AI Hope

Seth Goldstein, host of The Entrepreneur’s Enigma podcast, had his now-is-gone moment about two years ago when ChatGPT-4 landed when he finally sat down to test it against something he didn’t think it could do. It provided a working answer where he expected a wall.

The clearest example he offered was Fedideck. Fedideck is a TweetDeck-style app he built for the Fediverse, the network of apps including Mastodon and Blue Sky that filled the gap Twitter left when it killed off TweetDeck. Seth vibe coded it in Lovable, got a usable version in about an hour, and then kept going. In total, he spent roughly $500 in AI credits and close to 200 hours of iteration on what he and thousands of others now use.

What Seth kept coming back to was the shift in posture required to use AI well. Typing into a chat window instead of building it yourself means learning to manage rather than execute. This requires explaining the task coherently enough that the work doesn’t wander off and waste tokens on the wrong thing.

Seth was blunt about the alternative: One shot everything and you get something that looks finished and works terribly.

Seth’s hope for AI is unglamorous by design. He wants it to help people work faster, better, and more completely without ever removing the human from the loop, and he has little patience for framing that treats the technology as an inevitability rather than a tool. The proof, for him, is his own day-to-day workflow.

Listen today on Spotify, iTunes, YouTube, or wherever you get your podcasts!

0:01:20 The “now is gone” moment
0:02:55 Building Fedideck in Lovable
0:05:20 Learning to manage AI, not just use it
0:07:22 Pushing back on AI hype
0:09:18 Cutting podcast editing to one-to-one
0:10:31 Where to find Seth

How to find Seth: https://social.sethgoldstein.me
The Entrepreneur’s Enigma podcast: https://entrepreneursenigma.com
Fedideck: https://fedideck.app

What is AI Hope?

What's your AI success story? What made you realize you had to change? And what does a hopeful future with AI look like to you? AI Hope with Geoff Livingston asks three simple questions of people doing the hard work of adapting and thriving in the AI era.

[00:00:00] Hey everybody, my name is Jeff Livingston, and this is AI Hope My book, Now Is Gone, is about never-ending change this podcast is about what comes after the stories the people and the reason for hope.

Hey everybody. Today's episode is with Seth Goldstein, the host of The Entrepreneur's Enigma podcast. He has his own incredible application that he's built with AI, and he's a social media OG. Enjoy. Seth is amazing, and I think you'll find it to be worth your time. Cheers.

I'm doing great, buddy. Good to see you again. We just saw each other, I think, last month- Uh-huh ... in person. 100%. Jeff does exist. Jeff is a r- Jeff is a real person. Yes, and I gave you my book. You're not AI. I, it, I am not Max Headroom. Yes, I have your book. You're not Max Headroom. It's pretty- Exactly. I have his book.

You do, and you wrote a review really quickly, which [00:01:00] is awesome. And, uh, we're gonna- I read it in a day. It's a quick read ... that's awesome. I mean, I really appreciate you doing that. It's very kind. I feel like most people these days- Always, buddy ... don't read very much, right? And so I think it's a big deal when somebody reads a book.

Thank you very much. Well, you're my buddy. Of course, I'm gonna read your book quickly and get a good review out there for people. Yeah, yeah. I, uh, I appreciate that. So but we're here to talk about you and your AI journey, 'cause part of that book is about that, uh, embracing AI and- Mm-hmm ... kind of going through that change that a lot of us have had to deal with, this algorithmic promise to redo all of our work.

And with that in mind, I wanted to ask you, uh, you know, have you, you had a moment in time where you've really kinda had to say, "Wow, I got, I better start using this," kinda that now is gone moment? Oh, yeah. The now is gone moment is that, like, now is gone. No pun intended. Like, the way you were doing it previously, like web design, y- you know, [00:02:00] newsletters, everything is different now.

It doesn't mean you remove the human out of the loop. It just means that you adjust and you pivot. Um, the, the moment for me was really a, a, I think two years ago- Mm ... or whenever ChatGPT-4 was out, and I was like, "What is this?" I mean, I kept on hearing about it. I've been listening to This Week in Tech and those podcasts, and everyone's...

