Flying Cars, OpenAI Spies, Why AI Isn’t Sticking (Superhuman CEO)

This Week In Startups is made possible by: Lightfield: https://lightfield.app Every.io: https://every.io/ PayPal: https://paypal.com/business Plaud: https://plaud.ai/twist GoDaddy: https://www.godaddy.com/ Today’s show: Today’s college grads are showing up to work at Jason’s company unable to write three sentences. He’s asking what Shishir Mehrotra of Superhuman, the company behind Grammarly, can do to fix that. Shishir delves into Superhuman’s reasons for acquiring GPTZero, the most popular AI detector, and the AI notetaker Fathom, and the spreadsheet startup Rows. Plus, how Shishir responds to accusations that Grammarly helps students cheat in the classroom, why writers should disclose when they’re collaborating with an LLM, and Jason's CEO wish list for fresh Grammarly features. In the news, Tesla demoed its long-awaited Roadster, which used SpaceX thrusters to do some flips and propel itself over five other Teslas… And OpenAI warned staffers that spies may be trying to seduce them in order to steal company secrets… And the shutdown of Character AI prompts Jason to declare: “AI relationships are evil.” Guests: Shishir Mehrotra on X: https://x.com/shishirmehrotra Superhuman: https://superhuman.com Relevant Links: Rows, the spreadsheet startup Superhuman bought, announces the deal → https://rows.com/blog/post/rows-is-joining-superhuman Superhuman to acquire GPTZero (Business Wire) → https://www.businesswire.com/news/home/20260623083788/en/Superhuman-to-Acquire-GPTZero-AI-Authenticity-Platform Is the Internet AI, GPTZero's live tracker of AI content on Reddit, LinkedIn and X → https://istheinternetai.com GPTZero AI Vision, the extension that flags AI posts as you browse → https://gptzero.me/ai-vision Superhuman acquires YC-backed notetaker Fathom (TechCrunch) → https://techcrunch.com/2026/09/14/superhuman-acquires-yc-backed-notetaker-fathom-as-productivity-platforms-push-for-agentic-work/ Grammarly Authorship, which shows how much of your writing came from AI → https://www.grammarly.com/authorship GPTZero Authorship and Writing Replay → https://gptzero.me/authorship Stanley Druckenmiller says his WSJ op-ed was written with AI (NOTUS) → https://www.notus.org/media/stanley-druckenmillers-wsj-op-ed-bessent-ai Superhuman Go, the AI assistant that works wherever you type → https://superhuman.com/go Tesla reveals its new Roadster with SpaceX thrusters (TechCrunch) → https://techcrunch.com/2026/10/15/tesla-reveals-spacex-powered-roadster-that-flies-and-falls-with-style/ OpenAI employee arrested over a hidden listening device in an executive conference room (NBC News) → https://www.nbcnews.com/tech/tech-news/openai-employee-arrested-hidden-listening-device-company-says-rcna603262 Eight states sue OpenAI over ChatGPT's alleged harm to kids (NBC News) → https://www.nbcnews.com/tech/tech-news/openai-sued-eight-states-alleged-harm-chatgpt-rcna603614 Timestamps: 7:14 GPTZero and the fight over what's human 9:27 How much of Reddit, LinkedIn and X is AI 10:56 Fathom and the 90% of work that's spoken 12:05 Lightfield - Name one person who's ever enjoyed updating a CRM. Exactly. Lightfield's AI agent does it for you - it even prospects and books your meetings. Used by thousands of startups. Free at https://lightfield.app/ 16:27 Why Jason says every employee should use Grammarly 18:04 How to test whether new hires can actually write 22:08 Every - For all of your incorporation, banking, payroll, benefits, accounting, taxes or other back-office administration needs, visit https://every.io/. 24:22 Grammarly Authorship and proving the work is yours 25:53 How accurate are AI detectors 30:36 Jason wants AI writing flagged by default 32:34 Enforcing Grammarly and Go across a whole team 36:45 PayPal - PayPal is built for all businesses. Get started at https://paypal.com/business 37:28 Assist, chat and do, three kinds of AI agents 41:06 Jason's feedback on the Grammarly keyboard 1:13:27 Tesla demos the Roadster on SpaceX thrusters 1:21:20 OpenAI engineer charged with bugging an executive conference room 1:21:43 Why OpenAI warned employees about foreign spies 1:27:57 Eight states sue OpenAI over harm to kids Subscribe to our newsletters on Substack: TWiST: https://twistartups.substack.com/ TWiAI: https://www.thisweekinai.ai/ TWiVC: https://thisweekinvc.substack.com/ Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp Follow Lon: X: https://x.com/lons Follow Jason: X: https://twitter.com/Jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis Check out all our partner offers: https://partners.launch.co/ Great TWIST interviews: Will Guidara, Eoghan McCabe, Steve Huffman, Brian Chesky, Bob Moesta, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland Follow TWiST: Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin Instagram: https://www.instagram.com/thisweekinstartups TikTok: https://www.tiktok.com/@thisweekinstartups Substack: https://twistartups.substack.com

• 0:00Speaker 00

  All right, everybody. Welcome back. As you can see from my background, I'm back in the saddle. I'm back in Austin, Texas. And the weather lawn is perfect. This is the perfect weather. It's been like beautiful, 80 degrees, 75, 85 degrees on the ranch. I threw some wifi up on the horse pitch. I got a table outside and I've been trying to get through my jet lag. But yeah, I just had a really productive trip in Japan for two weeks. And I'm ready to get back to the regular schedule, everybody. I was taping episodes at all times. And you know what I always like to do if we're if we're this busy? Is I gotta keep my plot with me. Cut my plot with me on the back of my phone. I'm walking around.

• 0:44Speaker 00

  Tokyo. Man, I had some days, Lon, I walked six, seven miles.

• 0:48Speaker 01

  Walk a lot. It's a

• 0:49Speaker 00

  Do a lot.

• 0:50Speaker 01

  Trains, but then you get to the neighborhoods and you just want to mill around and go down all the little side streets and see everything there is to see. You can't see it all, so.

• 0:57Speaker 00

  Yeah. And if you're an old man like me and you get addicted to rucking, you throw the backpack on, you walk around. Bang, bang. I hit two different lunch places. Bang, bang. Two different cafes. I take my laptop out. I work for an hour or two, go to the next place. Just had a wonderful time. Visited every neighborhood. I went to the book district, the used book district. I went to the restaurant district. I mean, I just went to each of the different districts. Have my wife with me. We had a wonderful time. Stayed at an incredible hotel, the new Janu Hotel in Azubudai Hills. Absolutely. The most money I've ever spent on a

• 1:34Speaker 01

  Well, that's where the that's where the swells live in Tokyo. That's like the hot neighborhood is Azubudai. I heard about that when I was there.

