// transcript — 165 segments
0:00 Introduction
0:02 There's a new movement happening in SAS and it's not AI. Founders are waking up.
0:07 We're realizing that the convenience of big tech has come at the cost of our
0:11 margins and our control. So, we're seeing a massive migration from the
0:16 cloud back to bare metal infrastructure and open-source software. [music] So,
0:20 instead of fighting over users in the AI space, there is now a big opportunity to
0:25 capitalize on this movement. And in this video, I want to cover five SAS ideas in
0:35 If you're new to this channel, welcome here. I'm Simon Horberg and I run a SAS
0:41 portfolio of five SAS tools. And if you're wondering why I'm here handing
0:45 out SAS ideas for free instead of just building them myself, it's because I am.
0:51 My SAS portfolio founder stack contains five SAS tools already. And because I
0:56 work with SAS every single day, I get a ton of ideas I believe could have
1:01 amazing potential. But obviously, I cannot build them all. So instead of
1:05 keeping them to myself, I'd much rather talk about them here. So maybe you can
1:09 take one of these ideas and [music] run with it. So without further ado, let's
1:12 Idea #1 - Open-Source Affiliate Tool
1:13 get to it. The first idea is an open-source affiliate marketing tool.
1:19 Now, this one actually comes from a very real personal pain point. We recently
1:23 launched the partner program for Founder Stack. And naturally, we needed a tool
1:27 to handle affiliate links, tracking, and payouts. So, I tried a bunch of the most
1:32 well-known tools on the market. And don't get me wrong, they're great tools,
1:36 but they all had one major requirement that had me think again. [music] In
1:39 order for these tools to work, you have to connect them to your Stripe account.
1:43 You have to give them read access to your revenue, your customer emails, and
1:47 your transaction history. [music] And honestly, I just wasn't comfortable with
1:51 that. sharing that level of sensitive financial data with a third party tool
1:56 just to track some clicks just didn't feel right. So instead, we decided to
2:00 vibe code our own solution. It took us about a week and we now have a fully
2:04 custom partner platform running on our own infrastructure. But it got me
2:08 realizing something. If there had been a solid open-source affiliate tool that I
2:12 could have self-hosted on my own server, I would have picked that in a heartbeat.
2:16 And I think this is a massive opportunity for 2026. [music] First of all, it solves the privacy
2:22 issue. You keep your Stripe keys and your customer data on your own database.
2:26 No third party ever gets access to any of your sensitive data. And secondly, it
2:31 solves transparency. A huge [music] issue with affiliate programs is trust.
2:36 Are partners actually getting the payouts they've earned? And is the
2:40 attribution logic fair? With an open source platform, this code would be
2:43 public. [music] Your partners can inspect the logic and confirm that you
2:47 are being honest with sales attribution. [music] That builds a level of trust
2:52 that a blackbox says simply can't match. [music] Now, if I were to build this as
2:56 a business today, I would structure the monetization in three [music] layers.
3:01 One, a free community edition, fully open source. Anyone can fork it and host
3:05 it. [music] This gets you distribution and makes it the standard. Two, a pro
3:10 lifetime [music] license. you sell a supporter edition that comes with
3:14 advanced features, maybe multi- teamam support or advanced analytics for a
3:19 onetime fee. And [music] three, and this is the big one, a full done for you
3:23 service. This would be the equivalent of a merchant of record, but specifically
3:26 [music] for affiliate payouts. You offer service where you handle the tax compliance and
3:32 the actual money transfer to the affiliates while a software runs on
3:36 their server. I really think the market is begging for a solution like this. The
3:39 Idea #2 - Agentic Q/A Testing Tool
3:44 second idea is an agentic QA tester. So, not long ago, I saw this tweet and
3:48 it sparked [music] quite a bit of debate. But I'll be completely honest
3:52 with you, I don't look at all the code the AI is generating anymore either.