Leo's, like, very big on it and all that stuff. And I'm digging into it and I'm like, "I don't get this conversational, like, talking to this thing to get an output." Right. So I dove in and I tried it, and I said, "Let me try this. Let me try asking it for something that I don't think he can do." Holy crap, it did it.

It's pretty awesome. I was like, "Whoa, that's a, that's a moment. That's a moment." Yeah, for sure, right? I mean, it's- Now, it doesn't do things perfectly. Yeah. Yeah. So, so yeah. I mean, it's obviously clearly flawed, but... and everybody's done a great job of talking about those flaws on LinkedIn. It seems like that's ground zero for complaining, right?

Uh, but, you know, in [00:03:00] reality- Absolutely ... a lot of, a lot of people have had some successes. I know you've had some successes. Is there, like, a moment where you could say- Yes ... "Hey, man, this has been incredibly powerful for me"? Oh, absolutely ... or a use case. Yeah. So- So I'm very big on the... Yeah, exactly. I'm very big on the Fediverse.

You know, Mastodon and Blue Sky and all that. And one thing that was missing was a TweetDeck-like interface for it. So I was like, "Let me vibe code it. Let me see if I can do it." And sure enough- Wow ... Fedi- Fedideck came out, and it's, it's really moving, shaking. That's awesome. Tell me a little bit about that, 'cause for those of us that are old school, I mean, when I say old school, you're probably not that old.

If you're a Millennial or a Gen X-er, you probably remember this. But when Twitter was really big, there was all these interfaces would allow you to interface into Twitter, maybe Instagram. There were all of these others, secondary social networks like Pownce and all these other little crazy places. And Hootsuite.

Hootsuite, yeah. Yeah. Hootsuite, right? It's kind of like Hootsuite. [00:04:00] And, uh, but then Twitter basically killed TweetDeck, um, and- Mm-hmm ... and there's been a sore need for this. So you're, you're seeing some market adoption for that. Is that correct? Oh, yeah. And even, even if I didn't, I wanted it. Right. And I don't really care if no one ever adopts it.

I hope they do, and they use it, and they tell me what's wrong with it and how I can adjust things and fix things. But I use it every day, and it helps my use cases for using it. Helps me with my outreach on Blue Sky and f- and Mastodon and all those Fediverse apps. It makes it usable for me. And if someone else gets value out of it, great.

I mean, yeah, that's... I probably spent 500 bucks- Wow ... building this sucker out. Which when you think about, when you think about that, 500 bucks in the scheme of developing a software, like old school- Is nothing ... is nothing. Nothing. But spending on AI, that's a little bit of money. I mean, it's, it was expensive for [00:05:00] AI to build it out, but I was like, but then you think about it, I got a basic use case in one shot.

It was not good in one shot. Right. I got it usable in one shot, very clunky, in about an hour. What platform? Now, uh, he was using Lovable- Yeah, I love Lovable ... Deck. Love that. Love that. Love Lovable, no pun intended. Yes. But, so using that, and probab- I've probably put- 200 hours into this app. Wow. I'll probably easily put 200 hours into this app, 'cause I'm constantly tweaking it and saying, "Hey, L- lovable, this is not working the way it should."

Right. "Can we move this? Can we adjust this?" And what I'm realizing with AI is that it's like working with a junior developer. So when you type in the chat window, instead of just doing it yourself, you have to learn how to be a manager. Yes. So the manager set, I think, is where people need to focus on, is realizing that the [00:06:00] AI is just the little rogue worker, hopefully not rogue, but, you know, the little worker be putting things, something together.

But you have to be able to explain things coherently, otherwise it's gonna go way off- Yeah ... and waste tokens. Very true. Very true. All right. Very cool. Yeah. So, uh, so that's pretty awesome. I think there's an entrepreneur story in that as well, that you build things that you would want, and it's probably likely that other people- Yeah

would want it, too. Um, what's your hopeful vision- That's the idea ... for, for AI in the not too distant future? I mean, we have a lot of conjecture with these IPOs coming. There's all thisagenic stuff with co-work, blah, blah, blah, blah. What does it mean? Like, what do you think as an entrepreneur and as kind of like a flagship carrier for entrepreneurs, what's your vision?

I think AI is gonna help people, if they use it correctly, help people work faster, better, and more completely. You cannot remove the human out of the loop, and I don't think you ever will be able to remove the human out of the [00:07:00] loop. You're gonna have to adjust where the human sits in the loop. Right. But, but these one-shotters are not gonna be a thing.

