39:59Sabrina Ramonov 🍄 · Interview
Mom of 3 makes $48k/month with Claude
A 40-minute interview with Sandy Lee — mom of three, regional manager, and someone who made $48,000 in one month using AI with zero technical background.
June 6thThe argument in one line.
A non-developer built a self-improving content factory using Claude Code that turns one source video into four platform-native formats while automatically learning what works through nightly Thompson Sampling on real engagement data.
Read if. Skip if.
READ IF YOU ARE…
- You are a solopreneur or consultant who wants to maintain a multi-platform social presence without manually managing each channel.
- You are an agency owner managing content for multiple clients and need isolated per-client analytics and brand configs.
- You are a non-developer who wants to understand what is genuinely buildable with Claude Code as your coding engine.
- You already use Blotato and want to see what custom automation is possible on top of it.
SKIP IF…
- You are looking for a turnkey product to sign up for today -- OptimusFlow is a custom build, not a public SaaS.
- You need broadcast-quality multilingual TTS -- ElevenLabs has noted gaps on non-English languages.
- You have no interest in faceless AI-generated video content as a format.
The full version, fast.
Till Oberhumer, a non-developer AI consultant, built a complete multi-platform video factory using Claude Code, n8n, and Blotato. A source YouTube URL enters the system, gets transcribed, and is transformed into brand-customized shorts for TikTok, Instagram, Facebook, and YouTube -- rendered and published automatically. The differentiator is a nightly self-improvement loop: Thompson Sampling analyzes engagement signals per platform, identifies patterns that lift saves and views, and injects those learnings as golden rules into future generation prompts. Cost tracking runs per-image, per-video, and per-platform, giving operators a clear cost-per-view number across channels. The system supports agency multi-tenancy with isolated client credentials, brand kits, and analytics.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:23
01 · Introduction
Sabrina introduces Till and teases the faceless video factory
00:23 – 01:44
02 · Case Studies and Real Results
NGO medical paper explainers, startup SEO presence via content volume; YouTube SEO underestimated by most
01:44 – 03:03
03 · Why He Built a Faceless Video Factory
The four problems: generic AI slop, platform duplication chaos, no feedback loops, agency complexity
03:03 – 05:57
04 · The Tech Stack
Blotato at the core for render/publish/animation; Claude Code to build it; Gemini for creative generation; Kling AI for images
05:57 – 09:08
05 · The Self-Improving Analytics Loop
Thompson Sampling runs nightly, promotes winning engagement signals to golden rules in prompts, injects random tests to maintain baseline
09:08 – 14:00
06 · Live Demo: Brand Setup and Onboarding
Wizard UI, AI brand kit import, complexity/humor/visual world sliders, Creator Cameo, guardrails (claims, competitors, disclaimers)
14:00 – 17:14
07 · Running the Content Pipeline
Paste YouTube URL, trigger n8n pipeline, review scenes and captions, rerender individual frames, publish via Blotato
17:14 – 20:01
08 · Analytics, Cost Tracking, and Platform Insights
Per-platform follower/engagement/CTR, content insight signals (keyword overlay, loop bridge, plain language), cost per view breakdown
20:01 – 22:42
09 · Thompson Sampling and Prompt Optimization
Golden rules vs. testing candidates, n8n pipeline walkthrough showing full self-hosted automation
22:42 – 23:30
10 · How to Work With Till
CTA to optimusflow.consulting and LinkedIn
Lines worth screenshotting.
- A non-developer built a functioning multi-platform content factory with Claude Code -- the barrier to building custom SaaS is now skill-in-prompting, not coding background.
- Thompson Sampling, the algorithm behind multi-armed bandit ad optimization, can run nightly on social analytics to auto-tune content generation prompts.
- Platform content expectations differ enough that TikTok, Instagram, Facebook, and YouTube need separate voice-overs, captions, and tone -- not just resized versions of the same video.
- A question in the hook works well for YouTube; plain language underperforms on TikTok; keyword overlay lifts views 10x on YouTube -- these signals come from your own data.
