SaaS Factory: The Big Idea to Feature Pipeline
Plus some thoughts on the AI Hype Bubble that's building
Last week, I decided to set up a few automations for Build In Public University. I haven’t been leveraging the Twitter account I’d set up for it, so I wanted to start doing so a bit more.
Anatomy of an Automation
I’d deployed an N8N instance to render that I could use to build out my workflows, but hadn’t actually done anything with it.
There were a few easy wins that I wanted to implement:
When a new podcast episode is published, tweet a link
When a new YouTube video is published, tweet a link.
Those are almost identical workflows, but each actually requires two steps. First, I have to keep track of when a new podcast/video is published, and then I have to trigger a Tweet about it.
For part 1, I ended up setting up a couple of Airtable databases, one for podcasts and one for YouTube videos. Then I use the RSS feed for the podcast and the YouTube API to pull in all the episodes, and add a new row to the database if there has been a new episode published. This triggers hourly, although that’s probably way too often. But that’s fine, because it’s a self-hosted N8N instance, so I don’t need to worry about how many things I kick off.
The second part, then, is a trigger activated when a new row is added to those Airtable databases. It grabs the link and title from the table and tweets it out. Nice and easy.
All in all, took about 2-3 hours to get set up and running, most of which was setting up the Airtable and Twitter/YouTube OAuth accounts to allow me to authenticate my accounts.
This flow led me to start thinking about what else I could automate. Something I was thinking about quite a bit was how to automatically add pages to the blogs I run alongside the podcast, with a link to the episode, the description, and the transcript. I use Ghost as my primary publishing platform and N8N does support the Ghost API, so I started exploring how I could make that work. And as I did so, I found a blog post outlining something I thought was interesting.
It seemed to allow you to create an in-app audio player that supported the ability to include audio files that can play natively on Twitter, which seemed super exciting to me.
To me, this seemed like a huge idea. I’ve been exploring ways to help creators get their content viewed on Twitter, even though external links don’t get the same number of impressions as tweets without external links…
Turns out, that React component was not quite the panacea I thought it was though.
As you can see, no podcast episode. But I’m still pretty excited about the idea. Not going to be quite as easy as I hoped of course (probably why we don’t see anyone doing that, but I still think it’s possible.
So now, I’m planning on validating the idea in a series of steps.
How To Validate A New Idea
Step 1: How effective are video clips with just audio?
Create 15, 30, 45, and 60 second video clips featuring podcast snippets and post them in tweets at various times throughout various days of the week. I want to make sure that clips can outperform tweets with external links.
Step 2: Figure out how to generate video clips based on audio files.
This is the part I’m not 100% sure of, but I have done some programmatic video editing in the past, so I know it’s at least technically possible. Plan is to just use the podcast cover art as a static image while the audio plays.
Step 3: Profit
(Ok, maybe not profit, but start testing with other users.) If I can prove that there are ways to share clips easier on Twitter and that they perform better, that’s a good start.
Step 4: Enhance
I’ve got a few ideas for ways that users can automatically share the episode with people who like the video clip. I’ll need to test whether or not posting the link as a response to the video hurts the performance of the main tweet or not.
Hard Decisions
I’ve been wrestling with something for a few weeks now. Ever since I screwed up with Who Should I Unfollow?, I’ve been trying to figure out what to do. I’ve been in contact with a couple of people from Amazon about the locked account and unfortunately haven’t been able to get anywhere, but it looks like my account was completely deactivated, which means I’m not sure if I’ll be able to get my data out at all. So I’m leaning toward canceling all the active subscriptions and just making the app free. I might start using it as a lead magnet for the work I’m doing on Social Media Gardens, since that will be a higher-margin product. But I’ve got another call with someone on the AWS team, so I guess I’ll see where that goes before making a final decision on it.
Data Mining & Minecraft
This was something that popped into my head the other day so I started playing with the idea on Midjourney: what if we had to spend our days mining data in the data mines?
And that led me to another realization: I could actually share a lot of information about how the internet works in general by using Minecraft as a great example. So you might see some things come out from me in the next few weeks as I try it out to see if it is helpful to anyone else. It is kinda cool how it provides a new vocabulary that I can use to share certain ideas, and it should be decently effective even if someone doesn’t know a ton about Minecraft, because it’s easy enough to start out with very simple concepts and build on them. Let me know if that interests you at all, and I’ll share some ideas with you.
Closing Thoughts - Reflecting on the Role of AI
I’ve been spending a lot of time studying and reflecting over the past few weeks with a lot of focus on where the future of AI is going. There are a number of big changes in the air and I’ve been considering what they mean at a macro level. You’ve probably seen the hype around Lex, I just saw quite a few people trying out TweetHunter for the first time, and just in general, the AI space is pretty hyped up right now.
But what is the role of AI?
Most people seem to think it means that they need to replace everything they do with AI, just click a few buttons and send out their work.
That’s the completely wrong approach to take, because even as AI improves, it won’t replace human effort, it will just change the effort required to get to certain outputs.
I say this as someone who has become very familiar with AI in the past year working as an engineer at Copy.ai. I’ve seen how people can use it well and I’ve seen how people struggle with it.
It can seem magical at first, and you can use it to generate a ton of things very very quickly.
But ultimately, you need to have your thinking right first, otherwise, your results will not match your expectations. With Copy.ai, one of our biggest goals is to guide you to the right thing you need for your given stage of writing, and that’s not always an easy thing to do.
But when you have a specific goal in mind and a working approach that you’re familiar with, you can speed up that approach quite a bit.
So here’s what’s going to happen: there will be a lot of hype. People will be trying to replace everything with AI. That won’t work, the bubble will pop, and most people will go away angry that the AIs didn’t make their lives the magical unicorn rides they were hoping for.
But in the meantime, those who put real effort into understanding how to use these tools for specific purposes and integrated them with their specific workflows will be lightyears ahead, because they focused on the things that really matter when it comes to business/writing/etc.
So keep that in mind while you see the hype build over the coming weeks/months, because there’s a huge opportunity here if you can learn about what these tools actually do vs what the hype around them looks like.
And enjoy an AI model showing us what the magical unicorn ride could look like.