You're not gonna be like, "I could never one-shot Fedak." Yeah. No, there's no way in heck that could be one-shotted. Now, I one-shotted a piece of poop easily. Yeah, yeah, yeah. It was, it... Beginning part of it was great, but usability-wise, like I said, 200 hours into this project, and it's, and probably close to 500 hours in credits to Lovable.

Hmm. Hmm. So it's the future, but it's working with it. It's another tool. People are like, "Oh, my God, it's gonna take over this, take over that." And I really think Anthropic has really kicked themselves in the backside here by saying that, you know, the, the, "The end is near. The end is near. We're all gonna have to give over everything."

Oh, yeah. It's so stupid. Of course, then they get hit with something. Yeah. It isn't. It's so stupid, like- It isn't near. It's another tool. Well, the- People said this back when the internet first came about. This is just another rendition of the web. That's all it is. Yeah. I mean, I think the, the problem I have with it is that it's so [00:08:00] all-encompassing, the LLM technology in general- Yeah

that it's inevitably gonna have failures. It's the grocery store of AI, and literally, right? Yeah. Where it has way too much training knowledge to accomplish any individual little task. And as a result of that, you- Yeah ... you have something that's terrifically inefficient, not just from the energy and all, all the other perspectives, but with its answers.

It's off all the time. It, it constantly doesn't really know what you're talking about, because it's got too much information and it's working on- And it apologizes. It apologizes. "I'm so sorry. I'm so sorry." It kisses your ass. I actually told my cl- I told Claude to stop apologizing. Yeah. I just s- I said, "If you have to say something, say you're sorry and get to work."

Well- "Stop complaining." You know, like I, I would say to any junior employee, "Don't say sorry, just fix the damn problem." Right. Right. And the other thing too is that it constantly forgets its path with you, and that's 'cause they're resetting the model all the time because of degradation, right? Yeah. It's such a [00:09:00] challenge.

Mm-hmm. So there's just no way. It's not perfect, but neither was the web. Oh, no. It was never- The, the web wasn't perfect. It's great- It's growing ... when you take away that panacea label that they put around it, right? When you look at it for what it is, fantastic, really powerful. But when you make it out to be this, like- Absolutely

you know, Swiss Army knife that solves all things, it, it, it does it all poorly in that context. Yeah Yeah. Anyway- Absolutely, buddy ... but that being said, I agree with you. I love it within the workflows, and it can be really, really powerful when you automate tasks. In fact, my podcasting time, uh, for editing is...

I've automated literally three-quarters of it, just bam, gone. It's like an hour and a half every- Absolutely ... every time. People a- people a- people ask me how long does it, how long does it take me to edit, uh, Entrepreneur's Enigma podcast, and there's a plug right there. Um- ... it literally takes me 20 minutes.

Link your profile. Like, it... Exactly. Link your profile. But it literally is a one to one, one minute to one minute. I, literally I've [00:10:00] mapped it down to a minute to minute. Wow. That's incredible. Minute of podcast to a minute of editing. That's incredible. That's incredible. Anybody that's a podcaster knows that these things can be, or traditionally have been, very tedious to edit, and now it's become nothing.
But remember, I've also been doing this podcasting thing since 2010. I've been doing Entrepreneur's Enigma since 2021, so I have it down to a system as well. Not anyone's gonna be able to take AI and just get it one to one. You have your workflow. I have it down to a system. Yeah. And that's the reason why it's down one to one.

I have a workflow, you know? Exactly. I've done this before. It's not like you're gonna just start a podcast and it's gonna be one to one. No, you're gonna start a podcast, it's gonna be one minute to, like, 300 minutes. Just figuring out the damn program. Yes. You're an experienced producer. So on that happy note, Seth, where can people find you?

Just go over to s- social, that's sethgoldstein.me, and it's all my links are there. You can just- There you go ... you can stalk me f- [00:11:00] to high heaven there. Right. We'll put- Social.sethgoldstein.me. Sethgoldstein.me. We will put that in the links, the show notes, as well as The Entrepreneur's Enigma as a standalone link.

And with that, my friend, thank you so much for coming by AI Host.

Thank you for having me