- Tracking cost-per-view per platform changes the content investment conversation from budget to ROI per distribution channel.
- The self-improving loop needs a random baseline injection to avoid crediting strong topics as strong formats -- without randomness, you lose the control group.
- A consistent AI Creator Cameo in faceless video functions as a brand recognition device that works across platforms.
- Blotato multi-account publishing solves a real pain: Buffer charges /month for 20 accounts.
- For agencies, isolated per-client brand configs and API keys are non-negotiable -- shared keys create cost-tracking chaos and conflate analytics across clients.
- Building with Claude Code as a non-developer is real -- the system has a UI, multi-tenancy, analytics, cost tracking, and a self-improving ML loop.
One source video, four platforms, smarter every night.
WHAT TO LEARN
The hardest part of multi-platform content is building the feedback loop that tells you which platform wants what, and encoding that learning into future outputs automatically.
- You do not need a coding background to build a functioning SaaS platform -- Claude Code handles implementation if you can specify what you want clearly enough.
- Building one video per topic and only varying the voice-over per platform is a viable cost optimization for solopreneurs; full scene-level customization is for paid clients.
- Thompson Sampling works as a content optimization engine because it balances exploitation with exploration -- the same reason it beats simple A/B tests in advertising.
- The random baseline injection in a self-improving loop is not optional: without it, you cannot tell whether a content format won or whether the topic was just hot that week.
- Cost-per-view per platform is the unit economics metric that makes multi-platform publishing legible -- without it, you are optimizing by feel, not by data.
- A Creator Cameo in faceless video creates brand recognition that persists across platforms without requiring the creator to appear on camera every time.
- Guardrails at the brand config level -- no claims, no competitor mentions, required disclaimers -- separate an amateur AI content setup from an agency-grade one.
- For agencies, isolating per-client API keys and analytics is non-negotiable: shared credentials make cost attribution impossible and mix performance signals across different audiences.
Terms worth knowing.
- Thompson Sampling
- A statistical algorithm for balancing exploration and exploitation in A/B-style tests. It allocates more attempts to better-performing options while still occasionally testing weaker ones, making it well-suited for continuous learning loops.
- Creator Cameo
- An AI-generated recurring character or face injected into faceless videos to provide brand consistency without requiring the creator to appear on camera. Configured once at brand setup and rendered into selected scenes automatically.
- Golden Rule
- In the OptimusFlow system, a content signal that Thompson Sampling has identified as statistically lifting performance and is now actively injected into future generation prompts.
- Platform-native format
- A version of content adapted specifically to one platform's expectations -- distinct voice-over, post copy, tone, and visual complexity -- rather than a repurposed crop of a master video.
- Multi-tenant agency layer
- An architecture where multiple client accounts exist under a single parent organization, each with isolated brand configs, API keys, and analytics so performance and cost data never bleed between clients.
Things they pointed at.
09:07toolNana Banana Edit 2
Lines you could clip.
04:04
“The whole thing I built with Claude Code. I'm not a developer. I have no background there.”
06:11
“Every night, the system checks all the videos which have been published, has a look at what the metrics are... and then promotes them into the prompts. So you basically have better videos over time.”
20:38
“It is like your own mini A/B testing lab for content.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
00:00Today, I'm super excited to interview Till Oberhumer, who is the creator of Optimus Flow Shirts. It is an insane faceless video factory that he builds completely custom, completely personalized with analytics and everything. He's an AI consultant based in Europe, and today he's gonna walk through his entire setup.
00:17Really great to have you here. Thank you so much, Sabrina. I'm really excited to be here.
00:20Any case studies or testimonials you can share about it? Uh, several foundations, for example, like NGOs.
00:26One of them, they are specialized RTCs. They needed a channel where they can share the information about all these different medical papers in an easy way.
00:37And in the past, they basically did it manually by reading the papers themselves and then trying to summarize them. And now we do these explanatory videos.
00:45Someone who is a patient or who has a relative who is affected by that can really easily watch that in a few seconds. It may massively proves them and it releases a lot of workforce because now they can really focus on something which which adds value directly.
