.\u003C\u002Fp>\u003Cp>So it's a lot less manual work.\u003C\u002Fp>\u003Cp>Speaker 1: But we could have done the same thing by using the MCP thing. Right?\u003C\u002Fp>\u003Cp>Speaker 0: Yep. You can. Yeah. We just wanted to give people an in app way to do it a little faster and simpler for people who maybe aren't don't have a data model or AI model they like to use outside of here.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Definitely makes sense. It's just like in my mind, I really like the flow of using some like, I use Claude a lot. So when I use Claude, I'm like, okay. I want a complete end to end flow.\u003C\u002Fp>\u003Cp>Like, what what would it look like if I don't leave cloth? Can I actually do the stuff I wanna do? And it seems like the answer is yeah. But there's also And\u003C\u002Fp>\u003Cp>Speaker 0: that's actually that's what I like to do as well as I'll go in cursor, and you could start the whole front and back end directly in cursor using the direct SMCP, and it can create it in here for you. We're gonna do the front end in Cursor just so we're seeing that too a little bit, but I wanted to use a little of the AI assistant as well.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Yeah. I'm actually in the process of making sure that our experience of getting started to be super seamless. So we already have, like, a regular CLI. We have, like, Asian skills, and, we're going to essentially keep shipping more, features through the CLI.\u003C\u002Fp>\u003Cp>So, like, it should be able to you say, like, oh, I wanna deploy direct us. It should be able to go through the templates marketplace, find the most up to date template deployed for you, and then it should be able to also figure out, like, the deployment is finished. Let me continue. And then you just have, like, a complete workflow where you don't even have to go to a templates page. So that's kind of, like, what's\u003C\u002Fp>\u003Cp>Speaker 0: That's awesome.\u003C\u002Fp>\u003Cp>Speaker 1: In the works. But yeah.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. I'm excited for that. That'll be good.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. That's I think it asked a question at the end.\u003C\u002Fp>\u003Cp>Speaker 0: Yep. So now it is done. It's gonna, like if you've used this model, it constantly asks you other things it wants to do, but it is done. It's created the fields. So we've got that part done.\u003C\u002Fp>\u003Cp>Now we're gonna give it some fake content in these fields. So that's the database table. And now we're gonna fill these tables with actual, like, fake data. So I'm gonna paste that in. So when we go into here, you'll see no content yet, and it's gonna start filling those in.\u003C\u002Fp>\u003Cp>But you can see, like, these are the fields. So if a nontechnical user came in and wanted to give it a new name or a new tagline for their website, they can do it in here without having to touch the code at all. So it's a nice way to work with both your front end team and your back end team and your content team and not have, people who are super technical not able to get things done. While this is doing it, we are actually gonna get finished setting up for our MCP, actually. So we're gonna go here in our docs.\u003C\u002Fp>\u003Cp>We have a little section for setting up the MCP in different tools. We're gonna use cursor. So I'm gonna click add to cursor here. Nice easy button, and it will set us up into our settings. And we are going to go grab our URL from here.\u003C\u002Fp>\u003Cp>So this is our directest instance URL, and we'll pop that in here for our MCP setup. So it's just your instance URL slash MCP, and then you're gonna wanna set up a token. So let's make sure it has the permissions it needs. When you're connecting to external tools like cursor, it's always a good idea to scope your permissions so that it's not able to completely nuke your whole setup. You know, AI likes to be overeager.\u003C\u002Fp>\u003Cp>Speaker 1: So Yeah. Yeah. It's like, let's start all over.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. We need a fresh start. We we recommend setting up a token for it specifically and not using your admin token. So you wanna go to your user roles here. You'll create a new role, name it whatever you want, and you'll add a policy.\u003C\u002Fp>\u003Cp>We're gonna give it view only policy. So you'll add your collections, give it read only access. That way it can't, like, update or delete or anything when it's in the MCP. If you want the MCP to do that, you can totally give it permissions to create, read, and update, and do whatever you want with it. But I recommend giving it, like, the lowest level of access at the start so that, you know, it's not gonna go in and break your whole setup here.\u003C\u002Fp>\u003Cp>Mhmm. So we've got our policy. I'm gonna create a new user too to grab a token from. So if we scroll down to our tokens, we're gonna generate it and save. And then we'll go back in here, and that's where you put in this bear here, authorization code.\u003C\u002Fp>\u003Cp>You put that token, click install, and that will install our direct as MCP encursors. So now we're ready to go for that when this is finished. So let's go check on it.\u003C\u002Fp>\u003Cp>Speaker 1: So what we've done is create a role, and then the role has the token?\u003C\u002Fp>\u003Cp>Speaker 0: Yes. So the the user is what who has the token. So this MCP user has the role that we created. I created it all within the same page, but this person has the role, and then they have the token. Mhmm.\u003C\u002Fp>\u003Cp>So if you go into our settings here and look at the MCP role, we've got our view only policy that we set up Mhmm. For our different collections. And the users in the role is that MCP guy. So now it looks like it did finish the content. If we go back into our content page, now you can see it's filled it in with our sample data.\u003C\u002Fp>\u003Cp>I gave it a few guidelines of the type of things I wanted it to have, but, otherwise, it just kinda came up with this on its own, sample content. So it's given us a starting point. So now that we've got some sample content, we are going to go back in to cursor. I'm gonna grab this really long prompt I've created telling it basically we are an AI agency. We're creating our website.\u003C\u002Fp>\u003Cp>I gave it I told it to use next. I also made sure to tell it to use the Directus MCP server. This might take a minute, so I'm gonna go ahead and get it started. We are in a completely empty folder here, so I'm gonna have it do all of the coding for me. Let it create what we need here.\u003C\u002Fp>\u003Cp>It's gonna make sure to look at our schema via the MCP. So we want it to not guess on what structure we have of our data. We want it to actually look at it. Like we said, we could create the data directly from Cursor via the MCP. You would have to make sure you give it, the right permissions.\u003C\u002Fp>\u003Cp>So that that might have to be an admin token in order to generate new schema with your MCP. So if you're comfortable doing that, you totally can.\u003C\u002Fp>\u003Cp>Speaker 1: Interesting. So once that's done, we'll have something that's working like a front end running locally that's fully functional, pulling data from our deployed direct us instance. Correct?\u003C\u002Fp>\u003Cp>Speaker 0: Yep.\u003C\u002Fp>\u003Cp>Speaker 1: Okay.\u003C\u002Fp>\u003Cp>Speaker 0: Yep. And any changes you make in direct us would be reflected in your local front end that you've got here. And then as the final step, we are going to do to deploy this front end to railway so that you have a full production application front end and back end.