Machine transcription with reviewed terminology and names. Some recognition errors may remain. Timestamps refer to the livestream; subtitle downloads are timed to the session clips.
3:48:40People have been saying that SaaS is dead for a while now.
3:48:43It's not even a hot take at this point.
3:48:45Why pay someone to build and maintain the tools you use when you could just do it yourself?
3:48:50Just ask your agent to do it.
3:48:52And for me, there's one missing piece that rarely gets mentioned.
3:48:56And it's the fact that there's a gap, a big gap between working software that got one-shotted
3:49:01and software I would actually use.
3:49:04Software that sparks joy, that is polished, that feels good.
3:49:07And I actually went on a journey to replace a bunch of SaaS providers we were using internally
3:49:11at Railway for marketing purposes.
3:49:13And I just built an all-in-one tool that we use.
3:49:16The tool allows us to manage content
3:49:17for the Railway website, change log, blog.
3:49:20We can use it to send emails,
3:49:22post on all of our social platforms at the same time,
3:49:25manage the website's media.
3:49:26There's even a mini CRM so that when we're working
3:49:29with external creators, we can keep track of that.
3:49:32There is task tracking, launch planning,
3:49:34because you wanna coordinate multiple things
3:49:36and when they'll go out, there's a calendar.
3:49:38and at the same time there's like comments real-time editing and it's all built using
3:49:45agents i would say i felt the breakthrough when gpt6 came out and all of this is running on railway
3:49:52everything just auto deploys and let me essentially now walk you through the principles i use
3:49:58to actually make sure that when i want to add a feature it's you know actually polished and
3:50:04matches the rest of the app and at the same time the principles that you can pull from and apply
3:50:11when building tools of your own so instead of doing my talk from my boring office i decided to
3:50:16switch things up and do it outside it is absolutely beautiful right now i'm in edinburgh scotland
3:50:21and it's just unreal like i'm on a bench outside and this is my view definitely a 10 out of 10
3:50:30bench. Now, this type of talk for me feels a little bit different than what I'm used to, because
3:50:36I essentially will walk you through the journey of me building our own internal tools and the
3:50:42opinions I've developed. But I also understand that the landscape changes very, very quickly.
3:50:48So like anything I say might change completely in like the span of months. Who knows? So I don't
3:50:57know consider this my disclaimer like but the time of recording is fall 2026 it's october and
3:51:04yeah so anyway with that out of the way uh let's talk about the first point which is text acts don't
3:51:12matter now a lot of people say that it doesn't matter what programming language you use what
3:51:17framework you use you can just ask your agent to go do anything and those are able to do it
3:51:21And while I agree to some extent, what I found is when you're building a tool that you will actually use, when you give the agent a proper code base with solid abstractions, what ends up happening is whenever you're working, let's say, on a new feature, you end up with a much more consistent result that is similar to what you've already built before.
3:51:44And the reason is, like, it's very simple.
3:51:47The agent doesn't need to reinvent the same, like, wheel every time it tries to add a new feature,
3:51:53or it doesn't need to introduce new patterns.
3:51:55It will just essentially follow existing patterns.
3:51:59And when it comes to, like, what you choose, I feel like it kind of doesn't really matter
3:52:03as long as you have a code base that is properly structured.
3:52:08But my go-to choice today is, one, TanStack libraries for anything when building on the web.
3:52:15And if you're building anything content-related, I would say definitely go with payload.
3:52:19Solid experience, great abstraction.
3:52:21I went down a big rabbit hole of comparing CMSs at some point because I don't believe that I should build it from scratch, even though I can.
3:52:30But again, who knows?
3:52:32Maybe the future will change.
3:52:33but yeah and that's kind of like the main first item that i kind of ran into
3:52:44when you use like a lot of sas products out there that you actually love they use every day
3:52:49when you're a user there's like a lot of things you don't notice but when you actually try to like
3:52:54replicate an experience you quickly find out how much care and thought goes into different flows
3:53:01So something as simple as, oh, I want to be able to send an email or to publish something.