• 1:40Speaker 00

  Wait, who's the swell?

• 1:41Speaker 01

  You know, the the the elite, the the one percent.

• 1:44Speaker 00

  It is the elite area. Yeah, it's the 1% area and they have a 1% hotel, I can tell you that. But we've got a lot to get to and I just want to make sure I don't miss a moment. So I'm gonna take my plot on the back of my new iPhone 18. I got off the air. Because I needed a good photo, take photos of ramen. The air, the air wasn't taking good photos of ramen. I just pressed the button, boom, and now I'm in record mode, put it right on the table in front of me. I got all my notes. We use it for the notes for this week in startups.

• 2:10Speaker 01

  We do. And you can see I've got mine on right here. Any kind of conversation, interview, meeting, call, interaction you want to hang on to for later. You just press the button, save it. It's on your desktop. It's on your phone, wherever you need it. If you want to check out Plod, go to plodai dot twist and use the code twist. And they're going to give you 10% off your Plod note pin.

• 2:30Speaker 00

  There it is. All the transcripts come in. I've got a lot of people asking me, which one should I get? I always just tell people, get the two pros.

• 2:37Speaker 01

  Sure.

• 2:37Speaker 00

  Just buy both pros because they both sync automatically. You got it for two different situations. The one that goes on the back for meetings is just awesome. Also, you can record phone calls without having to like invoke another piece of software, crash your computer. You just have it on the back and it records it. I don't know how it does it with technology. Of course, make sure you are letting folks know you're, hey, may I record the call, as we say in the business these days? May I record the call?

• 3:05Speaker 01

  Yes.

• 3:05Speaker 00

  Always a good thing to do. Plod's the best. There it is. One press, record, and then you get the pin as well. And hey, our guest is here. Let's introduce our guest and get to work.

• 3:14Speaker 01

  Once again, joined by the great Shashir Marotra. He's the CEO of Superhuman. They of course make AI tools, including Grammarly, that help you write your documents, manage your email, get work done anywhere that you are, any, any site that you're on, any platform that you're using. You could fall back on these amazing productivity tools from Superhuman. Shashir, thanks for joining us.

• 3:36Speaker 02

  Great to be here. Thanks for having me.

• 3:38Speaker 00

  Yeah, and you've got a lot going on in your world. In some ways, are you becoming like the productivity version of bending spoons? Because every like three months, you seem to buy another productivity tool. You obviously Grammarly bought Superhuman, Raul's company, which we were the the first investors in, although some people will claim to be the first investors and we'll have that debate till the end of time.

• 4:04Speaker 02

  I won't dispute it here.

• 4:07Speaker 00

  I'm gonna start just getting like, Shashir, I'm gonna start requiring in the documents, you must accept my money first when I make these investments. Smart founders are like, yeah, if you get it in on this day, you'll be the first investor. But but tell me a little bit about what's going on in M and A land. You have bending spoons, buying air table and other assets for one times, two times revenue, and then managing them more effectively with AI. And I think you have, since the last time you've been here, maybe you bought two or three companies. So walk us through what's happening in M and A land for the companies that are cashed up like yours, companies that have, you know, big installed bases like yours and have the

• 4:52Speaker 00

  you know, ability like your management team does to integrate great companies, which is not an easy task.

• 4:57Speaker 02

  Yeah, I think since I last came on the show, we bought three companies, Rose, Fathom and GPT-0. The bending spoons comparison, first of all, I have enormous respect for bending spoons. I think they're doing amazing work. I we learn a lot from them. It's a comparison I both love and hate. I mean, on one side, I love it because, you know, there is a there's some camaraderie in the, you know, I think we're probably the two most active acquirers out there right now. And we definitely see it in the market when people are are looking to sell. Other way, I kind of hate it because their strategy is like nearly the opposite of ours. I mean, they, their, their strategy is mostly to to buy things

• 5:38Speaker 02

  that are sort of past their prime and extend the life. And I think they're they're doing a good job with that. And they've taken some brands that nobody thought could be rejuvenated and and try to squeeze the extra out of them. We mostly do the opposite. We take things that are growing but lack distribution and we give them a bigger platform. And, you know, we take the number one or number two AI detector on the planet and we give it a much bigger platform to go to go run on. We take a popular note taker with Fathom and give it a much broader base of users to run on, much broader set of surfaces to run on.

• 6:12Speaker 02

  But I think, you know, it's a love and hate comparison.

• 6:15Speaker 00

  So walk us through each one of the acquisitions now. There were three, I believe. So let's walk through each of those. Tell us why you bought it, what the thinking is here, and then how the acquisition might

• 6:30Speaker 00

  how you think about integrate or standalone, integrate or standalone. Obviously, Superhuman has now become super integrated. I noticed you redid that little bit of the design and the design and some features are starting to get in sync, but it does feel like when you have these, excuse me, robust programs, it's a little bit of work. And so you got to be thoughtful about it.

• 6:50Speaker 02

  Yeah, I think that's that's fair. I think the, so the three that we've done since we last chatted. So we bought a company in Portugal called Rose. They make a spreadsheet product that's being fully integrated. So that's turning into a new set of capabilities that'll show up in what's currently, Superhuman Doc says, we're building out that suite. It's Superhuman Spreadsheets, Superhuman Databases and and so on. And that'll be coming out soon. It's it's pretty exciting. It's a really fun product.

• 7:19Speaker 02

  The next one we did after that was GPT-0. For people who aren't as familiar with that category, GPT-0 is market-leading authenticity company. So the product they're most well-known for is it's the most popular AI detector out there. And they've surrounded that with a set of other products that are all associated with different forms of authenticity. The

• 7:43Speaker 00

  That's GPT-0. Yeah, now this is an interesting. Yeah, GPT-0 is very interesting. We have this issue now that people want to know. Lon, we we come across this a bunch. Is this AI slop? Is this written by AI? Was this written by a human? And it's a bit, it's a bit of a debate. In some areas, it's essential that it not be written by AI. In some situations, hey, if you're making copy for a website, maybe it's better that the AI be given that task. So maybe walk us through how people use that product.