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. I have cursor set up, so I could just automatically runs everything. I don't, like, manually approve stuff. I just\u003C\u002Fp>\u003Cp>Speaker 0: I don't trust it. Yeah. I gotta make sure I know what it's doing. I've been burned before.\u003C\u002Fp>\u003Cp>Speaker 1: I don't know. The the models are they keep getting better and it's like, I'm like, I would've done the same thing. So I don't know. I feel like it's just faster and you can just like, you know, sit back, relax, grab your favorite drink while it gets to work. You come back to it and you're like, woah, this thing actually works.\u003C\u002Fp>\u003Cp>So Yeah.\u003C\u002Fp>\u003Cp>Speaker 0: I think I'm a bit of a control freak. I can't let it just just\u003C\u002Fp>\u003Cp>Speaker 1: go. I'm telling you, like, once you let go and you see, like, it actually works and it's like, I think, I think you'll be plenty surprised. Then you're like, wow, that's actually a really fun. Cause like now, you know, if let's say for whatever reason, someone brings you on Slack, you can play for guys. Like, you come back to us, like, oh, I forgot to approve.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. That it's done.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Yeah. I I I think, it's a much more fun experience when you just let it kinda do its own thing without you steering it. But I do understand the, appeal of, you know, wanting to see what it does. But also, like, we're starting from scratch, so there's little, like, room for error, I would say.\u003C\u002Fp>\u003Cp>But yeah.\u003C\u002Fp>\u003Cp>Speaker 0: There's nothing for it to break. It's creating it from scratch. So\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Yeah.\u003C\u002Fp>\u003Cp>Speaker 0: We're pretty comfortable letting it just do what it wants to do here. So it looks like it actually has gotten the types if you look. So it pulled this from our MCP. It checked in, Directus, and it got our back end and what types we have there. So that's good.\u003C\u002Fp>\u003Cp>Looks like it's I did tell it also to create a little connection a net little connector to direct us here, a fetcher.\u003C\u002Fp>\u003Cp>Speaker 1: So it\u003C\u002Fp>\u003Cp>Speaker 0: looks like it's done that, getting different gets to get the data.\u003C\u002Fp>\u003Cp>Speaker 1: And it will be able to pull in our deployed, like, like, URL and actually set it as an environment variable?\u003C\u002Fp>\u003Cp>Speaker 0: Yes. It will. So I it might make me make the e n v, but I've got it. It's supposed to use the direct URL here, and it's supposed to use the token that we created. So, we'll see.\u003C\u002Fp>\u003Cp>I've done this a couple times just to see how it would do, and it it doesn't always know to grab the URL. If it's really smart, it can grab it from the cursor settings, and it can get the token and the URL, but it doesn't always manage that. You'll notice I'm in auto mode, so it's a different model each time, so it might not be a very smart model. But we at the end, we'll give it the proper e n v variables, and it'll be able to connect to our deployed instance.\u003C\u002Fp>\u003Cp>Speaker 1: Excellent. Yeah. Honestly, again, I I like my to choose my own models, You should use, like, a, you know, g p three five point four or, like, an Opus 4.6 and because I don't know. I don't like that, I would say I wouldn't trust the auto. Like, which one would affect?\u003C\u002Fp>\u003Cp>Like, I wanna be I want a consistent experience. So then if, like, anything starts feeling weird, then I can blame the model. I was like, oh, you know, like, Opus is acting really dumb lately, that sort of thing. So\u003C\u002Fp>\u003Cp>Speaker 0: It it does just feel like from one day to the next, though, even if you have a model that you know is good, it it doesn't always behave exactly the same.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Sometimes it feels like, you know, you're you're just rolling the dice. Sometimes it's like, woah. This is it. Like, this is the next big thing.\u003C\u002Fp>\u003Cp>I was like cause like, sometimes I, I, I asked it a question and I wanted it to check against like, you know, our internal, GitHub repo. And it's like, Hey, do we support this thing? It was like, yes, you do. And I told her like, oh, show me the code. It's like, after checking, apparently, this is not supported.\u003C\u002Fp>\u003Cp>I was like, come. Yeah. And this is like the smart and I told it, like, oh, you know, UltraThink, you know, work really hard, but it's like, I guess sometimes it's just, you know, not that type of day. But I think here it's like it's it's it's working well, based on what I'm saying.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. It looks like it's getting something.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah.\u003C\u002Fp>\u003Cp>Speaker 0: It's messing around with the configs now.\u003C\u002Fp>\u003Cp>Speaker 1: Interesting. What is what's it doing to the Next. Js config if we open it up? Just out of curiosity. Technically, you know, we're we're not supposed to look at the code, you know, and, because, you know, I was just shipping it, but I'm also curious.\u003C\u002Fp>\u003Cp>Okay. It's setting, like, the assets URL, I assume, for, like, files and stuff. That's pretty awesome.\u003C\u002Fp>\u003Cp>Speaker 0: Yep. And while it does that, I'm actually gonna grab my last prompt, which is a prompt to have it, push to GitHub so that we can get it ready for deploying the front end. I'm gonna grab that. Go back in here. I think it's almost done.\u003C\u002Fp>\u003Cp>It feels like it's it's doing the read me. So\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Probably at the end. But yeah. I would say, there is actually a like, you you can definitely, you know, start with this workflow. It's totally fine.\u003C\u002Fp>\u003Cp>But if you have, like, the AOS CLI setup, it could just deploy it to your existing project. And this way you don't even have to worry about git being installed, get, having a GitHub repo or anything. But when you have it with git, well, the benefit is, anytime you make, changes through the front end, it will just automatically deploy. So it's like it's it's all trade offs based on, you know, how, structured and organized you wanna be. But yeah.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. That makes sense. Alright. We did not create our E and V, so we're gonna create that real quick. And\u003C\u002Fp>\u003Cp>Speaker 1: I would say that's the responsible thing for it to do.\u003C\u002Fp>\u003Cp>Speaker 0: Yes. I don't didn't really like that it created its own ENV and, like, grabbed my things from the MCP. That was a little, uncomfortable. Yeah.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. It then gets better this way.\u003C\u002Fp>\u003Cp>Speaker 0: Yes. Look at our URL. Alright. And then we're gonna run it locally just to see it working.\u003C\u002Fp>\u003Cp>Speaker 1: What does the static token do?\u003C\u002Fp>\u003Cp>Speaker 0: So that's the one we created with for the MCP. We're using it here too. Okay. So that is the permissions that we allow in Directus. So if you give it, like, no read permission, it won't be able to see the content.\u003C\u002Fp>\u003Cp>If you give it access to read, like what we did, we gave it access to read all of these collections, then it can read the data. So that's where, like, you probably want a separate one from your MCP. We're just in the interest of time having it Yeah. One that is view all for all of our collections because we don't need to make changes with the MCP. But you probably would have permission to update and delete with your MCP as well.