01:02And I did the same thing for a startup, which was kind of like pushing into the market with their new product that they want to have, like, being present on a on a lot of channels. The idea was that if you Google them and you hear the name that you that they will be super present, which also works very well now if you Google my name, a lot of the YouTube videos are just coming up.
01:23You know, it's not even because I'm going viral or something, but it's just like there's a lot of content out there. Oh, yeah. People really underestimate like the SEO power in social contents,
01:33especially platforms like YouTube, Instagram, Facebook, even LinkedIn.
01:37They're like heavily indexed by search engines. So yeah, that's really cool sharing the two use cases. It's probably best to show the structure first and then go into the the nitty gritty.
01:48When I went into automation and then I also looked into social media, one of the things which I looked at was, like, different ways how to be present as a solopreneur, and then so I came across Protato. And I played around with it a lot. Still quite a manual process, but there's so much more to do with it than if you want to build out a presence online, which for sure was my intention, it was very helpful to kind of like make sure that you play around the different platforms.
02:14And these faceless videos, it really worked well to get across, uh, the message. What I was focusing first of doing the research and then having the brand script and then putting it on different, uh, platforms.
02:26My background is not social media or marketing, so it was also for me very important to have a self improving part in there, which are then basically built. So I touched face on the problem already. So it's, uh, it shouldn't look like, uh, generic.
02:38So I wanna have like a good quality on a video. There should be a platform duplication, but it should not be like the same.
02:45YouTube, TikTok, Instagram, Facebook, they're all different when it comes to the audience and what the expectations are. If I don't have feedback loops, I have to guess what is working. And if if it's if it's the agency, then you have the chaos because you have a number of brands, you have number of stacks, which you have to con contribute on.
03:03So I took one topic, I transformed it into four platform native videos, and I created a metrics driven loop, which basically helped me then to make it look like that. So that's the the the stack which I'm using. You can see potato is at its core because I I I use it for render.
03:19I use it for the publication. I even use it sometimes to do the the animations as well. It's just an amazing platform.
03:25Just the fact that you have the option to have so many accounts
03:29linked to so many different social media platforms at this price level. Yeah. I was just checking comparing it to buffers when doing some SEO, and it's, like, $80 Yeah.
03:38Per month if you want 20 accounts. And I really built this to like the solopreneur. Right?
03:44Be able to scale, be able to grow and not have to worry, right, about paying for this, all the infrastructure on doing this yourself. If you wanted to like post on TikTok, you have to get approval, and it's, like, such a huge headache.
03:58So, yeah, it's really cool to see what you've built on top of it. The whole thing I built with Cloud Code. So I'm I'm not a developer.
04:04I have no background there. For the most of it, I use Gemini actually because it's it's better on the creative parts, And I use KIE AI for the pictures because they are super cheap.
04:17I tried a lot monitoring the cost as well. They are at the moment at least unstoppable. So how does the end to end journey looks like for the user?
04:27So you have to the onboarding part where you basically build your brand world. You have the voice if you wanna have a Cameo. I wanted to achieve two things.
04:35I wanted to have social media presence linked to my topics. I wanted to my face to be known.
04:43I mean, you're working now on the voice as well so that that we even use my voice there. But at the moment, it's just me my face.
04:51I wanted to make sure that there's the research, so that's also something which I'm offering within the platform. You can basically say what kind of topics you're interested in, and then it will look into the trends and the transcripts.
05:03I'm using Protato here again. I have the triggers in there on the dashboard, so you can I will show this as well?
05:09On the background, I have an in a den pipeline where the whole process runs through. Protato is then again doing the rendering. I can review it.
05:16I can edit the caption. I can even rerender within the platform as well. And then publish publish again is going via potato.
05:23Works best. And then I have the analytics as well to understand the views and what was the cost. Brand customization is for sure important.
05:29So I have these visual words, like how should it look like. Now your platform evolved a little bit more, so you can do even more. You have the voice pics.
05:36For English speaking countries, it's perfect because 11 laps, the the standard voices are really great already. If you have different languages, like for me, it's German, it's a little bit harder because the pronunciation is not always on point.
