​

​

​Podcast, YouTube, Blog​

Dear founder,

​

AI has an image problem among founders, particularly indie hackers and within the online communities that we frequent.

In the builder community, we are constantly inundated with shallow copies of the same kinds of tools. There are GPT wrappers, social media auto repliers, AI starter kits, boilerplates, AI search, AI coding tools, AI, AI, AI… — and frankly, half of the products on the Product Hunt front page have AI in their name, let alone the subtitles or descriptions.

Everybody is building AI tools.

​🎙️ Enjoy the blog post and the podcast

​

From within the indie hacker community, right now, this looks like a massive Gold Rush where everybody is trying to peddle shovels to the ambitious miners looking for AI gold.

It feels like there’s almost a Ponzi scheme-like quality to how tools for AI founders are being built by other AI founders who are a couple of months ahead. Courses are being released on AI topics by people who just barely scratched the surface in their own exploration of AI.

And from inside a community that is usually very protective of its members, this feels like it’s a problem.

So let’s talk about it.

Maybe it is an overblown hype that we perceive with the topic of Artificial Intelligence, but it feels like there’s something going on. It appears as an outsized interest in a niche topic. Everyone is talking about AI all the time — me included, right now. And consequentially, a lot of us software founders who are trying to build lifestyle businesses and bootstrapped businesses, are overwhelmed by all of it. That can’t be good.

The AI avalanche causes two potential reactions in founders: either, you’re one of the founders that say “Well, I don’t care. I’m going to build an AI product because that is currently what is selling. That is what consumers are looking for. That is what I see my peers build I want to be part of this wave.” On the other hand, you’ll see founders say, “Let’s go back to traditional software as a service businesses. Let’s ignore the AI thing and let’s step away from it” — a kind of a signal of contrarianism to actively pursue “not AI” as a response to the popularity of AI.

Both approaches feel a bit reductive.

I think we’re looking at all of this through a particular bubble. That’s unfortunate because obviously there is value out there in building with AI technology. Otherwise, it would not be as popular beyond the indie hacker community as it clearly is.

But is there value in AI out there that can be meaningful for us as entrepreneurs to inform what businesses we should build without falling prey to the hype without buying shovels?

Let’s start with becoming aware of our own biases. What kind of myopic lens do we have as indie hackers? Why are we so upset with the hype? What are we not seeing?

The first stumbling block is that we see AI as this hyper-modern trend. But AI didn’t start with ChatGPT. It got a lot of attention from that, clearly. When OpenAI started to release Large Language Models, we witnessed a technological breakthrough, both in the capacity and the availability of AI for founders, particularly solo founders.

So for many of us, AI is something that started somewhere in late 2022 or early 2023 with GPT3 or even GPT4.

But artificial intelligence is has been a research subject for decades before that, and many technologies existed long before LLMs were conceptualized. Neural networks, machine translation, prediction systems: there was a lot of cool tech around when LLMs eventually hit the scene.

I actually used a machine translation system called Moses back in 2017 for my previous software startup FeedbackPanda: we were creating an automated system to turn texts written from a male perspective (he did, his car, they gave him) to a female perspective (she did, her car, they gave her) and vice versa. We needed that to allow our customers to write one version and have the other one auto-created. Building this system took me several months, and training it to work well enough for the margin of error we were comfortable with took over 24 hours for every new experimental version.

GPT4 can do this in under a second now.

And I believe that this is both the amazing novelty and the reason we have such a skewed perspective on AI.

We forget that a lot of these technologies are silently powering a lot of businesses behind the scenes. Yet all we see when we look at the front page of Product Hunt are Ai-front-and-center B2C and B2B applications. Chat with your documents. Create copy within 50ms. Render a video right in the browser from a prompt.

We see AI as interface, as the first point of contact between us and the service. AI is flashy, and it’s right there. And that causes our indie hackers myopic lens.

When we think of how we should use AI as a product, we think of AI as the thing working right in front of us to create some kind of magic. And the more tools are released, where AI is such an outstanding part of the software, the more we consider that this is how we should be building tools in the AI space.

And meanwhile, there are systems out there that look like they were built in the 1990s that have powerful AI algorithms working on their backends, their databases, and reporting and data analysis data stores, which are surfacing massively impactful business information — and you can chat with none of those. No cool prompts. No on-the-fly image generation. Nothing in there yells “current bleeding edge AI systems”, but everything in there is an artificial intelligence crunching massive amounts of data.

I’m seeing this myself right now with Podscan. The actual magic of this tool is something completely invisible to my customers. It is the ability to extract data from audio that is meaningful and impactful for the businesses that are using my product. I scan every podcast episode out there, and I look for host names, guest names, sponsors, and advertisers. What are the main topics? Is the tone of this episode friendly? Or is it aggressive? Are people kind to each other?

This is what my customers need, and AI systems can give it to them — because AI works “reliably enough” at scale. It gives the people who use Podscan for their business meaningful information.

Their core question is not “Can I chat with this tool?” Instead, it’s “Is there a podcast that I want to advertise on? Or is this a podcast that I want to pitch my client on?” This is how they make money: by getting their clients onto good podcasts that fit their profile.

And to be able to do this job, they need to have information that is uncollectable by individual people, but can be easily collected and extracted by completely non-flashy AI systems. Yet nothing in my interface suggests this. It’s just a list of things. It’s a search bar where people can search for something and then results come up. There’s nothing different between the interface of Podscan and the tools of the last 20 years (other than it maybe looking a bit more modern).

When it comes to how do you use it, my AI is not an interface at all. However, it is powering data collection 24/7. I have multiple dozens of servers doing nothing but data extraction, data analysis, question answering on large text documents, storing these documents, and searching these documents, all in the background — completely Invisible to my users. The only visibility they have are the results that are AI collected or maybe even generated when it comes to summaries and such.

But it has nothing to do with creating content or automatically replying to social media posts like all those hyped-up tools we find in our social feeds.

I think our problem is that indie hackers just see the wrong tools that are getting popularized because they’re easily adopted by the likes of us and the peers around us. Monkey see, monkey do, and we end up building similar things that miss out on the deep but invisible power of AI that actually creates massive value that people are willing to pay for.

I’ve mentioned this several times over the last year: AI businesses are risky. Businesses merely leveraging AI are less risky. Still risky, though. There’s platform dependency when you’re relying on some OpenAI API or using any of their competitors. Local LLMs solve this problem, but they will cost you. And even then, this technology changes so fast that the flashier it gets, the quicker it also loses its appeal.

Look for the boring applications of AI. Make it a workhorse, not a dancing poodle. Pay attention to the founders who use this technology without mentioning it in every single sentence of their sales copy. When in doubt, instead of wondering “What AI business can I build?”, think “What boring business can I build that leverages invisible AI behind the scenes?”

---

I'll share a few updates about my SaaS on the pod, and I'd love to know what you think about them! Please leave a voice message at podline.fm/arvid 🥰

​

And if you want to track your brand mentions on podcasts, check out podscan.fm!

​

Classifieds

I recently launched The Bootstrapper's Bundle, which contains Zero to Sold, The Embedded Entrepreneur, and Find your Following. If you want to start a bootstrapped business and build a validated product and a personal platform while doing it, check out this bundle. It contains all eBooks, audiobooks, video courses and extra materials I ever created. It's just $50, for now.

​

​

Thank you for reading this week’s essay edition of The Bootstrapped Founder. Did you enjoy it? If so, please spread the word and ​share this issue on Twitter.​

If you want to reach tens of thousands of creators, makers, and dreamers, you can ​apply to sponsor ​an episode of this newsletter. Or just reply to this email!

​