With Open Cards: Building Solo AI in Public

I joined the Twitter #buildinpublic community a little over a month ago when I started building RecapIt with Langchain. Shortly after the release of that small app (it summarised YouTube videos and let you chat with the content) I started working on the concept for Solo AI. In this blog post I want to tell everyone what I’ve learned so far, how I’ve utilised AI and other products from indiehackers and how the product development is going - being completely transparent with everything, including signup numbers.

The Idea

The idea for Solo AI is simple: Quickly get solo business owners to improve their workflow with AI, without spending days studying or reading a newsletter everyday. Instead of reading through 20 Twitter threads that each mention the “100 best prompts”, Solo AI will just simply walk you through an AI workflow with a clear goal.

Building a Landing Page

The first step in building Solo AI was creating a landing page. While I have extensive software development experience, I don’t really know much about design and frontend web development. After a quick conversation on Twitter with some people, I decided that I wanted to actually code my landing page (with the intent of easily launching other pages in the future by just changing the text). I decided on using Tailwind based on another recommendation on Twitter.

For design, I made my life easy and just decided to buy the excellent Indie Design kit from a Twitter connection. And then I started coding - with the help of a Youtube tutorial and ChatGPT. ChatGPT specifically helped me out quite a bit in at least 2 tricky situations: With my little CSS and React knowledge, finding out what I was doing wrong would have probably taken me hours or days. But with ChatGPT I had the issues resolved in under 10 minutes!

Due to my little React knowledge and my general desire to keep things simple, my waiting list is simply a popup of a Tally form. This “sign up” button will later be simply replaced with a button to the checkout page on Gumroad - no business logic coding needed!

Lastly, I needed a logo for Solo AI. Naturally, I’m using AI to generate it. I used the excellent tool AppLogoCreator from another indiehacker on Twitter and with a little help from the founder I eventually landed on the logo that the landing page currently uses (within maybe 3 hours of playing around with the tool).

Getting Feedback on yesRamen

Next step: Getting feedback to see what people think about the idea. The webapp yesRamen from yet another indiehacker is perfect for this! I gave feedback to a few other projects and then submitted Solo AI - within a short period of time I got very valuable feedback! The first set of feedback generally wasn’t so great - in part because the landing page copy wasn’t that great and too unclear. I took the feedback, restructured the sections and copy a bit and the feedback improved dramatically!

Setting Up ProductHunt

If you’re launching a small, bootstrapped project, listing it on ProductHunt is nearly required nowadays. To be frank, I wasn’t looking forward to this much - while it’s definitely helping projects succeed, I’ve heard from several people that it requires a lot of work. And I wasn’t sure if I was ready to already put in the work at this stage of the product. But thankfully you can “pre-launch” on ProductHunt - I setup a “notify me about this launch” page within a few minutes, which I can slowly use to spread awareness. I also listed the launch on hunted.space. But the real work for the ProductHunt launch will start roughly a week before the actual launch of Solo AI.

Building the Product

I’ve started building out the content for Solo AI a week or so ago after getting some initial validation. To be clear, it’s not a lot of validation - my Twitter audience is still quite small in size - but it is enough for me to commit to launching the product. In part this is also related to the fact that Solo AI is roughly “just” a month of work and I want to ship a product to go from 0 to 1 in revenue.

If you are a builder with a small audience and wondering what sort of validation others have, here’s a quick fully transparent rundown of the numbers:

• I have around ~130 Twitter followers. Twitter is the only place on which I’ve done any sort of “marketing” for the product (and I’ve kept it pretty light).
• Solo AI got 7 detailed feedbacks on yesRamen, with the later one being generally positive.
• 13 people clicked the “Notify me” button on the ProductHunt teaser page.
• 5 people signed up to the waitlist on the SoloAI homepage.

So far, I’ve finished the list of workflows that I want to have on launch and much of the intro page. Plus, I’m done with 3.5 out of 12 workflows. Writing the content has mostly been a mix of writing down what I know from experience and doing some additional research (eg when I wanted to try a different AI tool).

The todo list is still quite long considering that I want to launch Solo AI on the 1st of July:

• Finish all remaining workflows
• Write up section 1 - the introduction to ChatGPT and Midjourney
• Code the AI tutor (based in part on the work I’ve done for RecapIt)
• Record 3 “intro” videos that outline the tutorial

Plus, all the marketing and launch activities that I have to do - primarily making sure that the ProductHunt launch goes smoothly. But I’m positive that I can finish everything in time.

Trying out Twitter Ads

I’ve also tried out launching Twitter ads - it went absolutely terrible. For my experiment, I used only 10$ and created 2 ads with ChatGPT in a few minutes without thinking much or giving ChatGPT a lot of context. Plus, I also didn’t do a good job at setting the audience correctly. So while the ads got 4.9k impressions and an engagement rate of 9.56%, I don’t think I gained any signups from the ads. Interestingly enough, around 50% of the engagement has been in the form of “profile clicks”, which feels a bit strange considering that I chose “clicking the link” as primary objective for the ads (the other ~50% was detail expands).

Next Steps

Besides working on the content for the product, I’ll also spend a few dollars again on another Twitter ad. This time I’ll do some more research before setting it up though. In addition, I might try out a Facebook ad since I’ve seen other people have more success with it (compared to Twitter ads) and I think the product is more appealing to people who don’t know much about AI yet (which excludes a chunk of Twitter).

And on a final note: Last week I wrote about using personalised AI to fix issues that exist in blockchain DAO governance. Most likely this project will evolve as part of hackathon in the coming weeks. Stay tuned!