• 8:20Speaker 02

  Yeah, so it's, I mean, it's used in in all those different ways. So the, and I think the general view of the world is we should be using AI, we should be using it responsibly, and we should be clear when we're using it and when we're not. And and it's not the I don't think we're trying to take a hard stance on you should always do one or the other, but you should know and disclose when you do one or the other. And you should know when you're talking to the human or when you're talking to the the AI that was at the direction of the human. It's kind of like if I if you call me and I put my intern on the on the podcast, you'd be a little upset.

• 8:51Speaker 02

  So the, you know, you got, I got to show up to to to do this. So that's, that's a tool. It gets used in a wide set of industries. So it's very popular in education. So students and teachers use it a lot as a way to verify proof of work, proof of understanding. It gets used a lot in journalism. It gets used a lot in interestingly, the scientific community is using it a lot for for things like verifying paper submissions. The recruiting has been a big market. So as you're, as you're getting resumes, as you're getting

• 9:25Speaker 00

  Yes.

• 9:25Speaker 02

  submissions, like I just want to know, like, is this them or is this not them? And so it's, you know, it's interestingly broad of, you know, understanding authenticity. They have a really fun webpage. If you go to istheinternetai.com. It's a really interesting look at, we've been scanning Reddit, LinkedIn, Twitter, so on, and

• 9:46Speaker 00

  Oh.

• 9:47Speaker 02

  And, you know, we have a sense of what's, how much is AI versus not in every place.

• 9:53Speaker 00

  So look at that. 6% of Reddit is AI, 13% of X is AI today. And this is just doing today.

• 10:01Speaker 02

  If you, if you, if you scroll down, you'll see some charts plotting out how

• 10:08Speaker 02

  How the internet is filling up.

• 10:09Speaker 00

  Oh, no.

• 10:10Speaker 02

  And, uh, yeah, so you'll see. Yeah, which is, I mean, it's interesting, right? If you look at LinkedIn's at the top of that list, right? Reddit at the bottom is also really interesting. You know, if you, you want to know where to find humans right now, you're gonna, you're gonna find more of them. You're gonna find more of them there. Like I said, not necessarily a bad thing, just a thing you have to know. As a reader, you want to know, there's a, uh, GPT zero ships a tool called AI vision that while you're scanning those websites now, you get a little flag that says this is likely AI or this is not. You can suppress them, you can, you can read them. Yes, that, that, that's what you want. But if you're looking for humans, you can, you can go help find them. So that was acquisition number two. That is staying as a fairly independent product. It's very broadly deployed, but will run as an agent on Go.

• 10:52Speaker 02

  And, uh, uh, we can come back and talk more about, about Go, but Go is basically the platform layer of Grammarly allows you to build agents that work everywhere you work, just like, just like Grammarly does. And then the, the third acquisition was most recent, um, was Fathom. So Fathom is a popular, uh, note taking tool. The, uh, you know, the software version of the thing you were showing at the beginning of the, uh, of the show, uh, used by millions of users, uh, really popular product. Uh, and we're taking this and we'll be launching with this one a more integrated product. So we'll be launching a integrated note taker, right? I won't say too much about what we're doing because it's really awesome.

• 11:33Speaker 02

  But you're gonna love it and it's a, it's really taking the heart of what I think AI enables with note taking and what we enable with being everywhere. Like our products work where you work and every single surface. You know, imagine a note taker that really followed you everywhere you went. It's a slightly different take than the, the, the, the hardware solution.

• 11:54Speaker 00

  The Claude one, yeah.

• 11:55Speaker 02

  Yeah, but it's a, you know, then the same vein of, you know, I want a note taker that works the way I work. And I don't have to bend to it. Uh, we think, uh, we think there's a really amazing opportunity there.

• 12:05Speaker 00

  All right, I'm here with Keith Piris. He is the founder of Lightfield and they are a CRM system built natively in AI that takes all your emails, calendars, Slack messages, all the stuff you keep losing, puts it in one place and creates organization out of chaos. So you don't lose any customers. Keith, welcome to the program.

• 12:23Speaker 03

  Thanks for having me.

• 12:23Speaker 00

  Let's talk about lead enrichment. You mentioned that. What is lead enrichment for people who don't know? It's, it's kind of implied, but I think sometimes people hear that term of art and don't know what it means. And how do you do it?

• 12:33Speaker 03

  Lead enrichment comes from, I think, the, the early 2000s where you used to buy a, a database of names, email addresses and, you know, other parameters like job titles, company size. And then, uh, if you had one of those parameters, you could use it to find the other. Like maybe you had the name and you want to figure out, hey, where does this person work? What's their email address? Or maybe you had the email address and you're like, uh, what company do they work for? What kind of company is this? And, um, this used to be, uh, you know, I, I would say a pretty big business because you've had these, these companies like Zoom, although that would pay thousands of people to go and call and make notes and fill out this. Annoying people.

• 13:18Speaker 03

  Yeah, annoy people and fill out this database. And, uh, in the new world, um, you know, that's not how lead enrichment works at all. You, you could get to use one of these, um, web LLM tools, uh, go and scrape the web, find the, the newest updates and then connect it to your database. Uh, so sort of lead enrichment in, in our world is like use, you know, a modern LLM, uh, web provider like Exa and, uh, get them to go search through a million different documents and come back to you with an answer.

• 13:49Speaker 00

  All right, everybody go check out lightfield.ac and see what it can do for your company. Don't lose any customers and find some new ones, lightfield.app. Uh, I'm, I'm interested in, as far as the AI note taking goes, I mean, there's so many, we joke on the show all the time that before any of our guests show up, we get like, we have to check in there for AI note takers before the show because everybody's got a few of them. I mean, when you're going into that market, is it really just, well, we've, we've already got your emails. We've already got your documents, so the notes are going to be better. Or like, how do you sort of slice up that market and figure out here's where our AI note taker really needs to like excel? And what, you know, where, where do you focus when you're building a new one of those?

• 14:29Speaker 02

  I think, I think giving, if you think about the world as, as, um, you could think about it as pre-doing and post-meeting. You can think about it as, uh, uh, how you, how the note taker fits in your day. Giving more context to your note taker is definitely a, a positive thing. And we can do that through, through our other products. Like you mentioned, we have a very popular mail and calendar product. Uh, we also through Go, we integrate with just about everything. And so it's definitely the case that a note taker that has read your, uh, your CRM system and understands tapping your task tracking system has a better chance of taking good notes. But I think actually the, the experience, first off, in the, you know, what it feel like to take notes is an experience that is, uh, I think has been quite broken.

