Vibecode Showdown

[00:00:00] Welcome to the Newcomer Podcast. I'm Eric Newcomer. And I'm Madeline Renbarger. Each week, Eric and I discuss the VC deals and the drama that went down. Let's do it. Here we go. Ooh, a loyal supporter of the Newcomer Podcast. I think our first advertiser and Christina, the CEO, has been on the show.

[00:00:27] Anyway, without further ado, you're a startup founder. Finding product market fit is probably your number one priority. But to land bigger customers, you also need security compliance. And obtaining your SOC 2 or ISO 27001 certification can open those big doors. But they take time and energy, pulling you away from building and shipping. That's where Vanta comes in. Vanta is the all-in-one compliance solution, helping startups like yours get audit ready

[00:00:55] and build a strong security foundation quickly and painlessly. How? Vanta automates the manual security tasks that slow you down, helping you streamline your audit. And the platform connects you with trusted experts to build your program, auditors to get you through audits quickly, and a marketplace for essentials like pen testing. So whether you're closing your first deal or gearing up for growth, Vanta makes compliance easy.

[00:01:22] Join over 9,000 companies, including many Y Combinator and Techstar startups who trust Vanta. Simplify compliance and get $1,000 off at vanta.com slash newcomer. That's V-A-N-T-A dot com slash newcomer for $1,000 off. Eric, how is your vibe coding going? Well, first of all, Guillermo, the CEO of Vercel, having lunch with him at HumanX,

[00:01:50] it's like, oh, you could build everything in VO, their coders. So I started there. Honestly, I got stuck in VO. What was the problem? It was like some button, you know, it was just like make the button work, you know, and it's like, I was like, start over. I don't know. I just like got stuck. I moved to Lovable. The buzziest company of the hour right now. On the wing list we just published and very buzzy company. That did the best, honestly. Spoiler alert.

[00:02:20] So far, I've had the most success with Lovable, you know, and for the viewer, listener, I don't know how to code at all. Like, I don't know anything. When I went to college, I studied philosophy. I was into journalism. I had no idea that I would be covering tech ever. Did not even cross my mind. Yeah, I did not take CS. Anyway, so no, nothing about coding, but Lovable, you know, created, uploaded our design elements like our logo.

[00:02:48] And so it borrowed our green and made really a nice website. You know, it's all the front end. I don't want to say exactly what I'm trying to build because it is a real product that I would be excited about. But the front end sort of exists. There are still some buttons that don't do exactly what I want, but I was able to add sort of user comments and annotations and cool. So that was actually pretty amazing. So Lovable was my like, oh my God, like this is magical.

[00:03:17] Have you played with any of the others? Have you touched windsurf or any of these guys? I have not done Codium's windsurf, though I should. And I haven't done Cursor, which honestly, I feel like is maybe the elephant in the room. I was going to say, if you're going to do one, it feels like you should do Cursor. My sense of Cursor was that, you know, it's used more by coders where Lovable is used by, you know, plain text.

[00:03:40] I went to Claude Code, Anthropics product, which is insane in that like you use the terminal of your computer. So you're like using like command lines on your Mac, which doesn't seem good opsec, but I have like Anthropic over total access to my computer. And it's much more autonomous. Like you describe in text what you want.

[00:04:07] I was joking on Twitter that coding with Claude Code is just like hitting one because it keeps asking me like, do you want me to do this thing? It's like one, yes. And then going to Claude online, just to sort of like, well, Claude Code's asking me this, like, what do you think? You know? And so you're just sort of like facilitating a conversation between two bots. I did end up with a locally hosted design website. Can't remember if buttons worked.

[00:04:36] The design was way worse, but it did feel insane in terms of the vibe. You did that all based on just plain English prompting. So you can build a lot with just that. That is impressive. Lovable was easier to like, both Lovable and VO, I think you're like ripping, you know, I take a screenshot of some stuff, you know, it's like, do something like this, you know? So there's definitely a risk that it's a little derivative, but you move fast from an image.