05:49You have to do the Creator Cameo. You can upload your face. At the moment, I'm using Nana Banana Edit two, and then you have the guardrails.
05:56You want to make sure that you have brand safety. Here's the interesting part. It's the self improving loop.
06:00Basically, what it does, every night, the system checks all the videos which have been published, has a look on what the metrics are, the views, the saves, uh, the click through rate, this Thompson sampling, uh, in the background to understand, okay, what works well, what doesn't work well, identifies candidates of changes and so on, and then promotes them into the prompts.
06:20So you then basically have better videos over time. It does it randomly as well to make sure that you have the baseline so that just that doesn't go wrong. Because maybe you just picked an amazing topic several times in a row and it's actually not really linked to the way you did the video.
06:35It's just the topic which is thriving
06:37and to avoid issues there or mistakes there. I love that because that's one of the top requests I get from users. And also across all people on social media, it's like, how do we know how to get better?
06:50You have to collect all the analytics across the platforms. Maybe one type of post is doing well on Facebook, but it's not doing well on Instagram. Maybe carousels are doing well on Instagram, but they're they're not doing well at slideshows on TikTok.
07:02Right? It's it's really messy to consolidate all of that analytics so that you can learn and improve.
07:08And for a lot of people who aren't social media native, like this is a surprise to most people, but like I never used social media before I started posting content. Like I texted my younger cousin to teach me how to use TikTok and they just said no. They're like, it's just I see it in social media native now, but no, I just went off of data.
07:26It's like, this this these posts did really well, so we're going to try more of that next time. But imagine doing this for six different platforms while you're also trying to run a business. Like, it's it's very difficult.
07:38So really excited to get to that part of your platform. And have the data even seen by seen. With your own videos, you would have to run through this extra step,
07:46um, to kinda analyze, uh, how your scenes look like. The algorithm works the same.
07:51So it's it's insane actually what it what it does. I built it in a way that you actually can use it as an agency as well, so you can basically set up your different clients underneath. All of them should have a Protator account, then you also have just to take the decision as the agency if you want to share your API keys or if each client should have an individual as well.
08:12I support both. So the platform supports sharing API keys, but it also supports having individual setups. Same goes for sure for analytics and cost tracking because especially when I started the whole thing, it was kind of like an experiment to kind of, like, see how does it work.
08:29Will I have more clicks? Will I have more traffic on the website? But I also wanted to understand what the cost will be.
08:37So I actually track the cost on on each image which is created. I have the reality score, which I also have broken down per platform because otherwise it's a little bit unfair because sometimes TikTok is just outperforming everything else.
08:52And if you then compare, I don't know, Instagram with it, you basically kinda just throw away Instagram. But I also break it down in the per cost per view per platform, then I'll show you how it actually looks like in two steps.
09:08The first one is I will basically show it to you how it looks like if you're a new client. So if you're really going the whole thing the first time, and then I will actually show you how it looks like in the platform where I'm using it so you can actually see my numbers.
09:22And then I will also show you in the background how the the technical setup looks like. Start with this one. What it also does when you do it the first time, actually runs you through the whole thing.
09:31So it gives you a demo and explains everything and you have even the AI agents on the right side where you can ask questions in case you are confused on where to find something. Two things are really important here.
09:44So the one thing is like you basically do this you can do this quick wizard brand setup where you basically choose the niche which you are in, choose the platforms you want to publish on, choose the language which you're looking for, then you give it some samples to kinda, like, understand how your social purse post look like, what they sound like, and and and so on.
10:06And then you can have an analytics there. Or if you want to have a a deep dive, you basically can go in here.
10:14Because some clients have it. They have, like, this literally this brand booklets where they have everything in the their whole brand kit, and they just throw it in there. And then the AI will take everything out of it and basically adjust the whole thing.
10:25That's awesome. I love that. So you can just drop it in there, and then all of these other parameters will be customized based on your brand kit.
10:31Exactly. So it will immediately adjust it to what your style is based on your brand. You have, like, this easy settings.