• 15:12Speaker 02

  And, uh, you know, we, we, we end up with, uh, with note takers that really force you to change your behavior and how you think about taking notes. And then, but much more importantly is after. And I think the, like, one of our core insights was that if you look at any company or any team, you know, something like 10% of what you do is written down and 90% is spoken. It's all forgotten. And so where does that all go? And how do I make that actionable, uh, for you? And just, I mean, just to give you a very simplistic example is, um, you imagine that you're writing your email to a, to, to a customer. And, you know, today you have Grammarly running right alongside you and Grammarly comes in like your, like your grammar teacher in your pocket, saying, hey, you've got the incorrect spelling or grammar here,

• 15:55Speaker 02

  or you have a passive tone here and you're unlikely to close a deal. But now with everything we know from all of your notes, we can also say things like, hey, that's actually not what happened in that meeting. And you got this fact wrong. You, you said, you're, you're, you, you said they want a thousand seats. Actually, in the meeting, they said they want 2,000 seats. You're about to get half the deal. And so that idea of taking everything that you can do with a note taker and make it available to you everywhere you actually work is the, you know, one of the hearts of the, of the, that theory of generally of our product suite, but of what we'll, what we'll do with, uh, the note taker.

• 16:29Speaker 00

  There, there is an interesting dichotomy coming now, which is what's human, what's not human? And education is such an important moment. You, Grammarly is a very polarizing product in education. There are people who are like, you have to use Grammarly. And then there are people are, you, you cannot use Grammarly. You have to learn where the comma goes yourself. I am a part of the former camp, which is you have to use Grammarly. The biggest frustration I have is we have an associates in training program here. People are coming out of college, they can't write. She's here, people are coming out of college, they can't put together three sentences. And you're like, well, wait, didn't, didn't we figure this out in your job application?

• 17:14Speaker 00

  That you're incapable of math and writing? I mean, it's a serious problem in America. And then what I love about Grammarly is I don't care if you don't know the commas right now because Grammarly correcting you on a regular basis actually kind of makes you learn it o- over, over time. So I, I'm in that camp, but we now have to assess writing ability. So I'm curious if you've thought about writing assessment and training in addition to just, hey, let's fix it. Because we, as an investment firm, as just one example, we had this problem. People are like, this person can't write a deal memo. This person's great at a deal memo. We're going to have this person not write deal memos.

• 17:58Speaker 00

  We're going to have this person write the deal memos. And then in an organization, I said, wait a second, you cannot be a venture capitalist if you can't write a deal memo. So that's a prerequisite for the job. I told my team, how do we assess this before we hire somebody? And then what is it going to take to train somebody inside the company? You know what I came up with, she's sure? I said, put them on a, a Zoom with another person and watch them work. And then tell them to, you know, talk out loud what they're deciding to do. And we now have to watch people acoustically write. But talk to me a little bit about assessment. Yeah, and, and this issue that we're starting to see in the workforce, because I've talked to other folks. I'm not the only person with this issue. Sometimes we'll get somebody to write something and we're like, oh my God,

• 18:42Speaker 00

  this person is a college educated person who spent a quarter million dollars on an education. They can't write.

• 18:46Speaker 02

  So, uh, uh, I'll, I'll come back to assessment. You said a couple of things there that I, I are worth reacting to. Is on Grammarly being polarizing. Actually, I mean, every product is polarizing. I would say Grammarly is the least polarizing AI tool in the, in the, in the suite. So if you look at many schools and universities, we tend to be the one product that people are much more okay with than the others. And I think there's good reason for that. But I, I think it's an important thing to build on. Part of that is we've been able to show in many cases mathematically, what you said is true, that the students who use Grammarly actually learn better grammar. And, uh, we've worked with a set of universities to, to do some studies on it. In some cases, it's like meaningful changes and they, they do a holdback study and some people get Grammarly and some people don't.

• 19:31Speaker 02

  Then they look at the pass rates on the class and they look at the exams where you're not allowed to use anything. And then they look at what the scores are. And so it's definitely possible to use AI in a way that is sometimes is replacing you and sometimes is augmenting you. It's our view, it's like part of why we picked the name superhuman. It's our view that we want to build tools that augment you. We want to build tools that allow you to be the super version of you. I had this, this, uh, line I love in the company that we spent 17 years turning people into super writers. Now we're going to turn people into superhumans. And so I think it's a, I think philosophically is how I think about it. The assessment thing is really interesting and I spent a lot of time on this topic. As you might imagine, that, that category is very important to us. I spent a lot of time with provosts at universities, uh, with deans and professors and, and, you know,

• 20:15Speaker 02

  and with primary schools as well. I, I think, you know, the way I think about it is. It's helpful to use the calculator analogy because I think when calculators came out, there was a similar fear of nobody's going to be able to do math anymore. And there was definitely a camp that said, why should you do math? You always have a calculator. And it's a different group of people that says, that's silly. Like, doing math is just a fundamental skill. You have to be able to do good math and you're going to catch the errors and you're going to... And so what did we do? Like, at first, everybody freaked out. And then we did, like, the obvious things. We said, in this assignment, you can use calculators, and this one, you can't. The SATs have a whole section that is with calculators and a whole section that is without calculators. And, you know, you want people to be good at whatever the tools or the best tools in the marketplace,

• 20:57Speaker 02

  because we're not hiring you to, you know, use less tools. We want you to use the tools. And at the same time, you want fundamental skills. And so I think the answer is you do some mix of both. And so when you give them the assignment, you need to give clarity on this is what I expect. And this applies in school. This applies in interviewing. This applies on the job. And then at the same time, you want to assess that way. And you want to come back around and you want to say, hey, look, and the example you gave of, you know, oral observation is a very common technique. Oral exams after, you know, I just want to interrogate you after. My daughter's a junior in college now, and that's how it works. Like, she's got a bunch of classes where use whatever AI you like, but we're going to call you in and you're going to answer questions about it. And, you know, we want you to use AI.

• 21:42Speaker 02

  We want you to be prepared for the workforce that we're all headed towards. But we also want you to understand that you just hired AI. It's like you hired a team to work with you. You're responsible for their work.