3:56 [music] In fact, I stopped asking the AI to refactor or optimize for human
4:01 readability because I figured it's most likely an AI that will be reading and
4:04 maintaining that code in the future anyway. Instead, I simply do manual QA
4:10 on the solution that the AI built to see if everything still works. Though, I've
4:14 noticed one thing. Having an AI go over another AI's work, sometimes even
4:18 multiple times, drastically increases the quality of the code. So, why
4:23 wouldn't this principle apply to the QA part, too? Imagine a tool where instead
4:27 of writing all these comprehensive Cyprus or Playright tests that break
4:31 [music] every time you change a tiny little thing, you just give the AI a
4:36 mission. Go to the pricing page, select the pro plan, and make sure the checkout
4:41 flow works. This agent would spin up a headless browser, look at the screen
4:45 using computer vision, [music] and click through the application exactly like a
4:48 human would. It solves the biggest [music] pain in testing, which is
4:52 maintenance. You get the reliability of end-to-end testing without the nightmare
4:57 of constantly keeping your testing logic up to date. Now, how do you monetize
5:00 this? Well, this is the perfect candidate for a usage-based [music]
5:05 pricing model. You are selling a junior QA engineer. So, you charge per test run
5:10 or per agent minute or maybe just AI credits and then let the user top up.
5:14 You could also offer a higher tier enterprise plan where the agent runs
5:19 inside their own VPC. So unreleased products never touch your cloud. This is
5:23 a massive compliance requirement for big tech and some companies will pay a
5:28 premium for that privacy. Idea number three is what I call a shadow database
5:29 Idea #3 - Shadow Database Tool
5:34 tool. We all use these amazing no code tools to run our businesses. We use
5:38 tools for our project management, for our CRM, and for our internal vickies.
5:42 And the user experiences of these commercial products like Notion and Air
5:47 Table are incredible. They are miles ahead of most open- source alternatives
5:52 and certainly way ahead of anything we could ever build ourselves. But they all
5:57 come with one massive downside. They own your data. If one of these platforms
6:02 decide to double their pricing or if they decide to change their terms or
6:06 worst case, if they decide to ban your [music] account, you are completely
6:10 locked out. your business data is trapped inside their proprietary data
6:14 stores. [music] This is the classic vendor lock trap. So the idea here is to
6:19 build a data liberation engine. Think of it as a sync tool [music] that connects
6:23 to the APIs of these SAS platforms and pulls your data in real time into a
6:27 standard [music] raw Postgress database that you host on your server. Now you can still use the
6:34 SAS product exactly as you do today. You're not replacing it, but you are
6:39 creating a live readonly mirror of every single piece of data you create. It's
6:43 effectively a data insurance against vendor lock. If the SAS provider ever
6:47 becomes a problem or if they shut down, you don't just have a messy export dump.
6:51 You have a structured SQL database with all your customers tasks and notes ready
6:58 to be plugged into a new tool. This product would then have a bunch of
7:01 different connectors to other popular platforms. So [music] if you were to
7:06 move from notion to Monday, you would simply click a button and this tool
7:10 would normalize your data from notion and then import it into Monday
7:14 automatically. In terms of monetization, I would follow the open core model here.
7:19 The core sync engine is free and open source. This gets you the trust of the
7:22 developer community because they [music] can inspect the code and see that you
7:26 aren't stealing anyone's data. But then you monetize on the connectors. Maybe
7:30 the connectors for tools like notion or air table is free, but the connectors to
7:35 Salesforce, SAP [music] or Oracle, the enterprise connectors requires a paid
7:39 license. And of course, you can offer a managed cloud version where you host the
7:43 shadow database for them for a monthly subscription. That's a very proven, very
7:47 Idea #4 - Self-Hostable AI Image/Video Tool
7:49 scalable business model. SAS idea number four is a self-hosted AI media studio.
7:54 Right now, if you want to generate highquality AI images or videos, you
7:58 generally have two options. You can go to platforms like Midjourney or Runway.