[00:05:06] Claude Code, yeah, it was text-based. All of them, I have sort of lost steam. I mean, it took a fair bit of my time. It sounds like the front end stuff is working really well. And from you and other people I talk to who play with it. But I wonder, I mean, the skeptical journalist in me is always like, if you get 80% of the way there, that's great. But if the last point like doesn't work, is it? How amazing is it really? It seems like smoke and mirrors, flashy demo day,

[00:05:32] and the actual tech work still needs to be tinkered out, which, you know, we support. Where's your read on where this actually is overall? I mean, you know, we experienced it with self-driving cars. You can feel super close and then you're actually far away. Now, self-driving cars seem to have solved that. I think we'll talk about them a little bit more later in the show. But for a long time, it was like, oh, they're nearly here. And it's like, oh, I don't know. Getting all the little details right matters.

[00:06:01] And with my experience, you know, I don't even think it wasn't like an 80-20 situation. I think it's more like a 20-80 situation. It's like, oh, yeah, I had, you know, the nice design. And sort of, I had the concept, but no back end. I have no idea how to like post it. You know, some of the features weren't working. That almost sounds like a competitor to like a Figma or like a design tool rather than like coding for websites. Yeah. Well, that's why people are saying vibe coding. Yes.

[00:06:30] I think that you understand this, but I think that's what a lot of people aren't in it. Don't get. It's like, yes, it's very compelling for the design part of the website. So that's why it's fun with front end. You're like, here's a screenshot. Here's some assets. Like play around with it. Oh, I don't like that. Re-shuffle it this way. You know, that's where it's a threat to Figma because, you know, it's, well, you know, we're all playing around with, you know, what ChatGPT just came out with, you know, on 4.0. They're really a Studio Ghibli machine.

[00:07:00] My Twitter feed is dominated by Studio Ghibli images. And I have to say, I've been building them. I've sent them to my wife. Oh, did you make some of you and your wife? Aw. ChatGPT, this is a real reveal here. Red pandas are my favorite animal. It's realized that it should just like put red pandas. Some of these images, I'll like them better. It's like, oh, here's a family portrait. And a red panda is in it.

[00:07:31] It's learning so much about you, Eric. Yeah, exactly. I love it. Which is what I want. I'm always shouting at it. Remember me. We've talked many times before. Use the context. You have all my things. I've asked you and uploaded. This is the dream. You know, we just get pandered to. Have you played with it yet? I've played a little bit with it. I'm more experienced with some of the older image generators. I was a really big fan of when like Dolly first launched. I spent like days on that, just prompting it with the most random crap.

[00:07:58] I think I did like any of the memes with like trail cam footage of someone in the woods and make everything look like a grainy picture when that actually became visible. So this I need to still play with. You're like, I have to file my draft. I don't know. I'm writing. I wrote a piece earlier this week. You know, now I get to play around the little images for work. The beauty of covering AI.

[00:08:21] I do, since I've been kind of spoiler alert for a little future work, talking to some people in the legal tech world recently, all of this image stuff has been making me think a lot about copyright laws. Are we just throwing those out at this point? Like the Ghibli stuff is very clearly Studio Ghibli style. And literally the CEO, Sam Altman is tweeting out like images of him in Studio Ghibli. Even in the new profile picture, Ghibli Altman. Right. Do they have a deal? I assume not. No, I can't imagine.

[00:08:51] Also, yeah, I don't remember Miyazaki being particularly pro this kind of stuff anyway. It's very fun. But then it opens up this whole host of defensibility from like a business perspective. Do you think it is a legal loser or like, you know, there's no law anymore and it's only power? I think it's great. I kind of am feeling like that's the direction we're going. I mean, Grok was the first to have the image generator that was no regard for copyright at all.

[00:09:17] And for the most part, OpenAI is still on paper looking to do that. There's workarounds and you know how to prompt it to get it to generate the style. So it's easy to work around. But at least on paper, they have their lawyers saying, you know, you can't do this this way. But I don't know if that's going to hold. Imagine you're an artist and you have this genius style. I wouldn't want every Joe Schmo to be able to copy it. I mean, it's like it's one thing if talented artists can sort of pay homage to it.

