The Bootstrapped Founder

Hey. It's Arvid, and this is the Bootstrap founder. Earlier this week, I read a blog post by Bethany Crystal, and it was really, really interesting. It was about how CHAN GPT saved her life. She had developed some kind of symptoms that she didn't really know, she couldn't place, and had blood work done couple days before.

The results were available to her, but her physician hadn't looked into them yet. And she could spot that on the website apparently where she got those results. So she brainstormed with Shad GPT about what the outstanding and non regular values on the data might mean. She took photos of the bruises that she developed out of nowhere and the little red spots and whatever and then the AI diagnosed her to the point where it recommended that she go to an emergency room right away. She first thought, should I really?

But then it actually became bolder, it wrote in bold text. And when she went to the AR and she got there medical staff told her that she was there just in time. They took the same measurements again and he was like, wow, this is really crazy. We've never seen it this bad before. She's now on the path to recovery, but Chachapiti made her go.

And there's something amazing about this story on so many many levels. It illustrates to me how everybody now has access to this health care advisor or health care consultant that is constantly available with knowledge of pretty much every illness out there and how to potentially diagnose it. And these AI systems have read countless accounts of illnesses in so many variations and they likely encountered way more variants than any individual physician could possibly see in their career. So when I go to my doctor they've seen maybe 5,100 cases versions of a certain thing. AI systems have ingested papers and abstracts and data for millions of such cases, right?

They're programmed with guardrails though and that's a benefit I think at this point because instead of trying to over diagnose like Bethany's problem, Chechi Petit just told her to go to the ER right now. So they prevent over diagnosing and then quickly defer to actual experts in the field, to emergency room physicians and first responders, whoever you might need. And that's amazing. I just really enjoyed this story for how crazy it is that half a decade ago this was unthinkable that you had this advisor to any potential problem in your hand or in your pocket. And at the same time, we're using these exact same AI systems that help people survive, to write code for our software products, or to draft cheeky marketing emails or conduct research for us.

It's crazy to think I'm just amazed by this whenever I read an account like that to think that all of this is possible made with the same technology by the same thing. It's Incredible. And with the reasoning models that are now available the most recent o3 model by OpenAI and Claude's very very recent like two days ago recent Claude SONNET 3.7. These things can now pull information from a vast vast context of documents that you give them and documents that they found somewhere else and make use of the fact that they can remember almost anything they consume. Obviously, don't really talk about hallucinations at this point and illegally obtained materials.

That's a topic for a different day, but these systems have the data available and it also works for not so critical things. I recently had my own interesting AI experience here and this might come off as a little bit nerdy, but bear with me. It's kind of cool. So I took a photo of my collection of unpainted Warhammer miniatures. I have plenty of those.

The boxes in which the miniatures still unassembled like on the plastic sprue are just waiting to be built and painted and it was sitting on the shelf and I thought, I have all of these things. What can I make of it? So I asked the AI to set up a list of the order and configuration in which I should build these little models so I could use them to actually play games. Warhammer has a very extensive rule set so you need to really figure this out and I didn't want to do that. So I just took a photo of the boxes and gave it to Claude and just from this photo of merchandise containing miniatures that the AI had never seen the inside of.

Right? They didn't know what was on the sprue. It accurately created a veritable and very in-depth list for my little miniature army that I can now build paint and know exactly how to use in the game. And just a couple minutes after I did this and was equally amazed by how this system that is just built in a computer somewhere knows so much about Warhammer, I gave the same AI a significant portion of my code base and asked it to refactor a certain part so that the outcome formatted file format would be a better fit for a customer that had certain needs. And this same exact AI that just then diagnosed somebody's urgent need for emergency care and built an army list for my hobby was then able to build a component that fits seamlessly into my existing software product.

It's just mind blowing that there's so much variety in this one system and then there's variety between systems. You could ask the same question of different AI's and they would maybe come to a consensus, but they would give you slightly different perspectives. And to a smaller extent, but still very similarly, that variety exists in locally installable large language models as well. If you were to install LAMA 3.1 locally, the 8,000,000,000 parameter version or maybe a 3.2 version with a smaller text based model that's still as good. They could still create those results.

Maybe not as extensively or reasoned as the modern GPD o three or the very recently released Cloud Standard 3.7, but these versions all just get better. Every newer version is just a little better than the old one, but the old ones are still pretty good. And even the small language models are extremely good compared to what we had before, which was nothing, which was having to do it ourselves. So again, another thing that blows my mind is just the massive availability of this technology, both locally installable and on the API. I occasionally take screenshots of my interface too and the source code that kind of goes with it, that creates that interface, and I ask what's wrong with this?

Because something just doesn't work. Like what is wrong with the code? I want it to look differently and then I explain it. Then magically the image detection figures out how the visual components pictured in the image map onto the code that I provided and then fixes that code so that the image looks more like what I want it to look like, all within milliseconds. And you can do this locally too, if you use Ollama, o l l a m a, and install a sizable or a small model.