10:39For example, this one, like, complexity is a very important one. So for example, for me, I found out that, uh, play language works best on Facebook. Or if if your brand is like, okay.
10:48Well, I'm a specialist in some point, so then actually my my target group is not the broad community. It's already someone who is maybe very technical himself.
10:57Then you probably go there and say, okay. Well, I'm actually going full blown specialist, and therefore, it will not use the easy to understand language, and it will also not try to explain things.
11:08Because at the moment for my videos, it basically would take a concept and turn it into something easy to digest. People are really like if they are really starting from the basics in AI that they're not getting confused.
11:19And it was also learning because, uh, initially, I had people when I was recycling videos, um, I was, uh, I got the feedback actually that nobody really understood what the what the video was about. Same here for the content task because, for example, I have one client. Um, they have medical studies, which they want to publish for their audience, but they want to have it simplified.
11:41So, basically, that would be then, um, what it does and would then, uh, preset the the voice settings accordingly. So you can see that it then adjusts here and says, okay, humor.
11:51You don't wanna be funny when you do this. And the same goes if you then say, okay, well, you wanna have it here.
11:58It will always ask you, and then it will also adjust here. You can see, for example, for academic academic papers, it obviously came to conclusion it should be only analytical, not funny at all.
12:09Same goes here for the the visual work. So worlds, I call it in the worlds.
12:15You can say modern tech, and it just adjust everything. You can exclude things. You can give it your private primary colors, secondary, then the accent colors, then, for example, band elements, which you don't wanna see.
12:28Because initially, for some of my videos, there was always some some tools where I didn't really understand why they are there. Then you have here the option to have this Cameo.
12:37If you wanna have it in there, you basically can give some information, describe the the So a cameo here is like a kind of like an AI character.
12:47So you can have you can give it a name. Yeah. Yeah.
12:49So you can either could you can basically say you have a mask work or something, which you always want to be in there. I saw some some creator. She has like a a little monkey, which is in all her videos.
13:00Yeah. So it's it's really cute, and it's it's kind of really, like, distinctive. So when you when I watch something and I see the monkey, I know it's from her.
13:07It's actually very smart. I haven't been that smart. I just use myself.
13:11So I basically said, okay. I want to be in one scene. So and usually, I'm in the opening scenes.
13:17And then you can also choose the the voice. Try to kind of have this random as well because I wanted to have a a look if the voice has also impact on how long people are watching the videos.
13:28Oh, that's so interesting. Yeah. And my algorithm also picks it up and also considers the voice that has been used for the videos.
13:36It's not perfect. To really break it down to the voice, I would have to do the same video with two different voices. Overall, it still works.
13:42You could basically gonna see a pattern. You can see that some voices adjust the form better. For clients, I allow them to to really like individual scenes and videos for each platform.
13:53At the moment, what I do, I do one video with all the scenes and only the voice over changes. So and the the text so the post as such is adjusted to the platform.
14:06Initially, I did it really, like, for four platforms, but it just got too expensive for for a simple experiment. So but for clients, I do it because, uh, for sure, they they need to have it really individualized.
14:17Then here, for sure, you have the guardrails. So it says, okay. Well, it's health, then you have to make sure that, uh, you don't have any claims in there, then maybe you don't wanna mention competitors in there, um, or you allow, uh, competitor mentions.
14:35Um, Yeah. Disclaimers, restrictions. So all these things that you need if you really wanna do it properly, and then you have, like, the the overall branding.
14:42So it can Mhmm. Logo. So and then you just press save.
14:45And then you have, like, the dashboard, and you have these old videos in there. So let me just go back because yesterday, I created one demo video, which you can see here now.
14:55Yeah. So you can see there's one for review. You can see there's one pending.
14:59It has been done for these four platforms here. Is it typical workflow for a client that they come into dashboard and they see videos already populated populated to be reviewed, or do they have to do something to create those videos? No.
15:11They would create them. So I go into your YouTube account. I just take your recent video, and what I'm going to do now is I just post it here, and then I go and run pipeline.
15:21And what it does in the background, it now, uh, starts the process on, uh, NNN. And you can see here then transcript, snippet, TikTok, and so on and forth. It will take a while.