\u003C\u002Fp>\u003Cp>So you probably want wanted a separate one for your front end. So you'd probably wanna create a user role for your front end that gives it different permissions. Otherwise, you can have it set up to use your public permissions, which controls what the API data is, available without authenticating. But it's safer if you do give it a token so that your front end is only able to grab what you want, not, like, everything. So let's see if it worked.\u003C\u002Fp>\u003Cp>Speaker 1: Moment of truth.\u003C\u002Fp>\u003Cp>Speaker 0: Okay. There it is. Looks like some images are broken, but we've got a full website here. You can navigate around. It's got some prompts.\u003C\u002Fp>\u003Cp>It's got a button to get in touch.\u003C\u002Fp>\u003Cp>Speaker 1: We've got a\u003C\u002Fp>\u003Cp>Speaker 0: full website here. Obviously, you'd wanna iterate on it, fix the images, and, do a little honestly, it looks great. It doesn't need too much design or anything even. So that's a good starting point. We are going to go ahead and say good enough.\u003C\u002Fp>\u003Cp>The images can get fixed later. We are going to deploy this. And, of course, I copied other things, so I need to go back and get my prompt. So like you said, there's other ways to do this, but we are gonna have it pushed to a GitHub repo. We're pretending we have other people who wanna manage this code, and we want it in some place that we would have access to it as a team.\u003C\u002Fp>\u003Cp>So we've have a GitHub repo we've already set up. It's empty right now, and we're gonna push it into here. So I'm having it create a readme that'll tell users, our other team members, basically, is in our example here. We're deploying to Railway. Here's what you'll need.\u003C\u002Fp>\u003Cp>It'll mention the EMV variables we created here, because we're gonna need those. I'm actually gonna copy them now, so I have them ready. And then it it's gonna make sure everything's building properly with the prompt I gave it, and then it will push to GitHub for us.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. And I noticed it said, like, even if you don't have, like, the environment variable set up initially, like, it should still deploy. It should still work. But once we actually set the these values, we'll be able to actually fetch the thing and have the same result we have locally, but actually have it live and deployed on a URL.\u003C\u002Fp>\u003Cp>Speaker 0: Yep.\u003C\u002Fp>\u003Cp>Speaker 1: It's updating the ReadMe.\u003C\u002Fp>\u003Cp>Speaker 0: Yep. There's our ReadMe.\u003C\u002Fp>\u003Cp>Speaker 1: I remember the times where I had you know, I used to write stuff by hand.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah.\u003C\u002Fp>\u003Cp>Speaker 1: Now I just be like, you know, just instructions, just lines, like update, read me, and that's it. You know? Just just figure it out the rest.\u003C\u002Fp>\u003Cp>Speaker 0: Alright. I think it's gonna commit.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. I think it's already I believe I ran already a built. So, yeah, we see, like, the dot next directory. Maybe the build was under there. Yeah.\u003C\u002Fp>\u003Cp>It's running get the commands. Okay. I think it has committed everything here. So now\u003C\u002Fp>\u003Cp>Speaker 0: What did I do?\u003C\u002Fp>\u003Cp>Speaker 1: I think it's just the wrong link.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. It did. Okay. So we are gonna now go in here. We're gonna add a service.\u003C\u002Fp>\u003Cp>Speaker 1: Yep.\u003C\u002Fp>\u003Cp>Speaker 0: GitHub repository. And what did I call it?\u003C\u002Fp>\u003Cp>Speaker 1: Was it like AI agents or something like that?\u003C\u002Fp>\u003Cp>Speaker 0: AI agents. Not sure. Oh, boy. Hold on. I pasted it down here.\u003C\u002Fp>\u003Cp>Speaker 1: You should be able to see, like, all repos, but maybe you have too many. So, like\u003C\u002Fp>\u003Cp>Speaker 0: I have so many. I really need to clean it up.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Yeah. That's me.\u003C\u002Fp>\u003Cp>Speaker 0: Okay. Hold on. I'm gonna\u003C\u002Fp>\u003Cp>Speaker 1: Give me to refresh. Yeah. Yeah.\u003C\u002Fp>\u003Cp>Speaker 0: Just don't wanna have my mega list of repos up on the screen so one's.\u003C\u002Fp>\u003Cp>Speaker 1: But I think you might need to refresh the, so I tried to push the code. Right? Oh, it didn't.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. I don't know if it\u003C\u002Fp>\u003Cp>Speaker 1: pushed. I mean, I guess you need to stop. Yeah. See\u003C\u002Fp>\u003Cp>Speaker 0: It's getting me with the permission things. You're right.\u003C\u002Fp>\u003Cp>Speaker 1: I should\u003C\u002Fp>\u003Cp>Speaker 0: just let it do whatever it wants.\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. But now it's I don't know if it's pushing it, and then it should be good. So now if you go to right way, I would say if, if you refresh, you should be able to see it. Yeah. We have code.\u003C\u002Fp>\u003Cp>Speaker 0: The code is now here.\u003C\u002Fp>\u003Cp>Speaker 1: Yep. Now when you add, we might need to like refer, I think it should just fetch, but you can try.\u003C\u002Fp>\u003Cp>Speaker 0: Does it not like my\u003C\u002Fp>\u003Cp>Speaker 1: Maybe just try refreshing, like, the, entire Page. Dashboard. Yeah. Yeah. Just in case.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Okay. And Okay. Let's try again.\u003C\u002Fp>\u003Cp>Speaker 1: Interesting. I mean, if you want, what you could do is maybe it's just an issue with, like, get up not pulling. So, like, you can copy the full URL. And when you add, you can just paste that URL.\u003C\u002Fp>\u003Cp>Speaker 0: Okay. The it looks like it found it like that. So let's\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Yeah.\u003C\u002Fp>\u003Cp>Speaker 0: Alright. So that is how you connect it, and then\u003C\u002Fp>\u003Cp>Speaker 1: we're gonna just need to hit deploy. Yep.\u003C\u002Fp>\u003Cp>Speaker 0: Yes. But I wanna put it up by the other ones first. Yeah. And that was\u003C\u002Fp>\u003Cp>Speaker 1: in there.\u003C\u002Fp>\u003Cp>Speaker 0: Yes. So I wanna also add our e and v variables before we get started.\u003C\u002Fp>\u003Cp>Speaker 1: So they're actually automatically inferred, like, at the bottom. Oh. You can see them. You were right. Suggest variables, and then you can update these values and click add, then you're good to go.\u003C\u002Fp>\u003Cp>Speaker 0: There you go. So let me grab where I pasted them over here.\u003C\u002Fp>\u003Cp>Speaker 1: You can also have, like, go the route you were going through, like the raw editor, copy and paste that other thing, but just just wanted to point it out. Because, like, to me, it's like, it's one of those things that when you don't know it, you're like, oh, that's neat. That's kinda how I always felt about it the first time I tried it.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. You're right. That is cool that it already kind of pulled up what we needed.\u003C\u002Fp>\u003Cp>Speaker 1: But,\u003C\u002Fp>\u003Cp>Speaker 0: yeah, you could also open raw editor and just paste in the whole EMV file that we had ready.\u003C\u002Fp>\u003Cp>Speaker 1: Yep. And then you can click add to add everything.\u003C\u002Fp>\u003Cp>Speaker 0: Yep. So these this is the token we have in our local one, and then this is the URL of the same thing in Railway. You guys can also do internal URLs. Right? Is that something we could use here if we wanted to use the local URL\u003C\u002Fp>\u003Cp>Speaker 1: of this? So you can do, like, a dollar sign, two curly braces, and then you should get auto complete for, like, the URL. So, actually, you will be able to see a line where they're both connected if you wanna try it out. And then what will happen is, now whenever if the URL changes, it will just automatically update the other one as well and redeploy it.