[00:09:47] But of course, they've learned to paint or draw. And so now they have some skill. And so maybe they try to be creative themselves. You know, this is the ultimate couldn't be more derivative. And even though it's fun, you know, for everybody to be able to do their family, because you're not going to be able to hire the artist to do it for you. It just clearly is going to dilute the artistic style. So putting aside the law, it's just sort of, you know, it's happy and sad. I'm doing it.

[00:10:17] But, you know, I wouldn't want to have an iconic art style right now because it's going to get copied. Madeline, what's your forecast on copyright? Right. They're wild now as they're acquiring customers and they'll get more buttoned up. Or you think this is the new normal? I think that we're still in the early days of adoption. Like as much as this feels ubiquitous to us being really in this ecosystem, not everywhere has really taken advantage of this. Just speaking also with my lawyer mother that I keep referencing on this podcast.

[00:10:47] They're very cautious about how to implement this kind of stuff. And so the sort of like 20% early adopters to 80% mass thing is really real with this. I can tell in a lot of industries and circling back to copyright. A lot of people are very nervous about this and don't want this to get them in trouble. So I don't think it's a total free for all yet. Long term, will it be? That's something that I think is certainly going to have to be renegotiated. Yes. But for the immediate time right now,

[00:11:14] I think people are at least in the content space are a little bit wary of doing this as directly. It's bleeding over more now than it was before. Like I think, I don't think Sam Altman would have used copyright style images. What's funny that it's like the media channels. If you're like NBC, you're surely paranoid about ripping off somebody's art style. But open AI, which is definitely more valuable than NBC.

[00:11:40] It's like, oh, the old guard companies that are responsible, but not as valuable, are more cautious than the startup worth, you know, 300 billion. I don't know. It's classic. Moving on. Different drama, different day. 11x, this sales automation AI startup backed by Andreessen and Benchmark. So some pretty heavy hitters was accused of falsely claiming certain companies as its customers. Eric, what do you make of this?

[00:12:07] I think short-term customers who turned off what were still advertised as customers on the website, maybe I think some annual ARR, I don't know if it was recurring revenue or runway revenue, but ARR maybe inflated potentially. But the really charged idea at the top of the story was that Andreessen Horowitz was going to perhaps mulling, suing the startup they backed, 11x.ai.

[00:12:37] And just like knowing anything about Andreessen, you need to have that like lock solid, I think, to lead off the story saying that Andreessen is going to sue the startup. I mean, Andreessen came out defending the startup and Sarah Tovel and Benchmark did. And they're not as hostile to the media. And so I feel like it was very predictable as soon as it came out. Like, oh, people are talking about whether Andreessen might sue. The other thing that was obvious in the story is like,

[00:13:07] clearly there are a lot of angry former employees. Like, you know, I think there was some picture in the story. And it's like, everybody in this picture is gone besides the founder. And as someone who has written about these types of companies, I mean, it is employees can get themselves incredulous that like, wow, we all left. And like, it's clearly a disaster not going to work. Sometimes they're right and sometimes they're wrong.

[00:13:33] And clearly, you know, VCs back the founder, especially Andreessen. Yeah, the VCs are doubling down. Do you think churn like that at a startup like this that's super fast growing is like an immediate red flag? I mean, based on your history of reporting, I feel like it really depends if there's something fraudulent going on for sure. That's obviously sketchy. But with this, it just seems like maybe not quite to that tier. It's hard. It's hard to know. You know, the business person in me, churn is bad, obviously.