15:30And, alternatively, what you can also do, can go into research input, paste your the text or whatever. You can even put a PDF in there. Sometimes the the scraping doesn't work, so the transcript is not really working.
15:41Then you can go in there and copy paste it and just paste it in there, uh, put it in a title, and then start the pipeline as well. It will also work. You can see as I did already the snippet.
15:50That's cool. Yeah. I like the progress bar.
15:52Ultimately, when it's when it's done, it will show up here. Uh, and if you don't have just a lot of videos, which I will then show you in my other account, um, then you go into this old video overview, and then you can go into the review here.
16:05So in this particular case, I'm very open, so I took this video from Nate. Once yours is done, it will also say that it's from you. I will say say your original title, what it was, and then what it what it made out of it.
16:18And then I can have here this review where I can see all the scenes. So I have the full explanation, what it was in there. So this basically took that sample YouTube video and then made it into a nice social media ready format, in this case, like, technical explainer type of shorts.
16:35Exactly. And it gives me the caption for each of them. For YouTube, for example, it also gives me the text, And then I'm actually able to add this tags automatically.
16:47After I posted it, I can then basically just push the tags also into the YouTube description, and I can also rerender. So if, for example, I think there is a mistake, there is some pictures which I don't like, rerender individual scenes, and then get it redone completely again on Plotato.
17:02Here, I always have the link to Plotato in the database.
17:08And here, it then will also show the the link to the post in which voice has been used. Now I would actually show you the the data.
17:17That's now my own account. And there you can just quickly see how the dashboard looks like, where you can see how many videos have been created, videos have been posted. Yeah.
17:28We'd love to see the analytics as well, like how you close the loop. So here we are with the analytics. You can see if you wanna have all of them, if you want to break it down to platform.
17:37It shows me how many followers I I gained. It shows me the engagement performance over time. It breaks it down to the individual platform as well.
17:45Here's actually the interesting part because that's the content insight. Okay. Well, if you go on TikTok and I want to see, okay, what works for views, then it basically understood negative signals here and says, okay.
17:58Plain language didn't really work. It faces in one. As already said, that's my face.
18:05So, obviously, that's not really popular on, uh, on TikTok. But it changes. You can see it really does look into sales, for example.
18:13There it says number and statistics actually doesn't really work. For engagement also that kinda like these are the indicators. They're not very super strong, but I can compare it, for example, with YouTube, there you have positive signals as well with 10 x here.
18:28So keyword overlay works well. Loop bridge works well, meaning that the the ends fits perfect into the beginning.
18:36Plain language, for whatever reason, doesn't really work so well. And, again, it changes also depending on what you want. So if a question in the hook works very well for YouTube.
18:47And then here, Facebook, for example, you can see there's even more. So it really looks into it. It runs every night, and it looks into all the data which is collected down here where I can see the individual views.
18:58I can see the engagement rate. I can see the cost per video. Yeah.
19:01That's one of my favorite parts, how you break it down cost per video, cost per view, etcetera. So they have the average cost.
19:08They have the the variable, then the fixed per month. And you can really adjust that because it really fits into what kind of platform you're using, and you might want to use more videos. Now, Higgsil, for example, is a new option because it has now this MCP support, which it didn't have in the past.
19:24A company that has a lot of budget, maybe they want to do really expensive avatars. Right? It's obviously gonna be a higher cost per video, but if that's what they want, and they could track everything.
19:33But let's say you're a company that's on budget, maybe you wanna stick to carousels
19:37because you don't have to generate AI images or any of the videos.
19:41Or maybe you wanna blend like your own assets and a couple AI images or video clips per each one to bring the cost down. I like how it's really clear, you know, what's everything costing you, what is the return. By the way, you can also do just carousels.
19:55You don't have to do videos. I have support in the background. I just didn't use it here.
19:59So I think from a techie perspective, before I show you the n n n one, I just wanted to show you the the prompts as well. So you can see that this bit is actually quite specific. Once it identified something as a candidate, here prompt rules.