\u003C\u002Fp>\u003Cp>Speaker 0: Okay. So that's like using the internal\u003C\u002Fp>\u003Cp>Speaker 1: URL reference variable. So I I believe you the way it would work is you just do, so the on directives and the variables themselves, like, if you go to variables. I believe what's the yeah. I guess it's public URL is the value that we will want. So wait.\u003C\u002Fp>\u003Cp>If you go now to the AI agency service and then go to the next public directors URL, because that's the URL value you wanna set. Right? So that more menu and then edit. And then if you do, dollar sign and then double curly braces, you'll be able to actually pull in so I believe the name of the service is directed. So you just do direct us, and then it's gonna be dot yep.\u003C\u002Fp>\u003Cp>Do you see? And then you just choose the public URL Yep. From this list. Got it. So now when you choose it, it's just gonna automatically give you the right URL and it will just be able to read it.\u003C\u002Fp>\u003Cp>So if the other one updates, you just do it. But for this, you're gonna need to hit redeploy again. So it will take some time. So I'd say we can do this after you can cancel. But yeah.\u003C\u002Fp>\u003Cp>And I would say there's actually another cool feature because might as well, if you go to settings, like, okay. It's done. So we can actually test the thing. But we can update, like, the domain to, like, a actual domain because you can now buy domains on railway.\u003C\u002Fp>\u003Cp>Speaker 0: Nice. Okay.\u003C\u002Fp>\u003Cp>Speaker 1: Yep. So this will give to you just a domain for, the front end itself, and it will have, like, a .up.railway.app at the end. You'll need to specify a port, which I'm not sure what it is on. I I think it should be 3,000. Like, 3,000 should just work.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Yeah.\u003C\u002Fp>\u003Cp>Speaker 1: And if we visit it, might be the incorrect port. You could try eighty eighty as well because that's what I was suggesting in the beginning. Yeah.\u003C\u002Fp>\u003Cp>Speaker 0: There we go.\u003C\u002Fp>\u003Cp>Speaker 1: There we go.\u003C\u002Fp>\u003Cp>Speaker 0: So there is our front end deployed.\u003C\u002Fp>\u003Cp>Speaker 1: Yep. And we have a live app.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Awesome. So that was pretty fast. We went from nothing to a fully deployed front and back end here. Let's see.\u003C\u002Fp>\u003Cp>Was there anything else we should know about Railway and getting this set up here?\u003C\u002Fp>\u003Cp>Speaker 1: I mean, if you hit deploy now, it will just redeploy, and you'll actually be able to see on the canvas. There's, like, an arrow, and this, service, like our front end is, referencing the direct us. And that's how you have, like, a connection between the two. And, yeah, other than that, it's like, you can just have everything in one place. Again, like, if you go now to the AI agency service and you click on settings and you also go to networking, from the right.\u003C\u002Fp>\u003Cp>Yeah. And then you can actually choose a custom domain. So when you click on it, you should actually be able to buy a domain straight from here. So like, if you come up with that domain, you can just do it from here as well. Or if you go to like ferry.com\u002Fdomains, you can do it.\u003C\u002Fp>\u003Cp>But yeah, I would say that's kinda like you now have fully like everything in one place, but you know, took us not a lot of time, and we have, like, a fully, fully deployed working back end, front end. Just, yeah, everything is in one place.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. So why don't I recap what we did? We started with a simple idea launching a fictional AI automation agency. We launched infrastructure on Railway. We generated a structured back end with Directus.\u003C\u002Fp>\u003Cp>We populated it with fake content. We generated the front end using Cursor. We deployed the whole thing back to Railway. That's the build it part of today's theme. The key takeaway is that when interfaces can be generated quickly, the back end data model becomes the strategic layer.\u003C\u002Fp>\u003Cp>If someone watching wants to try this themselves, what's just the easiest way to get started on the railway side of things?\u003C\u002Fp>\u003Cp>Speaker 1: The easiest way, I would say going to the template you shared would be the easiest place. And if they, you know, wanna use direct this. And also, by the way, I forgot to mention. So if you actually, like, close the, service here, there's like an agent in the top right corner as well, where you can actually also ask questions about your stuff, like actually ask it to make changes and it should be able to make changes, directly. So I believe if you just say like, oh, update, make this update, it should figure things out on its own.\u003C\u002Fp>\u003Cp>Speaker 0: Awesome. So you guys got some AI in here too?\u003C\u002Fp>\u003Cp>Speaker 1: Yeah. Yeah. In case someone wants everything in the dashboard, they can, but also if they want, the, to use whatever their coding agent of choice, they should be able to do it. But yeah.\u003C\u002Fp>\u003Cp>Speaker 0: Cool. Well, thank you for joining us. I hope everybody has a good rest of your Leap Week.\u003C\u002Fp>\u003Cp>Speaker 1: Sounds great. Thank you so much for having me.\u003C\u002Fp>\u003Cp>Speaker 0: Yeah. Thank you for joining us.\u003C\u002Fp>","Hi, everyone, and welcome to Leap Week. My name is Lindsay, and I'm an engineer on the direct us team where I mostly work on integrations and developer experience. For day two of Leap Week, the theme is build it, run it, sell it. The idea is to show how quickly you can go from an idea to a real production application using modern tools. We'll be focusing on the first part of that story, build it. For this session, we're going to build a fictional AI automation agency. The idea is that this agency helps companies implement AI workflows for things like marketing, automation, research, sales outreach, and internal knowledge assistance. So by the end of the session, we'll have a working back end for this agency, generate a front end website for it, and deploy the whole thing live. But before we can build anything, we need somewhere to run our back end and infrastructure. I'm joined today by someone from the Railway team. I'll let them introduce themselves and tell us a little bit about Railway before we start building. Awesome. So, thank you so much for having me. My name is Mohammed. I'm a general engineer at Railway. And at Railway is I would say in a nutshell, it's a all in one intelligent cloud provider. You can deploy pretty much anything, databases, back ends, front ends, queues, whatever. Like, you can just deploy it. You can just run it. And, yeah, essentially, we allow you to deploy anything, and we really focus on making it easy for you to go from zero to deploy it as easy as possible. Yeah. So excited about what we'll build today today. Awesome. So the first step in getting any project off the ground is gonna be your infrastructure. And instead of spending hours setting up servers and databases, Railways gonna let us start from a working template. For this demo, we have a template we've already created. It's a pretty simple one. It's gonna have everything you need to get started. We do actually have another version as well that launches with CMS collections already started if you want a starting point. But for this, we are going to use our blank one. So I'll go ahead and launch that template here. Yeah. So a little bit of context. Templates essentially encapsulate a set of infrastructure components, all the configuration. And here, we essentially have, like, a database set up, post press red as a bucket for object storage, as well as directives, which runs our back end. So, like, directives here would be, a server that we actually use. And now, like, everything here is neatly on the railway canvas. You can see everything grouped