[00:14:03] It makes it very hard to keep growing if you have to replace all your old customers. But if you're sort of early on in figuring out what works, you're obviously going to churn out customers who weren't the right fit for you. So it's hard to know just like churn on its own. Yeah. I mean, the issue is that it was easy for people to abandon. They had sort of a loose break clause, but then they baked in those customers to their ARR. You know, but again, this is sort of the type of thing,

[00:14:31] which is why we have staged venture capital funding. It's like if you sort of fake the ARR today and those people churn, unless you're really pushing it, you're going to have trouble the next round. So I think a lot of VCs are just sort of, this is the process. Why write an expose on a Series B company? Well, also a lot of these even really these early AI startups, there's been a ton of reporting about how a lot of customers for these early AI startups

[00:14:59] have a lot of customer interest and then a lot of churn, but then still retain a lot of customer interest. Like it's not like these companies are failing. VCs are still backing them and they're growing. It's just like a high, it seems like a lot of these companies just have higher turnover than maybe traditional SaaS companies did at this time. If you need fewer employees, you can churn the ones that are annoying to you as a CEO, I guess. I don't know, a couple wild tweets that I would never put in a story, but in a podcast I can read to you.

[00:15:28] One from a non-account basically is like, the fake revenue expose has begun. Theranos will be forgotten. Long live the AI sales agent. Next is Merkur, then scale perplexity, open AI. I mean, that's bombastic, but I do think there's like, we don't know anything about the nature of this revenue, but Arthur Rock, who has a track record of good reporting, said an AI company with a similar quote revenue trajectory

[00:15:57] likes to include early equity raise as part of their ARR calculation. Raised 20 million so far and is quite literally everywhere. I'm pretty sure none of its investors are aware of this. Good litmus test on who is doing most basic DD due diligence. Clearly, I mean, money's flying and there are gonna be some startups that do shady things and inflating revenue could be one of them. I think TechCrunch needed to be more buttoned up here

[00:16:24] about having this lawsuit idea at the top of their story when they had other interesting claims that they could have focused on. Regardless, this is a TechCrunch story that drums up a ton of media attention and hits just a week after their big sale, which had a lot of people scratching their heads as to what that means. Yeah, I'm interested to see. Obviously, TechCrunch has been an institution. I think selling to private equity does not make people more optimistic

[00:16:52] about the fate of a beloved brand, but Apollo owned Yahoo, which owned TechCrunch. So one to another. We'll see. Hopefully, they figure it out. I said earlier that we might come back to self-driving cars. It is now the time to do so. Waymo, my favorite. Near and dear to my heart. Waymo's planned robo-taxi launch in Washington, D.C. Intends to introduce by 2026. I mean, is D.C. going to be part

[00:17:22] of the Trump federal administration by then? I don't know. We'll be a separate city. I don't know. But anyway, they want to go to D.C. What do you think, Madeline? I think this is cool. Of course, the launch is dependent on if they can change some local laws that currently require human drivers in autonomous vehicles. So obviously, they're going to be pushing for that. So that it does go through. I think that D.C. as a neighborhood makes a lot of sense for Waymo. It's a bigger city. Obviously, it's nation's capital. The access of power is Silicon Valley

[00:17:52] moves closer in, you know, maybe probably a good call there. But I also think it's just exciting to see self-driving cars in growing cities. One step closer to getting to New York City. Am I right? Though I was reading something, I worry that New York is going to be a particularly big hurdle. Taxi cabs are powerful and highly regulated. Also, just the culture of jaywalking and the chaos of New York streets, I feel like is something to consider too. Not to be...

[00:18:21] You're buying car propaganda there. This is why self-driving cars, they're much better. I was looking at an analysis by Timothy Lee. He analyzed the 38 most serious crashes Waymo experienced between July 2024 and February 2025. 16 crashes where another vehicle crashed into a stationary Waymo. I think the only ones where Waymo is sort of responsible are passengers

[00:18:50] getting out and like, dooring like a bicyclist or something. Okay, for sure. Again, also human error. Yeah, in a certain way, though a product design. I think Waymo is trying to warn people when people are coming. But anyway, by any standard, he predicts that there would have been way more incidents with human drivers. I mean, I think it's pretty clear at this point that Waymo is superior. I think if you're riding them, you're like, yeah, this is definitely safer.

[00:19:20] So if you want to protect pedestrians in the streets of New York, we need Waymo. That's our show. Thanks for listening. See you next week.