It all works. It's incredible. I believe that if you don't yet use AI as a founder or as a hobbyist or just as a human being on any level, you're actually missing out on something very meaningful. And I don't want to do FOMO stuff here, but you have access to an expert in all things all the time by using AI systems. That is novel and to make this more about the entrepreneurial side of things as we are in the Bootstrap Founder podcast, the mindset of being able to prompt any AI system to fulfill almost any task and to go AI first in brainstorming and figuring things out and planning and thinking that is something that we can map onto our business ventures.

And I believe we must map this mindset shift onto our business ventures. So instead of collecting information like going out there and doing the research or coming up with strategies from what we already know. I have found it to be way more effective and the results of it be way more impactful if I approach a conversation with AI first and I act as if that AI were more experienced than I was in the field as an expert. I use AI systems like this to figure out if my theories about what the next strategy should be or which tactic I can now use right now for any given purpose are valid or invalid. I trust that the collective knowledge that is instilled in any AI response because the system knows so much because that's what it draws from is wiser than whatever individual knowledge I might have to this point.

I know that I know nothing or very little. That's really what this is and I open up the realm of possibility to have conversations with somebody who may not be real but knows way more than I do. So when I have a challenging problem in marketing or I want to reach a particular customer segment because I see a new and improved benefit for them using my product instead of trying to figure it out on my own, I open up a conversation with an AI system and I use different AI's for this purpose just to see how responses vary. But with Claude in particular, because that's what I've been using and will keep using, particularly now with a new reasoning model. Very interesting.

I go into conversation mode. And here's exactly how I do this. I often record just a solo brainstorm that I do all by myself into a microphone just like this. And I talk about something for five to ten minutes, anything. And I put everything I know, everything I assume about the thing, about what the consequences might be, what the underlying reasons are, and everything that I'm not quite sure about, what I wonder about, and where I have conflicting views into just this conversation with myself.

And then I take the full transcript and I prompt Claude to start a back and forth exchange with me, a conversation about this topic, and then brainstorming strategies trying to invalidate any theories I might have. I put the full transcript of the solo brainstorming in there, and that's when I hit enter, and that's when the conversation starts. And usually, particularly if I instruct Claude to respond in a more structured way, but giving examples for ideas or giving me blog post URLs to do more research with or sharing narrative examples that it remembers from having likely illegally scoured the whole Internet for these kind of things for companies that have used the strategy to great effect or companies that struggled with it. I get a conversation out of that that I feel I could not have had with any other human being for two reasons. First, it would have been very expensive to find an expert in all these things and pay them for their time.

Whereas, Claude is effectively free. It might cost me 20 some bucks a month but that's just domino for a subscription that wouldn't even get me a minute with the expert that I would need to talk to to get the same results. And second, more importantly and more on topic here, it would have taken me weeks to set up. It would have taken days to find out that expert and find a date and then figuring out how to reach that person and navigate their availability? But with an AI system, it's immediately available within milliseconds to start this conversation and it remembers everything forever.

It's all on my own terms. Nobody else is involved in making it a priority. I get to do this when and how I want. That's very alluring. And for a solo founder, it's really needed.

Previously, I would have had to find the expert and do all the research myself while still running, building, operating my business, and that just takes hours for the initial research to just figure out what a potential strategy might be to be interested in for the next marketing push. Take a recent example here. We introduced the ability for podcasters to claim their own podcast on PodScan just this week and this feature came out of exactly such a conversation with an AI system. I've been telling it about my last couple months of marketing activities, opening up PodScan's data more to the public more SEO available and the several feature requests where people talk to me about wanting to update information about their podcast and how they found it through Google and want to kind of fix something. I would get messages like hey this isn't fully correct the publisher here should have this letter in the name or I recently changed this and your system hasn't picked it up yet or my podcast does not accept guests why does it say it does?

That kind of stuff, right? So I quickly update the data whenever this happens and that's just maybe once or twice a day because it is pretty accurate, but when it happens I do this. But in the conversation with my marketing project on Claude, my kind of virtual chief marketing officer, it came up that maybe we should allow these people to just do it themselves. We could let them prove that they actually do own the podcast and then give them limited but impactful management control over all of this data like editing the name if there's a weird letter and coding issue or updating the description formats publishing frequencies and they could also update the size of their audience was a suggestion, which would be very beneficial. And the moment Claude mentioned this, I was like, yeah, this is amazing.

This is not just good because it helps people be more accurate, but it would really help the system, the platform. Because by now, our machine learning system that figures out audience sizes from all kinds of data on a podcast, that's been very good and reliable, but it's not perfect. Sometimes people would say, hey, I have way more people in my audience than this or I have much fewer than this. Don't say this is accurate because people will hold me accountable. So now they can call in the podcast, put in the real number and show proof that they have it to me so we can actually verify it.