under, like, the directors group. And, essentially, like, here, you have a project, and a project is where you can have essentially everything deployed. So, like, in this example here, you have practice running. Maybe you will have, like, the front end, so that will be a separate service. And that's pretty much all we need to get started. So, like, all you need to do is you can sign up for free, and then you can go to the link to deploy the template. You click deploy now. You can either deploy to a new project or an existing project. If you choose, like, the path we're going through right now is deploying to a new project, this is the experience you'll get. And now you can see, actually, it's like we have the difference for versus, they're being deployed. So, like, Post JS, that's the database. And then we have Redis and as well, we have, DirectThis, which is also just cute. So, like, one thing after another will be deployed. So, like, in a, hopefully, a minute or two, everything should be ready, configured. You don't even need to, like, touch an editor, and it just works. So that's pretty much yeah. That's pretty much the overview for templates. Yeah. So it looks like Like, this is also deploying I see it's so this one is actually deploying from a Docker image, and this is the official one, for directors. I assume this is the latest version. Right? Yep. This is our latest version. Yep. We make sure to keep this updated. That's awesome. But, yeah, if, let's say, for example, you deploy and you want to get the latest version, of, let's say, Directus. If you actually go to settings, you'll be able to see the source image. There's like automatic updates. You'll actually be able to see if there's an automatic update. You just click apply. You deploy it. It just works if you want like the latest version. So it should be very easy to set up, but so far we can now we see, we have the database ready. Red is ready. The bucket is created, and now the last component direct list is being deployed, which hopefully should be ready Yep. Very soon. And we're live. We'll click that. So in just a couple minutes, we've gone from nothing to a running back end with a database, API, and admin interface. So as part of our setup process, we'll create our first admin user here. Give it a password. So I assume this happens as the first initial step. Is this correct? Yes. Yep. Okay. When you first set up, it'll ask you to create your admin user. And here we go. We're in our, UI here. And to get started, we are gonna do a little bit of, setup for our AI assistant because we are going to be using AI to create most of this. So Directus has a built in AI assistant that can help you. It can generate schema content, and it has knowledge of what's going on in your instance. So if you go into settings here, we'll give it a key. Oops. Let me grab my key. Yeah. I think you just said One password. Gotta have that security. Yeah. Alright. And then we are also going to set up our MCP for later because we'll be using that when we do our front end. So here, this is, like, the Directus MCP or, like, this is an MCP server for our deployed direct the Directus instance? Yes. It is. Yep. So our cursor, which we'll use to do the front end, will be able to access all of our data here. And it'll depend on the permissions we give it, what it will be able to do. But we also have this in app AI assistant here. So we're gonna use that to generate our schema. So I already have some So quick question. For someone who's not a 100 to 100% familiar, let's say, with all of the features, essentially, what we're doing is now we have a like, our complete back end deployed running. And part of this back end, you can actually expose an MTP server, and then I can connect to this MTP server via call code, code x cursor, whatever coding agent of my choice, and then just tell it to do stuff for me. That'll be reflected in my live deployed instance, and now it's just that's it. Yeah. Yeah. We'll see that a little bit later, but we're gonna do something similar here. We also have our AI assistant that's in the UI here, and we are gonna use that to build our back end, our collections, our schema. So right now, we have no collections created. We don't have any data tables, basically. So I'm gonna go ahead and prompt it with a prompt that I created earlier, and it's going to create the structure of our back end. And the idea is, like, a collection would be, like, an object, right, for something. Yeah. So it's gonna be, like, your data tables. So it'll have, like, an overlying day data table where we'll have one called site settings, and that's gonna be the settings for the website. We'll have one for case studies. Mhmm. That's gonna have case studies for our AI agency. Actually, I can show you the prompt here a little closer. So we've we're creating our collections, and we're giving it different fields. So we're giving it, like, a tagline, a headline. We're gonna give it different services that this AI agency does, give it some fields there, and we're gonna create playbooks and case studies out of those, so things you might need in an AI agency marketing website. And the AI on our end is going to do that work for us. So we've told it in our prompt what we want. We didn't really tell it what type of fields they need to be, if they're strings or numbs or any of that. It'll infer that if we want. And, actually, one nice thing is if you don't know what schema you want, you can talk it through with the AI assistant. It can give you suggestions of things that you might wanna add for a data table. So it's gonna ask questions too back to the user. So it's saying it's ready to create these, and it's saying, do we wanna proceed with its defaults? Do we wanna review each one and decide? Or do we wanna have, like, it be super minimal? I'm gonna say just create it. Let's see what it comes up with. It's asking how we should store the icon. So let's just say a single file upload. How do we wanna store the pricing start? We don't really care about that. We'll just pick something. Yeah. Yeah. There's a there's a recommendation. So I think we can go with it. Yeah. It's collaborative. It's gonna ask you how you wanna do things. It's not gonna just bowl over your user and, like, do everything for you. It's it's not just an AI tool that's gonna do it for you. It is a human in the loop type of experience where you can make sure it's doing something you actually want it to do. And it's gonna tell you everything it's doing while it's doing it. So you can see here. It's planning it out. You can see the thinking if you wanna know, like, what it's thinking while it does it. And you can do the different models if you're you have a different one. We're currently just doing, GPT five. In those settings we set up earlier, you can change that out. Like, if you want something different going on, you can say which ones are allowed, because this UI is actually not just for your admin. You can allow other people here. So you can have other people with different access levels able to come in and modify your content once it gets created or your data tables. So you can let them use the AI assistant, and you can set what permission levels they have. So it's gonna ask for different tools like creating the collections if it can do it before it does it. You'll see here's our data models now. It's created a couple of them. We've got our case studies one. It's starting to add the fields into these. In the past, you would have to do this all manually and create collections. Now you can just use the AI to do it without having to worry too much. So it's a lot less manual work. But we could have done the same thing by using the MCP thing. Right? Yep. You can. Yeah. We just wanted to give people an in app way to do it a little faster and simpler for