• 21:52Speaker 00

  All right, listen, if you're running a company and you're the founder, hey, you're going to have so much work to do. You got to hire a team. You got to build the product. You got to do customer interviews and understand what your product or service is going to do for your customer. Well, you're also going to need to incorporate and you're going to have to do your taxes. You have so much work to do on payroll, benefits, taxes, banking, incorporation, and you need a partner. And that partner is Every. Every has worked with over 1,000 startups. You're going to make mistakes, and some of those mistakes are going to be costly. That's why you need a partner like Every. They're going to handle your Delaware C corp, your EIN, and getting a registered agent with no

• 22:37Speaker 00

  legal fees and no delays. They're going to get all that done for you. And if you're further along and you need to scale up your team, well, Every will help you hire contractors in over 200 countries and do it right. Tight is right. Plus, their banking platform keeps your funds board compliant while making it easy to collect receipts, pay reimbursements, and stay on top of your spending. So head over to every.io and stop wasting time on stuff that isn't growing your company. That's every.io. I think you got to just, if you can provide this free, Shishir, it's going to be an incredible growth tool, just like you have that landing page for like, what percentage... If you just had a thing, you know, what's my writing ability? And you gave it to your corporate customers where I could just give it to everybody on

• 23:22Speaker 00

  the team, force them to do a writing assignment where it tested it, and just benchmarked everybody on the team? Because I've literally gone down the rabbit hole and I've looked at all the assessment tools. There's no good assessment tool. Like, if you find an assessment tool, it's like you have to buy an assessment platform for $3,000 per person, and then you have to write the content and build it. It'd be like, I would need to have an assessment person in the company. I need a lightweight assessment tool, just in Grammarly, in Superhuman, boom. Or just look at everybody's output and then just rate them. What is the actual output of this person's slacks and Superhuman email output? And when they're typing stuff? Yeah, just give me Slack, Word, whatever.

• 24:07Speaker 00

  Just tell me the person's a 72, and then I can give them coaching. Am I crazy, Shishir? Doesn't Grammarly already do something like... I feel like I get scores back from Grammarly where it's like, your vocab's in the top 97% of all Grammarly users. That's a true stat, folks, by the way.

• 24:23Speaker 02

  We do some of that. We also have a tool that's like, it's not exactly what you're asking for, Jacob, but it's called Authorship. And actually, GPT-0 has a similar tool called Replay. And that is, but we put it all in the hands of the person. So they're basically, the way Authorship works is if you're a student, for example, you say, I'm writing this paper, you turn on Authorship, and it will watch what you're doing and vouch for how much AI you used and didn't. It's not quite what you're saying about what is the quality of your writing, but it's vouching for how much AI you used and didn't. But you then take the report and then you hand it to your professor, to your boss, or to the person interviewing you. And that's pretty effective.

• 25:06Speaker 02

  We just watch your screen and figure out everything that's happening there. The, I think the idea of doing that as a writing assessment, we're working on something that's close, but I'll register the idea.

• 25:15Speaker 00

  Here's the thing. You're selling to the CEO ultimately. Like, we have to, some CEOs are going to get this package. There is something very appealing to CEOs or bosses about just knowing on the team where everybody's at and having a general assessment of people, as opposed to like what it does now, which is like passively tell Lon, hey, you're great. So Lon can be like, I'm great. We know Lon's great. That's why he's editorial director. What we need is to pull everybody up. I need to start from the bottom and take the people who are bad or weak and make them OK to good. Then I got to get the OK good people to great. But I don't need to worry about the great people, you know, getting, oh, I'm 97%, I'm 98%.

• 25:56Speaker 00

  I need to get the 50s, 60s, 70s up, but that needs to be like a dashboard. Lon, you got a question?

• 26:03Speaker 01

  I do. I want to ask about, because I, now I'm, I sort of edit all of our Substacks. And so one thing that Substack now allows you to do is check everybody, like, how, what percentage of all of these posts were AI? And I get a lot of, there's some pushback from the team sometimes where they'll be like, I wrote every word of this and I'm still getting an 85% human written score. And I see this on social media all the time, too, that people are like arguing. So, one, my question would be like, how accurate is GPT-0? Like, how, like, how much should people be trusting these tools? And then, two, are you hearing that, too, that people are complaining that like, it's not accurate? Like, what can we do to make these, establish these as like a source of truth more for people, rather than just the next level of the AI slop debate?

• 26:47Speaker 02

  I mean, there's definitely, Substack doesn't use us. They use a competitor. They, and so I can't vouch for them. I think there are some standard benchmarks out there, and there are, in particular, a couple of organizations that are rating AI detectors. And it's not easy to, you know, to rate the detector of the, you know, it's a sort of a cursive problem. But on those benchmarks, you can get pretty high. I mean, these products get to, you know, high 90s in terms of being able to detect AI. It does better when it's longer, like, it's easier to detect AI. When it's shorter, it's harder. And I'll say that in many cases where people feel like, hey, this is misassessing me, it's

• 27:30Speaker 02

  usually on the margin, and it's, I would say I trust the AI detector more. It's more likely the AI detector is right. 85% may, you know, may be plenty. And it may be the case that it's, that that's appropriate. And you, you pulled a quote from somewhere and you didn't know where the quote came from, or you pulled a concept from somewhere else and you didn't realize that that had been refined by AI somewhere. And it's nothing necessarily wrong with it. I mean, I think generally the idea of like, you're at 85%, like, I want to be at 100, back to what you were saying about the proficiency score. Like, the thing we're trying to avoid is slapping your name on something that's 0%.

• 28:05Speaker 00

  And getting busted. I mean, that's actually the big issue here is trust. The, Sir Druckenmiller said he used AI to write this piece, and the Wall Street Journal let it through. So my take on that, I'm interested in hearing yours, Shishir, my take on that was shame on the Wall Street Journal. They're not checking the submissions. And then I was like, eh, you know, usually like someone like Druckenmiller would have a co-writer. So, you know, if I'm writing a piece, I'll have Lon jump in and read it for me or Jackie for my team. But I'm a writer, so I write it myself, then I pull them in. But I understand some people might not be great writers, so, you know, they might have somebody work with them on a draft, in a conversation. OK, fair enough. But that just felt like a trust issue.

• 28:49Speaker 00

  So what was your take when you saw that?

• 28:52Speaker 02

  I think that's the right analogy. I mean, if you have someone that's working with you, like, I'm in the middle of writing a book, I have someone writing it with me, they're going to get a byline. They're gonna show up there and it's gonna say, you know, this is by Shimer Otra with help from Aaron Dame. And I think that's, I think that's okay. I don't, I don't think using tools is wrong. I think that you have to own your ideas and you have to disclose your collaborators. And I think at the end of the day, you put your name on it, you know, take Drucker Miller as an example, he stood by it. He said, look, I, yeah, I used AI. I, I, I think that's fine. He should have disclosed it. I think he said, written with the help of, I think that's okay.