3:53:09You can easily do it by having a good old server-side rendered app with forms that you submit.
3:53:18But if you try to build something that is going to be used every day that you will reach for
3:53:23a lot, you're like, oh, it shouldn't feel that way.
3:53:27And there are kind of like a bunch of patterns that I found that agents don't typically do unless you explicitly ask them to.
3:53:35And this is my list. So we'll go through them one by one.
3:53:39And the first one is like autosave. I found that if, like, let's say I'm building some form of document editor, autosave is kind of like a table states feature.
3:53:49So rather than manually clicking save to save things, you just ask it to automatically save progress in the background and to make sure that it's like a non-blocking action.
3:54:02So that you can, let's say, as you're typing, you're able to type and then it just automatically saves.
3:54:08And one other kind of like step further that you could do is to also prevent the browser tab from closing if there are unsafe changes.
3:54:20So it's kind of like these two layers that kind of guarantee that whatever it is that you're working on is just not going to be lost.
3:54:27Now, you can also go a step further and to add like real time collaboration so that you can have like concurrent edits.
3:54:35And again, my agent didn't struggle at all with adding the feature.
3:54:39It was like, I don't know, asking about it, reviewing how it would work.
3:54:45And then once that's done, just 700 million tokens later, I had real time.
3:54:50So I would say in general, this is the first thing that I find agents just don't do well from a product experience perspective.
3:54:58Another pattern I found that agents completely forget to implement is optimistic updates.
3:55:02So rather than showing your users a bunch of loading spinner whenever they're taking any action,
3:55:07you essentially show them the success state optimistically.
3:55:11And in the background, you're going to sync whatever change they made.
3:55:15Now, if the sync or the operation fails for whatever reason,
3:55:19you just roll it back and then tell them, like, oh, the operation didn't work.
3:55:23Now, of course, you shouldn't blindly have optimistic updates for everything.
3:55:27There are some scenarios where you want to show the user a loading spinner
3:55:31and they're going to wait to check that everything worked properly.
3:55:35But in general, having optimistic updates
3:55:37makes the overall experience feel significantly faster,
3:55:41much snappier, and it will just feel much more instant.
3:55:45So pretty sure you can just ask your agent like,
3:55:49hey, I would like to implement optimistic updates where it makes sense.
3:55:52Let's come up with a plan and it will tell you all the areas
3:55:54where it makes sense to do.
3:55:55Again, it's like it's not difficult for the agent to build it.
3:55:58you just need to ask and know exactly what to ask for. Now an extension of optimistic updates in my
3:56:04opinion is giving users the ability to undo certain actions. Now for example if you ask your agent to
3:56:11implement the ability to delete records what it will typically do is it will give you a delete
3:56:16button you click it might show a confirmation dialog you hit confirm and then it will actually
3:56:21do the deletion. But a much nicer experience is actually allowing the user to undo that action
3:56:28so they could delete you immediately remove the item they want to delete from the ui but you tell
3:56:34them like hey you have like a couple of seconds where you could change your mind and if they don't
3:56:40do anything that means you actually want to delete a thing then you actually implement the operation
3:56:45otherwise if they hit undo you just bring back the change so it's kind of like you do the operation
3:56:52in the background as well it's just kind of a delayed action which kind of ties into the next
3:56:57point, which is archiving, you should tell your agent you want the ability to archive records
3:57:03and be able to unarchive them. Because sometimes you want to hide something from the UI completely,
3:57:09or let's say from search results, and archiving will be the way to do it. And you shouldn't
3:57:14force your user to be like, well, if you don't want to see this, you have to delete it.
3:57:18Any much better experience is just the ability to archive records. So I think it is a pattern
3:57:24that should definitely exist.
3:57:26Another big one for me is URL state.
3:57:27You should ask your agent to persist in the URL
3:57:30whenever it makes sense, of course.
3:57:31And it's so useful because whenever you have,
3:57:33let's say, a list of records
3:57:34and you want the ability to filter or sort them
3:57:39and you just have a bunch of conditions,
3:57:42maintaining these things in the URL
3:57:45makes it trivial for you to give them to someone else.