8:03 These are completely closed ecosystems running their own proprietary models. Or
8:08 you can use aggregators like Open Art. Now, Open Art is great because it lets
8:12 you use and combine different models in one place. But their business model is
8:17 based on credits. Essentially, they are reselling you the compute, but with a
8:21 markup. You pay a subscription, you get a balance, and when it runs out, you pay
8:25 again. But if we look [music] at the textbased AI world, we have this
8:29 incredible tool called typing mic. If you haven't seen it, it's a beautiful UI
8:34 that let you chat with GBT5, cla or Gemini. But instead of paying a monthly
8:38 subscription for credits, you simply buy a one-time license for the software and
8:43 then you plug in your own API key. [music] This means you are paying wholesale
8:48 prices for the AI inference directly to the provider. No markup, no middleman.
8:51 [music] We need this exact same model, but for AI video and images. The idea is to
8:58 build a self-hostable UI that sits on top of AI pass providers like Replicate
9:03 or Fell AI. It would be a studio tool that let you chain together the most
9:07 popular models like [music] Flux or Nano Banana for images or cling or Pixieverse
9:11 or Sora for video. You wouldn't be selling credits. You wouldn't be selling
9:15 compute. You are simply selling the interface. The user brings their own API
9:20 key from replicate. They pay replicate directly for the generation. You just
9:25 provide an awesome UI workflow builder to combine these models. And in terms of
9:28 monetization, this is a perfect [music] candidate for the lifetime license
9:33 model. Just like typing mine, you sell the software for a onetime fee. For
9:38 instance, $79 or $99. Since you don't have any server cost because the user
9:41 connects directly to the AI provider, your recurring cost is effectively zero.
9:47 It's pure software play. You build it once, you update it occasionally, and
9:50 you sell the license forever. This is exactly the kind of low overhead, high
9:55 margin business that is perfect for a solo founder in 2026. The last idea is
9:58 Idea #5 - Self-Hostable YouTube Clone
10:01 unfortunately a bit more serious. One of my favorite YouTubers, Pat Walls from
10:06 Starter Story, recently had a terrible experience with YouTube. Out of nowhere,
10:10 his content started getting flagged [music] as dangerous. And he started
10:14 receiving strikes on his channel. And as you know, after three strikes, YouTube
10:18 doesn't just suspend you, they may delete your entire channel. Years of
10:23 work gone overnight. He talked about this on Twitter. And as it turns out, it
10:27 wasn't just him. This started happening to a bunch of business YouTubers
10:31 seemingly due to some AI moderation error. Now, fortunately, I haven't
10:36 experienced any of this myself, but seeing that happen to the starter story
10:39 channel made me incredibly uncomfortable, and it is a harsh reminder that we cannot really rely on
10:46 big tech for our businesses. We are building our houses on rented land. So,
10:51 the idea is to build a self-hostable YouTube clone. I'm talking about a
10:56 platform you can spin up on your own server that functions exactly like
11:00 [music] YouTube. It handles the video hosting, the streaming, the quality
11:04 adjustments, and it has a full UI for likes, [music] comments, and related
11:08 video suggestions, and so on. Now, to be clear, this obviously cannot replace
11:12 YouTube as a platform and the network. But it's a place where your viewers can
11:16 go and watch your content if the unthinkable happens. [music]
11:20 Place where you own the platform, you own the database of comments, and you
11:23 own the subscribe button. In terms of monetization, this is the perfect
11:28 creator economy tool. You could release the community edition for free, which
11:33 just handles video hosting, perfect for personal archives or small channels, but
11:37 then you could monetize by selling membership features as a paid license.
11:41 This would allow creators to turn their self-hosted platform into their own
11:45 version of Patreon or YouTube Premium. They can put their videos behind a payw
11:49 wall, [music] connect their own Stripe account, and keep 100% of the revenue
11:54 from their biggest fans. In 2026, creators are waking up to the platform
11:57 risk. [music] Selling them the ultimate backup plan is, in my opinion, a very powerful value
12:03 What's next
12:03 proposition. [music] Now, as you know, these are just ideas.
12:07 They are completely worthless if you don't execute. And if you feel the
12:11 analysis paralysis kicking in already, let me [music] help you out. Cuz in this
12:15 video, I'm showing you how to build a niche AI SAS with lovable and replicate
12:20 step by step. So take one of these ideas, go watch this video next, and