• 29:26Speaker 00

  You really would have put it that Drucker Miller with Claude Fable 5-1.

• 29:30Speaker 02

  I mean, you could stick it at the bottom, like enhanced by, you could be proud of it. You can say like, you know, there's definitely, you know, there's definitely bloggers I read that like regularly talk about how this image was, you know, I couldn't figure out how to visualize this. And then I, I asked, you know, I asked Claude to do this thing and it came up with a perfect visual. And it's a great way to describe it like.

• 29:48Speaker 00

  I'm okay with that. Yes. That, to me, that seems fine. Maybe I'm a bit biased for words.

• 29:54Speaker 02

  But I think for me, the bigger chess is like disclosure of your collaborators is very important. Just that's just a normal human decency thing. But you gotta own your content. You put your name on it, you gotta own it. And yeah, like you're right. Some people need, need help. They, they, they need a, they need a co-writer. I'm like you. I, I write myself first and then, then I let my co-writers in. Otherwise, my ideas don't ever get to the right, the right spot. My team complains about it. It's a.

• 30:19Speaker 00

  Well, because you're, have to wait for the inspiration. You have to wait for the words to come out. And they're like, well, we could just give you the words. And that's not how great writing occurs.

• 30:29Speaker 02

  That's not what I would say, right? I'm not putting my name on something I wouldn't say.

• 30:32Speaker 00

  You, you now have this AI detector. You have Grammarly. It's checking my text. Why can't I go on the offensive? Why when I'm on X myself or I'm in an email, why aren't you checking in Superhuman if the person who sent me the email is using AI?

• 30:48Speaker 02

  We, we were starting to. So that's all coming. So the, the, the, in, in mail, some of the systems that were integrated already, they're silently showing up as auto labels. And so our spam detection is getting much better. And we can, we can start looking for those things. Again, it's not necessarily bad, right? You, you, you wanna use it. But you know, for me, Superhuman Mail is like a, a, a lifeline for me. I don't know how, I don't know how I would live without it. I mean, it's the, the, I would be inundated. And my role, it's by my inbox would be completely overfilling. And it's very helpful to say, look, I'm gonna deal with the humans first. It's okay. You use AI, I'm gonna, I'll get to you.

• 31:27Speaker 00

  Yes. Sorted by percentage human. I love that. Human stuff moves up, but you gotta just, you gotta show me. I think you should be controversial here. I'm gonna say something super controversial. I think you should label the tweets that I'm replying to and tell me percentage, uh, chance. Same thing. You should be on my side. I'm paying for the products this year. I'm the customer, not the person sending me information. Rat them out, Shishir. Rat them out.

• 31:54Speaker 02

  That product exists. It's called AI Vision. And, and if you, if you install a GPT-0 extension, uh, you can flip, flip it on. It'll, it'll start flagging things on LinkedIn and X and so on. And it'll tell you.

• 32:06Speaker 00

  Really?

• 32:07Speaker 02

  Yeah. Yeah. So it's, it is a big question about how, how we treat that. But I agree with you. Tell me what's human is a, it's a hard part in a world full of, you know, we're watching a world get overrun and with, with slop. And I think there's gonna actually be this return to humanness. I think it's gonna return to shows like this. Like it's like really hard to fake this, right? And so I wanna know what the.

• 32:29Speaker 00

  Trust me, I've been trying for 12 years.

• 32:30Speaker 02

  We've got 11 labs. There's smoke coming out of the back of the computer, but it's not working yet.

• 32:38Speaker 00

  You have no idea how controversial this feature is gonna be. When, when, cause I have it, I have Grammarly on. I, the biggest issue I have in my company with Grammarly is enforcement. I need you to, there's another product thing, since I have here. I need you to tell me when it's not running for more than 24 hours on a paid account. So if it's not running, I need you to email the admins and say it's not running on Lon's computer. Constantly, because you know how computers work, software gets updated, people forget to auto start, whatever. I need it to tell me it's not running on a paid account. Or you should punish yourself. If it's not running every day, you should not charge me. And you should say, hey, it was running less than 15, 50% of the time. We're not gonna

• 33:21Speaker 00

  charge you for that account. Because people on my team are constantly, I catch them with mistakes. And they go, oh, I forgot, I forgot to turn Grammarly on. And I'm like, that's your job. But I need you to enforce.

• 33:34Speaker 02

  Are you pushing everybody to use Go yet?

• 33:36Speaker 00

  Yes. Yes. But you know, there's no enforcement tool here, you know? So.

• 33:40Speaker 02

  The thing we're seeing a lot with, by the way, Go is, is going like a rocket ship. I mean, it's a, it's, it's amazing. Uh, you know, multiple points a week on week growth. It's a, it's a really, really fast growing product. But one of the things we're finding is that especially in team environments, there's often someone like you that says, I have a set of things that I really want people to do. Um, yeah, there you go. Um, there's a set of things that I want people to do that are, um, standard across the company. And so it might be grammar, but it might also be, you know, uh, I have the style that we wanna use. If we're working with a large, uh, large newspaper and they've got a series of, uh, style guides that are, there are sets of things. We only talk this way. We,

• 34:24Speaker 02

  we need to check the facts this way. We write titles this way. And they go and they deploy an agent on Go. It looks just like Grammarly. So like that, you know, it shows up, the underlines show up that, you know, gives you hints right where you're working, uninterrupted, invisible AI. Um, and you can then go enforce it. It's also been used by regulated, regulated industries. So people come in and say, never, ever say X. Like we work at a bank, you are never allowed to promise a stock price. Um, and they'll come in and.

• 34:50Speaker 00

  Compliance, right?

• 34:51Speaker 02

  Yeah, compliance, right? You know, I'm a lawyer. Like, these are things you're not, you know, lawyers can't say or can't say. And then you can connect it to your systems. And you can, you can have it pull all the information right back in front of people. That set of products, I think this, I'm gonna require everybody to have it on all the time. It's a very common request.