people who maybe aren't don't have a data model or AI model they like to use outside of here. Yeah. Definitely makes sense. It's just like in my mind, I really like the flow of using some like, I use Claude a lot. So when I use Claude, I'm like, okay. I want a complete end to end flow. Like, what what would it look like if I don't leave cloth? Can I actually do the stuff I wanna do? And it seems like the answer is yeah. But there's also And that's actually that's what I like to do as well as I'll go in cursor, and you could start the whole front and back end directly in cursor using the direct SMCP, and it can create it in here for you. We're gonna do the front end in Cursor just so we're seeing that too a little bit, but I wanted to use a little of the AI assistant as well. Yeah. Yeah. I'm actually in the process of making sure that our experience of getting started to be super seamless. So we already have, like, a regular CLI. We have, like, Asian skills, and, we're going to essentially keep shipping more, features through the CLI. So, like, it should be able to you say, like, oh, I wanna deploy direct us. It should be able to go through the templates marketplace, find the most up to date template deployed for you, and then it should be able to also figure out, like, the deployment is finished. Let me continue. And then you just have, like, a complete workflow where you don't even have to go to a templates page. So that's kind of, like, what's That's awesome. In the works. But yeah. Yeah. I'm excited for that. That'll be good. Yeah. That's I think it asked a question at the end. Yep. So now it is done. It's gonna, like if you've used this model, it constantly asks you other things it wants to do, but it is done. It's created the fields. So we've got that part done. Now we're gonna give it some fake content in these fields. So that's the database table. And now we're gonna fill these tables with actual, like, fake data. So I'm gonna paste that in. So when we go into here, you'll see no content yet, and it's gonna start filling those in. But you can see, like, these are the fields. So if a nontechnical user came in and wanted to give it a new name or a new tagline for their website, they can do it in here without having to touch the code at all. So it's a nice way to work with both your front end team and your back end team and your content team and not have, people who are super technical not able to get things done. While this is doing it, we are actually gonna get finished setting up for our MCP, actually. So we're gonna go here in our docs. We have a little section for setting up the MCP in different tools. We're gonna use cursor. So I'm gonna click add to cursor here. Nice easy button, and it will set us up into our settings. And we are going to go grab our URL from here. So this is our directest instance URL, and we'll pop that in here for our MCP setup. So it's just your instance URL slash MCP, and then you're gonna wanna set up a token. So let's make sure it has the permissions it needs. When you're connecting to external tools like cursor, it's always a good idea to scope your permissions so that it's not able to completely nuke your whole setup. You know, AI likes to be overeager. So Yeah. Yeah. It's like, let's start all over. Yeah. We need a fresh start. We we recommend setting up a token for it specifically and not using your admin token. So you wanna go to your user roles here. You'll create a new role, name it whatever you want, and you'll add a policy. We're gonna give it view only policy. So you'll add your collections, give it read only access. That way it can't, like, update or delete or anything when it's in the MCP. If you want the MCP to do that, you can totally give it permissions to create, read, and update, and do whatever you want with it. But I recommend giving it, like, the lowest level of access at the start so that, you know, it's not gonna go in and break your whole setup here. Mhmm. So we've got our policy. I'm gonna create a new user too to grab a token from. So if we scroll down to our tokens, we're gonna generate it and save. And then we'll go back in here, and that's where you put in this bear here, authorization code. You put that token, click install, and that will install our direct as MCP encursors. So now we're ready to go for that when this is finished. So let's go check on it. So what we've done is create a role, and then the role has the token? Yes. So the the user is what who has the token. So this MCP user has the role that we created. I created it all within the same page, but this person has the role, and then they have the token. Mhmm. So if you go into our settings here and look at the MCP role, we've got our view only policy that we set up Mhmm. For our different collections. And the users in the role is that MCP guy. So now it looks like it did finish the content. If we go back into our content page, now you can see it's filled it in with our sample data. I gave it a few guidelines of the type of things I wanted it to have, but, otherwise, it just kinda came up with this on its own, sample content. So it's given us a starting point. So now that we've got some sample content, we are going to go back in to cursor. I'm gonna grab this really long prompt I've created telling it basically we are an AI agency. We're creating our website. I gave it I told it to use next. I also made sure to tell it to use the Directus MCP server. This might take a minute, so I'm gonna go ahead and get it started. We are in a completely empty folder here, so I'm gonna have it do all of the coding for me. Let it create what we need here. It's gonna make sure to look at our schema via the MCP. So we want it to not guess on what structure we have of our data. We want it to actually look at it. Like we said, we could create the data directly from Cursor via the MCP. You would have to make sure you give it, the right permissions. So that that might have to be an admin token in order to generate new schema with your MCP. So if you're comfortable doing that, you totally can. Interesting. So once that's done, we'll have something that's working like a front end running locally that's fully functional, pulling data from our deployed direct us instance. Correct? Yep. Okay. Yep. And any changes you make in direct us would be reflected in your local front end that you've got here. And then as the final step, we are going to do to deploy this front end to railway so that you have a full production application front end and back end. Yeah. I have cursor set up, so I could just automatically runs everything. I don't, like, manually approve stuff. I just I don't trust it. Yeah. I gotta make sure I know what it's doing. I've been burned before. I don't know. The the models are they keep getting better and it's like, I'm like, I would've done the same thing. So I don't know. I feel like it's just faster and you can just like, you know, sit back, relax, grab your favorite drink while it gets to work. You come back to it and you're like, woah, this thing actually works. So Yeah. I think I'm a bit of a control freak. I can't let it just just go. I'm telling you, like, once you let go and you see, like, it actually works and it's like, I think, I think you'll be plenty surprised. Then you're like, wow, that's actually a really fun. Cause like now, you know, if let's say for whatever reason, someone brings you on Slack, you can play for guys. Like, you come back to us, like, oh, I forgot to approve. Yeah. That it's done. Yeah. Yeah. I I I think, it's a much more fun experience when you just let it kinda do its own thing without you steering it. But I do understand the, appeal of, you know, wanting to see what it does. But also, like, we're starting from scratch, so there's little, like, room for error, I would