So they are happy because accurate information is properly reflected on the PodScan platform, which other people get to see. And I am happy because now I have customer verified audience numbers, something that is extremely hard to get in this industry, that I can then use to train my machine learning model to be even more accurate. So as customers stream in claiming their podcasts and they do, I get more accurate data which makes it easier for people to see PodScan as an interesting and reliable product and data source. So all of those benefits, which is kinda mind blowing again, I'm sorry, I'm using this inflationary, but it is, are a consequence of a conversation about what the next step should be in making the product more visible to its target audience with Claude. Is that crazy?

Like, that conversation maybe cost me effectively 1.5 to $2, and it was a conversation with an expert that knows everything about marketing, about product development, software product adoption, podcasting. And that same expert helped me build my Warhammer army list and helped Bethany Krystl get to the ER in time to deal with an illness that her doctor hadn't even looked at the blood results for. That is incredible. If you take anything away from this, it's the concept that at this point, it is much cheaper and in aggregate more effective to think with AI and do the work with people than to do the work with AI and think with people. That's something that I've kinda realized.

This brainstorming, this planning and back and forth between me and AI, that's really just limited to how fast I can type because AI responds so quickly that I'll never be able to absorb it at the speed that it's responding. If I have a thought and I express it, AI instantly responds. The moment I send it in, it already starts responding because that's how these large language models work. They respond one token at a time and they respond very quickly. And the speed of that conversation is really just kept by my capacity to keep it up.

The work, the actual implementation, that is still very complicated and it's context dependent, which is something that AI systems struggle with, people struggle with less. Particularly when it comes to code implementations and larger code bases, AIs are still suboptimal. They're great. They do a lot of work faster than people do and they might do it more reliably than some developers but that still often needs hand holding correction and investigation particularly in large code bases. Smaller ones usually fine but the moment there's complex interdependency it becomes problematic.

So the work part here is still better done by humans who use AI as a tool, not by AI agents working completely independently. We'll get there. And I feel we might even get there this year still, which is gonna be weird. But for now, the thinking part, the brainstorming, idea generation, verification, validation, invalidation, that stuff, anything that has to do with getting an outside perspective, I will always go AI first. Because it's cheaper, faster, gives me the whole perspective, or gives me many different perspectives if I instruct the AI accordingly.

And after that, I talk to my team. And that's how the claim your podcast feature actually happened. I talked to the AI, set it up, built prototypes, experiments, and then I had a conversation with my UX designer Nick who wears many other hats too. And after the whole thing was strategized, we discussed how to monetize it. Do we monetize it at all?

Do we allow access to podcast claims for free? In perpetuity as a freemium part of the business, We had, like, a people conversation around things that we need to be fine with. And that operational conversation involving multiple moving parts was what I had to have with a human. But the first level was all between me and the AI. And the result is that podcast claiming is now a freemium part of the product of PodScan.

It makes so much sense from a business perspective to allow anybody who has a podcast to maintain its data for more data fidelity and accuracy on the platform without having to to pay for anything. It's their data. They will only have access to their podcast, obviously. If they start a trial, they get access to everything, every feature within PodScan. But when the trial runs out, all they can do is edit their podcast.

And if they wanna do more, they can subscribe. If they want alerts for their podcast, I think that would be an interesting way to upsell the product. We can remind them that it's a possibility. We could even track alerts for them, so they could see what they're missing by not subscribing. All of these potentials exist outside of the freemium plan.

Because we now have podcasters on the platform, we can also present them with information about competing podcasts or potential guests that they might want to invite. There's value in having podcasters on the platform, not just marketers and researchers and journalists, which are my main audience and have been so far, or builders who need podcast data for their own products. So it's a freemium part of the product. And if you have a podcast, it's likely already listed on PodScan because we are at, like, 3,200,000 podcasts, which is every single one. So feel free to claim it and check it out.

The feature came up because of AI and was implemented through AI. I was clawed with the results of the brainstorming and the existing data models around the podcasts and users and tasks to build the back end logic and front end for claiming podcasts by checking RSS feeds for tokens and having a background checking system. All of this was caught. So everything about this feature from inception to implementation I mean, implementation was still me pulling it all together, hence the thing with AI built with people. But everything around it was AI conceptualized and then with the assistance of an AI system implemented by me.

PodScan is an AI business on many sides. It uses AI heavily in its internal operations to extract data, figure things out, summarize, but it's also built on through and with AI. So any business that doesn't look to AI first at this point is wasting money and time. Because AI is going to be there immediately for you and be good enough for the first steps. Maybe even good enough for the whole thing.

If you don't use this, you're actually hamstringing your own operations. And that's it for today. Thank you so much for listening to the Bootstrap founder. You can find me on Twitter at AvidKahl. If you want to support me on this show, please share PodScan.fm with your professional peers and those who you think will benefit from tracking mentions of their brands, businesses, and names on podcasts out there.

PodScan is a near real time podcast database with a stellar API, so please share the word with those who need to stay on top of the podcast ecosystem. Thank you so much for listening. Have a wonderful day, and bye bye.