• 35:09Speaker 00

  You know what I do? I go to my general channel on Slack. I say, take a screenshot that you have, uh, Superhuman Grammarly turned on. And then if somebody doesn't reply, then I have somebody go break their chops and be like, what's going on here? Why didn't you sure reply? So I, a screenshot proof of life is what I do. But it needs to send the admin of, an engagement report, right? Because if it's sent an engagement report, who's using, who's not, then people, it will reduce your churn. I'm angry customer number one for Superhuman now. I'm, I'm cantankerous customer one. Is there anything else I'm cantankerous about? I've been in a cantankerous mood. What is my, I am, I'm cantankerous

• 35:51Speaker 00

  about people not complying, Lon. I'm cantankerous about people not learning to, uh, write in college.

• 36:01Speaker 01

  I got, I, you, you like, you like getting responses in Slack quickly. You're cantankerous when you feel like you're not getting prompt Slack replies.

• 36:10Speaker 00

  Being ignored in Slack is, is, is, not replying in thread. That's another pet peeve I have. You have this issue, Shishir, the not replying in thread. This is Lon. Lon doesn't reply in thread. Look at this. By the way, look, August 10th, you see what's on the screen here, Shishir? Please make sure you have Grammarly Desktop downloaded and installed on your machine. Please take a screenshot and show me here that it's working and you have, have the paid version. I didn't use.

• 36:35Speaker 01

  Yeah. I got, I, you, you like, you like getting responses in Slack quickly. You're cantankerous when you feel like you're not getting prompt Slack replies.

• 36:45Speaker 00

  This holiday season, don't miss your chance to turn just looking into just bought it. Make it easy for your customers to say yes on the spot. PayPal Checkout is a trusted conversion engine fueled by 400 million active PayPal and Venmo accounts. And when you add Pay Later, you turn holiday shoppers into regulars. 96% of Pay Later shoppers are repeat users. Give them Pay Later flexibility now and it could help earn their loyalty all year long. PayPal is built for all business. Get started at paypal.com/business.

• 37:19Speaker 04

  PayPal Later Data, eight markets, March 2025 to February 2026. Repeat users defined as more than one Pay Later purchase in the past year.

• 37:27Speaker 01

  Uh, so my first thought was a lot of the things that we at Launcher are using Grammarly Superhuman for, you, you start to hear them associated with, well, I could, you could do that with Muse, you could do that with Dots, you could do that with Grokbot instead. I could have them proofread everything or whatever. Like, how at Superhuman are you thinking about that? Do you see them as direct rivals? Or are you just, you know, like that? It's a fad. It'll come and go. We're, we're focused on our mission.

• 37:53Speaker 02

  No, no, it's not. I, I don't think it's a fad at all. And I think, and I do think it's a different category. And I think, you know, the way I look at it is, yeah, I think we've developed three main metaphors for AI. Uh, we call it assist chat and do. And they're, you could almost think about them like the human metaphors. Right? So we all started with the middle one with chat. And so ChatGPT came out, you know, uh, you know, uh, ruined or amplified our Thanksgivings four years ago. And, uh, we all fell in love with this ability to talk with this virtual human. So there's a lot of AI products that have a chat metaphor and that, that's very common. Um, there's a set of products that have a do metaphor. I always describe that as like having a task list that you have shared with, with an, with an AI tool. I'd say some of the ones you mentioned, Muse Instinct. So I think that's what, that's what they feel like. Cause I'm, I'm giving

• 38:36Speaker 02

  them tasks. They go off and do the task. I don't care how long it takes them. I don't care how they do it. I just want the task to be done. It's get me the damn reservation, you know, answer this, uh, answer this particular, uh, person, whatever, whatever it is. Um, and so I think that's really important. What we work on primarily, we do all those things. Our agent harness does all those things. But the one we do uniquely well, we call assist. And that feels like a collaborator sitting right next to you everywhere you work. And then, you know, the key thing to know about Grammarly, what's now, now Go. So, you know, Grammarly works in, uh, it's about 40 million or so daily active users. We see a million unique surfaces a day. So we work in every single surface, every web app, desktop app and mobile app that you can work in. Uh, we generate over 125 billion

• 39:19Speaker 02

  LLM queries per week. So if you do, think about that per user, that's 3,000, uh, LLM queries per user per day. So every time, and that's because it works at the speed of your cursor. Right? So as you type anywhere, you could be in Salesforce, you could be in Apple Notes, you could be in Slack, you could be anywhere, you have Grammarly and now Go, watching everything you do, annotating in a way that's unobtrusive to you and to the application and making changes on your behalf. And so that's what we call an assistive agent. So there's, you know, assist, chat and do, I think are three very different metaphors for agents. As far as I know, we're the only ones that really focus on, on that assist framework, at least to that level.

• 40:01Speaker 02

  And I think it's interesting because the volume, if you use Grammarly, we are almost certainly your number one AI tool. We are generating more L on traffic for, on your behalf than any of the other tools. But we're doing it next to you as a collaborator, right next to you. And what we did with Go, we just opened that up and we said, it's no longer just about grammar. You can now write a prompt, you can write, you can connect it to anything and you can deploy it to your entire team. And you can now say, in addition to, you know, J Cal, your, your, your Slack message was, I want a grammar chat. It's like you said, I want a grammar checker sitting next to every single one of you. I paid for the grammar checker, make sure they're sitting next to you. Now you can also hire the fact checker to sit next to everybody and the, and the new idea generator sitting next to everybody.

• 40:46Speaker 02

  And that, you know, and the note taker sitting next to everybody. And I want them each to be sitting right next to you, bringing, bringing that work to where, where you actually work. So I, in my mind, that's the biggest difference of what we're doing.

• 40:58Speaker 00

  Let's get back to product notes here. I, your poor team, you're going to, everybody's working this weekend. I'm just letting everybody at Superhuman know, cancel your plans, J Cal has requests. Everybody is coming in Saturday. Shishir, you're going to get everybody donuts for Saturday? Are you ordering donuts for the team?

• 41:15Speaker 02

  Let's see, let's see what the request is.

• 41:16Speaker 00

  I want everybody in the office this weekend. Here's another thing, you have a keyboard, Grammarly keyboard on iOS. It's incredible for doing Grammarly, but I also have the iOS one. Now that's got approved ratings, yada, yada. And I have Whisperer. So I have three different keyboards that I am rotating between. It's infuriating. Now, when you do dictation on Grammarly, it doesn't do whisper flow, level dictation, it does something else where you kind of describe what you want. And then it gives it to you. It doesn't feel like it's pure dictation. Now, you may not have used this.

