Building Your Own Substack Repurposing App - 40 Pieces, 45 Seconds, $0.30 (Part 2: Case Study)
I was spending hours writing my Substack letters and barely repurposing any of it. So I built an app that turns each letter into 40 content pieces in under a minute - for $0.30 a run. Here's exactly..
If you’re just landing here, this is Part 2 of a two-part series on vibe coding your own business apps. Part 1 covers FlowSignal, my outreach tracker, and the one piece of advice I ignored to build it so it’s worth reading first.
![[023] Should you vibe code your own apps? (Part 1: Case Study)](https://substackcdn.com/image/fetch/$s_!Qinu!,w_140,h_140,c_fill,f_webp,q_auto:good,fl_progressive:steep,g_auto/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd7fa0005-804d-41b8-9ae3-14049b9fc3d2_2528x1682.png)
Once FlowSignal was running, I had a second operational Founder Block (im)patiently waiting.
Every week I spend real hours on Substack letters. Then the letter goes out and that’s more or less where it ends.
I had a full distribution strategy sitting in a document and I was doing almost none of it because turning one piece into many felt like starting over every time.
The Founder Block here wasn’t creativity. It was the conversion step between “letter published” and “content actually distributed.”
So the next question was obvious:
Could I build something that handled that step for me?
1/ Single Source Of Truth
I know there are different opinions about content repurposing and being present on multiple platforms.
Some of you repurpose a ton of content and publish on every possible channel daily.
Others swear by going all in on one platform.
I’m currently somewhere in between, and what got me thinking was my old mentor’s Ryan Levesque concept of becoming the Category of One.
Read more about it directly from him here:
Essentially, what I liked is the approach of having your “Single Source Of Truth” piece - which for me are weekly Substack letters.
And then having a distribution system (Ryan calls it Strategic Content Ecosystem) for all the other platforms.
Just to be clear (if you skipped Ryan’s article) - becoming the Category Of One is much more than just having a distribution system. But, distribution is essential part of sharing your unique angle and message wide.
2/ Introducing FlowOne
The idea was simple enough to say out loud: paste a Substack link or a raw draft, and get structured content chunks ready for every platform in my distribution system.
No reformatting by hand or starting from scratch.
What made it more complex was the feedback layer. Without the ability to rate, refine, and teach - I’d just be running fancy prompts in a loop. That’s already possible in your regular ChatGPT chats.
So the second question I built around was:
What if each piece of content could be rated, refined, and those learnings carried forward automatically?
It took about five hours to get the first working version live. A few more hours of testing to get it stable.
The biggest new challenge was connecting to the Claude API and making decisions about how sub-agents should actually work together - this was my first time building at that layer.
The setup is the same as FlowSignal:
Claude Code for the actual programming
Claude Chat for brainstorming the simplest viable approach
Claude API to give the sub-agents access to the model’s reasoning
Each repurposing run costs around €0.30 and takes 45 to 60 seconds. That’s less than a minute to generate roughly 40 content pieces.
3/ FlowOne - Draft Mode
This one exists for a specific moment in my workflow that used to create friction every time.
I finish a draft, and I’m not quite ready to hit publish.
There are a few things I still need before the letter goes out: section image ideas, SEO-friendly title and description, and a few Substack note teasers ready to go so I’m not writing those in a rush after publishing.
Draft Mode handles exactly that. Paste your raw draft, and in under a minute I have those elements ready to review. It’s not glamorous, but it removed a consistent drag from the publishing process.
(If building something that removes your own version of this friction sounds interesting so far - I’ve opened two extra slots over the next few weeks where I can help you discover your Founder Block and build a tool to solve it. More info at the bottom.)
4/ FlowOne - Repurpose Mode
This is the core of the app.
Once a letter is published, I paste the link and FlowOne pulls the content and runs it through the full distribution system. I can preview how each piece looks on the actual platform - where the text breaks and what needs tightening before it goes into my posting calendar.
If you Zoom in on my screen you can see my repurposing system that covers multiple channels and formats:
- Website blog
- LinkedIn newsletter (plus caption post)
- Carousel slides text (this one is big for me!)
- Instagram
- Facebook
- YouTube (video script, description and thumbnail prompts)
- Substack
- Reels / Shorts (captions and scripts)
…just to name a few (something has to stay behind the scenes and for client eyes only)
5/ FlowOne - Brain
The repurposing output impressed me immediately, but the part that actually matters for long-term use is what happens after.
Every piece of content has a feedback field. I can rate it, add notes, and tell the agent what to fix before I lock the letter as finished.
Once a letter is marked complete, every sub-agent records its learnings from that session: what worked, what didn’t, and what needed correction.
By FlowOne’s own estimate, it takes around 20 finished letters to reach approximately 95% of author-level output. I’m on letter four right now, and the difference from letter one is already visible.
The mechanism here is what separates this from a prompt-based workflow, because this compounds and it’s easy to track progress and navigate the learning process.
6/ What I’m not showing you
There’s a Stats section I’m keeping out of this letter - that one is for clients only.
I needed a way to see which platforms are improving the fastest and where the agents are still making consistent mistakes. It gives an immediate visual of the system getting smarter over time, or occasionally, going in the wrong direction.
There are also a couple of new features in testing: a content calendar and a content library - both aimed at making daily publishing easier to maintain.
The goal for all of it is the same: not to auto-publish, but to have everything ready and reviewed in one place so there’s always a human interaction before anything goes live.
That distinction matters to me because faster doesn’t have to mean automated.
7/ Should you build something like this?
FlowSignal and FlowOne are proving to been real day-to-day helpers and have already saved me a lot of time and nerves on daily basis.
Will I keep building more apps for my business just because I can? No.
I set out to find my two biggest Founder Blocks and see if custom builds could solve them - that’s done.
The more important phase now is actually using both tools, improving them through real use, and not letting the building become its own distraction.
Anything else at this point would just be a shinier version of the same avoidance.
That's why you always have to know exactly which Founder Block you're solving before you write a single line coding prompt.
Want your own Substack content repurposing app?
I’m not turning FlowOne into a SaaS product. But if your biggest Founder Block looks anything like the one I described in sections two and three, there are two ways I can help you build your own version:
Option 1: FlowBuild (1:1 session) A focused four-hour Zoom where we identify your one Founder Block and the tool that would actually solve it. I prepare a systematic brief for Claude Code, then you share your screen and build the first working version while I provide feedback and guidance throughout. By the end, you have something that works and a method you can keep using.
I’ve opened two extra slots at founding pricing over the next few weeks. DM me if you want to talk through whether it’s the right fit.
Option 2: Done For You (beta) If you’d rather skip the build process and have your own version deployed and ready, reach out and let me know what you’re working with. I may take on one done for you project (pricing starts at 3-5k).
-Filip “FlowOne” Sardi