say. But yeah. There's nothing for it to break. It's creating it from scratch. So Yeah. Yeah. We're pretty comfortable letting it just do what it wants to do here. So it looks like it actually has gotten the types if you look. So it pulled this from our MCP. It checked in, Directus, and it got our back end and what types we have there. So that's good. Looks like it's I did tell it also to create a little connection a net little connector to direct us here, a fetcher. So it looks like it's done that, getting different gets to get the data. And it will be able to pull in our deployed, like, like, URL and actually set it as an environment variable? Yes. It will. So I it might make me make the e n v, but I've got it. It's supposed to use the direct URL here, and it's supposed to use the token that we created. So, we'll see. I've done this a couple times just to see how it would do, and it it doesn't always know to grab the URL. If it's really smart, it can grab it from the cursor settings, and it can get the token and the URL, but it doesn't always manage that. You'll notice I'm in auto mode, so it's a different model each time, so it might not be a very smart model. But we at the end, we'll give it the proper e n v variables, and it'll be able to connect to our deployed instance. Excellent. Yeah. Honestly, again, I I like my to choose my own models, You should use, like, a, you know, g p three five point four or, like, an Opus 4.6 and because I don't know. I don't like that, I would say I wouldn't trust the auto. Like, which one would affect? Like, I wanna be I want a consistent experience. So then if, like, anything starts feeling weird, then I can blame the model. I was like, oh, you know, like, Opus is acting really dumb lately, that sort of thing. So It it does just feel like from one day to the next, though, even if you have a model that you know is good, it it doesn't always behave exactly the same. Yeah. Sometimes it feels like, you know, you're you're just rolling the dice. Sometimes it's like, woah. This is it. Like, this is the next big thing. I was like cause like, sometimes I, I, I asked it a question and I wanted it to check against like, you know, our internal, GitHub repo. And it's like, Hey, do we support this thing? It was like, yes, you do. And I told her like, oh, show me the code. It's like, after checking, apparently, this is not supported. I was like, come. Yeah. And this is like the smart and I told it, like, oh, you know, UltraThink, you know, work really hard, but it's like, I guess sometimes it's just, you know, not that type of day. But I think here it's like it's it's it's working well, based on what I'm saying. Yeah. It looks like it's getting something. Yeah. It's messing around with the configs now. Interesting. What is what's it doing to the Next. Js config if we open it up? Just out of curiosity. Technically, you know, we're we're not supposed to look at the code, you know, and, because, you know, I was just shipping it, but I'm also curious. Okay. It's setting, like, the assets URL, I assume, for, like, files and stuff. That's pretty awesome. Yep. And while it does that, I'm actually gonna grab my last prompt, which is a prompt to have it, push to GitHub so that we can get it ready for deploying the front end. I'm gonna grab that. Go back in here. I think it's almost done. It feels like it's it's doing the read me. So Yeah. Probably at the end. But yeah. I would say, there is actually a like, you you can definitely, you know, start with this workflow. It's totally fine. But if you have, like, the AOS CLI setup, it could just deploy it to your existing project. And this way you don't even have to worry about git being installed, get, having a GitHub repo or anything. But when you have it with git, well, the benefit is, anytime you make, changes through the front end, it will just automatically deploy. So it's like it's it's all trade offs based on, you know, how, structured and organized you wanna be. But yeah. Yeah. That makes sense. Alright. We did not create our E and V, so we're gonna create that real quick. And I would say that's the responsible thing for it to do. Yes. I don't didn't really like that it created its own ENV and, like, grabbed my things from the MCP. That was a little, uncomfortable. Yeah. Yeah. It then gets better this way. Yes. Look at our URL. Alright. And then we're gonna run it locally just to see it working. What does the static token do? So that's the one we created with for the MCP. We're using it here too. Okay. So that is the permissions that we allow in Directus. So if you give it, like, no read permission, it won't be able to see the content. If you give it access to read, like what we did, we gave it access to read all of these collections, then it can read the data. So that's where, like, you probably want a separate one from your MCP. We're just in the interest of time having it Yeah. One that is view all for all of our collections because we don't need to make changes with the MCP. But you probably would have permission to update and delete with your MCP as well. So you probably want wanted a separate one for your front end. So you'd probably wanna create a user role for your front end that gives it different permissions. Otherwise, you can have it set up to use your public permissions, which controls what the API data is, available without authenticating. But it's safer if you do give it a token so that your front end is only able to grab what you want, not, like, everything. So let's see if it worked. Moment of truth. Okay. There it is. Looks like some images are broken, but we've got a full website here. You can navigate around. It's got some prompts. It's got a button to get in touch. We've got a full website here. Obviously, you'd wanna iterate on it, fix the images, and, do a little honestly, it looks great. It doesn't need too much design or anything even. So that's a good starting point. We are going to go ahead and say good enough. The images can get fixed later. We are going to deploy this. And, of course, I copied other things, so I need to go back and get my prompt. So like you said, there's other ways to do this, but we are gonna have it pushed to a GitHub repo. We're pretending we have other people who wanna manage this code, and we want it in some place that we would have access to it as a team. So we've have a GitHub repo we've already set up. It's empty right now, and we're gonna push it into here. So I'm having it create a readme that'll tell users, our other team members, basically, is in our example here. We're deploying to Railway. Here's what you'll need. It'll mention the EMV variables we created here, because we're gonna need those. I'm actually gonna copy them now, so I have them ready. And then it it's gonna make sure everything's building properly with the prompt I gave it, and then it will push to GitHub for us. Yeah. And I noticed it said, like, even if you don't have, like, the environment variable set up initially, like, it should still deploy. It should still work. But once we actually set the these values, we'll be able to actually fetch the thing and have the same result we have locally, but actually have it live and deployed on a URL. Yep. It's updating the ReadMe. Yep. There's our ReadMe. I remember the times where I had you know, I used to write stuff by hand. Yeah. Now I just be like, you know, just instructions, just lines, like update, read me, and that's it. You know? Just just figure it out the rest. Alright. I think it's gonna commit. Yeah. I think it's already I believe I ran already a built. So, yeah, we see, like, the dot next directory. Maybe the build was under there. Yeah. It's running get the commands. Okay. I think it has committed everything here. So now What did I do? I think it's just the wrong link. Yeah. It