20:13So for example, here it says it's a golden one, so it really started already to inject that. Words is not working anymore. It's declining, but it was a golden rule.
20:22And then he has a testing one. That's a candidate. So this one means it's already active.
20:27This one means it has been identified out of the random numbers of posts that actually there's a lift in there in the in the saves on what it's actually expecting.
20:38This is really cool. It's like your I mean, it's your own, like, mini AB testing lab for content. But to be able to do that, you have to have a quite significant
20:46prompt here where you have all these options to inject all this additional information into the the prompts,
20:53which the system generates for the clients. Yeah. That's what I wanted to ask.
20:57So, like, if I'm a new client, I don't have to set any of this up. Right? It comes out of the box with your system.
21:02Exactly. Exactly. It's strongly
21:05linked to what you have been putting in into into into the settings. Once you have done that, uh, it basically creates everything for you. Uh, and this prompt option is usually not even available because it's it's advanced.
21:16So for the normal clients, I don't give them the option to look into the prompts directly just for the the people who are always saying n n n is dead. That's still a very nice backbone.
21:27I mean, this is self hosted, but you can see basically what this whole automation looks like. It's getting triggered over here with the webhook.
21:36It runs through. I'm using, as I said, rotator here for the the scraping of the the transcripts.
21:45And then I have like this in between steps where it identifies here if there should be a blueprint in there, if there's something to be explained. It makes sure that the blueprint is visible. Then here, it decides which type of format should it be.
22:00Should it be explainer video? Should it be short or should it be a deep dive?
22:05And then it basically goes through this whole process here to then come back and render one video after the other once the the pictures are here. So I basically create the pictures here, and then I sent them over to Plotator.
22:22And just to be clear for folks watching this, like, you clients don't have to do anything
22:28related to this. They don't have to do any of that. That's all done and coming out of the box for them.
22:33Just plug it in into into their own end and end, or they can use our system in the background depending on what their what their preference is. For folks watching this video who wanna work with you, collaborate with you, or want help setting system up like this for themselves, how can they reach you?
22:48The best way is either to just go to our website. So wwoptimusflow.consulting or Google my name.
22:58You can reach me also on LinkedIn, which is my channel where I'm the most present.
23:04And, yeah, I really look forward to take on some inspiration projects, which is Okay. Be fun. Awesome.
23:09And we'll have all your links in the video description. So, guys, if you're watching this and you want to connect with Till, make sure you click the video description to connect with him there. Thanks so much for joining us.
23:18This was super awesome to see. So glad to see Plotato, uh, out in the wild, people building stuff on top of it. Super cool for me to see that.
23:25So thank you again for the time. Thank you for having me, Sabrina. Really, really enjoy
The Hook
The bait, then the rug-pull.
Till Oberhumer is not a developer. He has no coding background. And yet he built a complete multi-platform content factory -- brand setup, AI-generated videos, nightly self-improvement, per-platform cost tracking -- entirely using Claude Code, n8n, and Blotato.
Named ideas worth stealing.
03:03concept
The Four-Platform Duplication Rule
One source topic becomes four platform-native outputs -- each with adjusted voice-over, post copy, and tone. Not the same video resized.
Steal forany multi-platform content operation
05:57model
Thompson Sampling Content Loop
- Collect nightly engagement signals
- Run Thompson Sampling
- Promote winners to golden rules
- Inject random tests for baseline
- Feed learnings back into generation prompts
Nightly loop that auto-tunes content generation prompts based on real platform performance data.
Steal forany AI content system that generates at volume
09:08list
Brand Setup Dimensions
- Tone/complexity (plain language to specialist)
- Humor dial (none to high)
- Visual worlds (e.g., modern tech)
- Creator Cameo (recurring AI character)
- Guardrails (no claims, no competitor mentions, disclaimers)
The five config axes that determine how the system generates content for a given brand.
Steal forany AI content personalization system
How they asked for the click.
VERBAL ASK
22:42link
“The best way is either to just go to our website. So www.optimusflow.consulting.”
Soft handoff at end of demo -- natural, not pushy. Host also promises links in description.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