3:57:47And then, let's say, another team member
3:57:49is just going to be able to see
3:57:51exactly what you want them to see.
3:57:52Now another small detail that agents don't implement out of the box is infinite scroll.
3:57:56I don't want to be clicking on lobe more over and over and over or like a next page button.
3:58:02As I scroll, I should just get the data.
3:58:05I shouldn't need to click next page.
3:58:07And then in the end, I'll just see that, oh, I've reached the end.
3:58:09It's a very minor thing.
3:58:11But also when you use something every day, I think it makes a huge difference.
3:58:15Now another pattern I found that made the output significantly better is telling the
3:58:20agent explicitly to use existing pages as references when it comes to the ui the data
3:58:26fetching and how it works and how the thing is overall structured and that made it possible for
3:58:32me to essentially define what good looks like once and then just reuse it for more features now
3:58:39to be completely honest sometimes even when you tell the agent like hey use page x as a reference
3:58:46sometimes it just doesn't follow it i don't know if models like get dumber sometimes but sometimes
3:58:52it does happen but overall i would say that my success rate is higher whenever i tell the agent
3:58:57to like well use an existing page as a reference because you also just end up with a more consistent
3:59:03experience so like as you're using your app you're like oh this just matches the other part of the
3:59:08app if they both share like the same concept so like if two parts of the app are essentially
3:59:14lists of records that you can create, update, and delete and list. Well, both of these parts
3:59:21should be pretty much identical. This list kind of brings me to my next point, which is whenever
3:59:26you have opinions about something, they should be packaged as agent skills. And then this way,
3:59:30your agent, whenever it's building something, it will actually follow those skills pretty closely.
3:59:36And as I was building all of our internal tooling, I found myself reaching for three agent skills.
3:59:41two of them are related to design engineering so one made it easier to have better ui better
3:59:47typography better layouts as well as like better error messages in general and the second one was
3:59:53a set of design engineering skills that made it easier to just have better animations and other
3:59:58things like that and then last but not least is the railway agent skills of course so like
4:00:02everything that i built for us internally at railway runs on railway of course like everything
4:00:09So like buckets, databases, the app itself, the real-time server, everything.
4:00:14And for the agent to effectively work with Railway and be able to deploy everything,
4:00:19you just give it the Railway agent skills and it was just able to do it.
4:00:23And yeah, so like once you have all of your agent skills and everything,
4:00:27in my opinion, the final step, and I think the most important step, is verification.
4:00:34So the agents typically, whenever you ask it to build something,
4:00:37it's going to be like, okay, I implemented the feature.
4:00:39build pass, test pass, all good, right?
4:00:42But what I found is when you actually try out what happens,
4:00:46you sometimes end up with just like a very broken experience.
4:00:51So once you have like a verification loop in place,
4:00:54it's like a nine day difference
4:00:55because the agent, you can literally see it,
4:00:58be like, okay, I built this thing.
4:01:00I tried it.
4:01:01It failed.
4:01:01I'm going to fix it.
4:01:03And it just keeps going.
4:01:04And then in the end, it's like, okay, I fully tested it.
4:01:06And I felt this breakthrough with computer use once it got really, really good.
4:01:11And now every time I want to add a new feature, I literally just in the end say, well, you should verify that the feature works end-to-end and use it by actually running infrastructure and to test things end-to-end.
4:01:26And the agents are very capable of doing this.
4:01:28You just have to tell them.
4:01:30and once you have all those pieces,
4:01:32I would say you will end up with solid internal tools
4:01:36that you can build quickly,
4:01:38that work properly and work in a consistent way.
4:01:40And that's it.
4:01:41This was my list of principles that I use
4:01:43whenever I'm building any internal tools.
4:01:44Hope you found them useful.
4:01:45But more importantly,
4:01:47that you're inspired to build your own internal tools
4:01:50that feel good, that are polished and spark joy.
4:01:54And I hope that you too can become your company's SaaS layer.
4:01:57And with that, thank you so much for watching, and I'll see you in the next one.