did. Okay. So we are gonna now go in here. We're gonna add a service. Yep. GitHub repository. And what did I call it? Was it like AI agents or something like that? AI agents. Not sure. Oh, boy. Hold on. I pasted it down here. You should be able to see, like, all repos, but maybe you have too many. So, like I have so many. I really need to clean it up. Yeah. Yeah. That's me. Okay. Hold on. I'm gonna Give me to refresh. Yeah. Yeah. Just don't wanna have my mega list of repos up on the screen so one's. But I think you might need to refresh the, so I tried to push the code. Right? Oh, it didn't. Yeah. I don't know if it pushed. I mean, I guess you need to stop. Yeah. See It's getting me with the permission things. You're right. I should just let it do whatever it wants. Yeah. But now it's I don't know if it's pushing it, and then it should be good. So now if you go to right way, I would say if, if you refresh, you should be able to see it. Yeah. We have code. The code is now here. Yep. Now when you add, we might need to like refer, I think it should just fetch, but you can try. Does it not like my Maybe just try refreshing, like, the, entire Page. Dashboard. Yeah. Yeah. Just in case. Yeah. Okay. And Okay. Let's try again. Interesting. I mean, if you want, what you could do is maybe it's just an issue with, like, get up not pulling. So, like, you can copy the full URL. And when you add, you can just paste that URL. Okay. The it looks like it found it like that. So let's Yeah. Yeah. Alright. So that is how you connect it, and then we're gonna just need to hit deploy. Yep. Yes. But I wanna put it up by the other ones first. Yeah. And that was in there. Yes. So I wanna also add our e and v variables before we get started. So they're actually automatically inferred, like, at the bottom. Oh. You can see them. You were right. Suggest variables, and then you can update these values and click add, then you're good to go. There you go. So let me grab where I pasted them over here. You can also have, like, go the route you were going through, like the raw editor, copy and paste that other thing, but just just wanted to point it out. Because, like, to me, it's like, it's one of those things that when you don't know it, you're like, oh, that's neat. That's kinda how I always felt about it the first time I tried it. Yeah. You're right. That is cool that it already kind of pulled up what we needed. But, yeah, you could also open raw editor and just paste in the whole EMV file that we had ready. Yep. And then you can click add to add everything. Yep. So these this is the token we have in our local one, and then this is the URL of the same thing in Railway. You guys can also do internal URLs. Right? Is that something we could use here if we wanted to use the local URL of this? So you can do, like, a dollar sign, two curly braces, and then you should get auto complete for, like, the URL. So, actually, you will be able to see a line where they're both connected if you wanna try it out. And then what will happen is, now whenever if the URL changes, it will just automatically update the other one as well and redeploy it. Okay. So that's like using the internal URL reference variable. So I I believe you the way it would work is you just do, so the on directives and the variables themselves, like, if you go to variables. I believe what's the yeah. I guess it's public URL is the value that we will want. So wait. If you go now to the AI agency service and then go to the next public directors URL, because that's the URL value you wanna set. Right? So that more menu and then edit. And then if you do, dollar sign and then double curly braces, you'll be able to actually pull in so I believe the name of the service is directed. So you just do direct us, and then it's gonna be dot yep. Do you see? And then you just choose the public URL Yep. From this list. Got it. So now when you choose it, it's just gonna automatically give you the right URL and it will just be able to read it. So if the other one updates, you just do it. But for this, you're gonna need to hit redeploy again. So it will take some time. So I'd say we can do this after you can cancel. But yeah. And I would say there's actually another cool feature because might as well, if you go to settings, like, okay. It's done. So we can actually test the thing. But we can update, like, the domain to, like, a actual domain because you can now buy domains on railway. Nice. Okay. Yep. So this will give to you just a domain for, the front end itself, and it will have, like, a .up.railway.app at the end. You'll need to specify a port, which I'm not sure what it is on. I I think it should be 3,000. Like, 3,000 should just work. Yeah. Yeah. And if we visit it, might be the incorrect port. You could try eighty eighty as well because that's what I was suggesting in the beginning. Yeah. There we go. There we go. So there is our front end deployed. Yep. And we have a live app. Yeah. Awesome. So that was pretty fast. We went from nothing to a fully deployed front and back end here. Let's see. Was there anything else we should know about Railway and getting this set up here? I mean, if you hit deploy now, it will just redeploy, and you'll actually be able to see on the canvas. There's, like, an arrow, and this, service, like our front end is, referencing the direct us. And that's how you have, like, a connection between the two. And, yeah, other than that, it's like, you can just have everything in one place. Again, like, if you go now to the AI agency service and you click on settings and you also go to networking, from the right. Yeah. And then you can actually choose a custom domain. So when you click on it, you should actually be able to buy a domain straight from here. So like, if you come up with that domain, you can just do it from here as well. Or if you go to like ferry.com\u002Fdomains, you can do it. But yeah, I would say that's kinda like you now have fully like everything in one place, but you know, took us not a lot of time, and we have, like, a fully, fully deployed working back end, front end. Just, yeah, everything is in one place. Yeah. So why don't I recap what we did? We started with a simple idea launching a fictional AI automation agency. We launched infrastructure on Railway. We generated a structured back end with Directus. We populated it with fake content. We generated the front end using Cursor. We deployed the whole thing back to Railway. That's the build it part of today's theme. The key takeaway is that when interfaces can be generated quickly, the back end data model becomes the strategic layer. If someone watching wants to try this themselves, what's just the easiest way to get started on the railway side of things? The easiest way, I would say going to the template you shared would be the easiest place. And if they, you know, wanna use direct this. And also, by the way, I forgot to mention. So if you actually, like, close the, service here, there's like an agent in the top right corner as well, where you can actually also ask questions about your stuff, like actually ask it to make changes and it should be able to make changes, directly. So I believe if you just say like, oh, update, make this update, it should figure things out on its own. Awesome. So you guys got some AI in here too? Yeah. Yeah. In case someone wants everything in the dashboard, they can, but also if they want, the, to use whatever their coding agent of choice, they should be able to do it. But yeah. Cool. Well, thank you for joining us. I hope everybody has a good rest of your Leap Week. Sounds great. Thank you so much for having me. Yeah. Thank you for joining us.","3d92d748-e9d2-4d03-8a5b-cc78b8273739",[70,71],"533d89f7-b567-47f2-8dbf-940e4ff396e0","37090185-0684-40ef-97c0-e860f2b2df77",[],1781213209958]