• 41:57Speaker 00

  It needs to be pure, yeah, but it's not pure dictation. It's like, I'm want to write something and you kind of, it's doing that help me collaborate feature by default. That's not what people want. People don't want you to write it for them. They want dictation when they're on the go. That's needs to be the default. Defaults matter. So take a look at this keyboard because I think there's a huge opportunity for you to, to get the dictation moment.

• 42:22Speaker 02

  I mean, first off, thanks for the plug on the keyboard and the fact that it includes the voice feature. That's a, a completely magical feature.

• 42:33Speaker 02

  I, it's interesting. I, I think we definitely get feedback on both sides and we'll, we'll definitely get feedback on the side of, I want dictation. We also get feedback on the fact that it feels like an agent that you can direct. And you can say, hey, just, I just want a response that accepts this and get me my latest details and so on. But I think this is more of a case of we're not learning you well. You clearly don't like that and we should learn you. And maybe it's defaults, but it's also the case that you should tell us once, hey, stop that. I just want dictation. And we should, we should adjust. I mean, I think it's one of the things a personalized AI system should be better at. So that's, that's, that's the note I'll take back to the team.

• 43:11Speaker 00

  Or maybe you have two, you have like, one is the dictation button because it's your keyboard. You can put whatever you want on it. You have a dictation button and then you have a question mark over the microphone, which is kind of like ask AI.

• 43:24Speaker 02

  Yeah, it's, it's hard. Multiple buttons on a small keyboard is always hard, but yes, it's the, uh, the, the feature is amazing. I mean, it is, uh, and as that gets connected to the rest of your connectors and so on, you can give it, we, I have a, a demo of it that I, I give all the time. That is, you can now ask that things that are incredible. They, they connect to every surface, uh, every application you have and you can tell it to do anything. But I understand if what you're trying to do is type and/or voice type, you just want something that voice types better than everything else and, uh, feedback heard.

• 43:57Speaker 00

  Also, can we get, uh, the product manager of Slack on next? I'd like, I'd like to abuse them for half an hour with my personal pet peeves of their product. Uh, if you haven't tried Superhuman or Grammarly, where, what are you doing, folks? It's the best. Go, uh, go download it right now. It is a must have for your team. This is the easiest decision you're ever going to make is to have Grammarly and Superhuman running. Because you're going to save minimum two hours a week on email. And at a minimum, uh, your most unprofessional employees will sound professional. And that's probably a third of your employees when it comes to writing. I don't know what we're doing.

• 44:42Speaker 00

  And I honestly, or you, or you won't get fired. I, I just, I cannot take it anymore with the people not being able to write out of school. I am literally now taking this over with my three daughters. I'm like, I, I just don't trust schools to teach writing. I am, I'm, I'm flabbergasted. And I put Grammarly on my 16 year old's computer and I'm like, going through this with her, you have to learn to write acoustically. Great job, Shishir. Okay, everybody go, get Superhuman grammar. Great seeing you, brother. Okay, let's keep going here.

• 45:15Speaker 01

  Great guest, great domain, Superhuman, right?

• 45:18Speaker 00

  As you know, for background, uh, Lon, we were the first investors in Superhuman, the email program, Grammarly was going. Shishir had all these other assets, Coda, they put it all together and they slapped the Superhuman uh, label on the entire suite, which makes sense. It's the best name. But you know, some people have an ego about it.

• 45:37Speaker 01

  I was just going to say, it speaks to the importance of a great domain name. You know, a lot of founders, when they're getting started, they might try to, well, you know, like, here's the name we want. We're going to put try in front or we're going to put get in front. Or we're going to go dot E G C or whatever, you know, to try to get the name they want. And, uh, the importance of a, a poor domain name, it can't be overstated. You don't want it to be messy. And that's why you want to go and get help from our friends, Jason, at GoDaddy. They're going to hook you up.

• 46:05Speaker 00

  GoDaddy is my default. Wait a second, hold on a second. We, we're having a legitimate conversation about domain names and branding.

• 46:12Speaker 01

  Yes.

• 46:13Speaker 00

  Grammarly has been my domain registrar for 20 years, I think.

• 46:20Speaker 01

  GoDaddy.

• 46:20Speaker 00

  I might be 20 years into this. I was using it in a weblogs thing. That was 20 years ago.

• 46:24Speaker 01

  I've been using.

• 46:25Speaker 00

  Are you saying they're partnering and sponsoring the show now?

• 46:27Speaker 01

  I am saying, yes, this is, we're bringing you this show today.

• 46:31Speaker 00

  What could be more natural?

• 46:32Speaker 01

  To our friends at GoDaddy. I'm also, I've been a customer for Lon Harris.com has been under GoDaddy for close to 20 years now, I would say.

• 46:39Speaker 00

  Incredible. So what, uh, what, what, what, what, what, what's the offer here from our friends at GoDaddy?

• 46:45Speaker 01

  We, we, we just want people to go learn more about how they can get the best domain name possible at Godaddy.com.

• 46:53Speaker 00

  That's it. It's a very simple, uh, well, okay. Well, then let me tell you some things about GoDaddy.

• 46:58Speaker 01

  Please.

• 46:59Speaker 00

  Cause I use this product day in and day out. You never want to lose a domain name. Okay. That's key, right? You, you take the time to buy. I have all of my domains restricted so that they cannot be transferred without like multiple steps. You never want your domain name to get hacked. It's an asset, right? So that's one thing. It has incredible security. That's why you use GoDaddy and they will not transfer. Second, you can hide your like domain information and they will become like a proxy for it. So people don't get your address or phone number. So you have anonymity, right? They will shield you from that. That's great. Also, if you want to buy a domain, they have an incredible service for like a hundred bucks or 200 bucks. I don't know what it is. I've used it

• 47:41Speaker 00

  no less than 50 times, I think. We will inquire about a domain name using their domain. They have a great auction for domain names. So there are many different things you can use GoDaddy for. Also they, um, you can buy a domain and just forward it. So disgraziad I have. And I just go into GoDaddy and I forward it as a joke to whoever's my disgraziad of the week, somebody who I'm upset with.

• 48:09Speaker 01

  Sure, yeah.

• 48:09Speaker 00

  That they did something disgraceful. These are all the different uses for GoDaddy.

• 48:15Speaker 01

  There is no promo code, just go to GoDaddy.

• 48:17Speaker 00

  So affordable already. That's the other thing is it's affordable.

• 48:20Speaker 01

  What promo do you, what more promo do you need to use the best domain registrar in history?

• 48:25Speaker 00

  I have asked Lon as part of this program. I was like, hey, you know, we're doing this weekend startups. This is kind of my life's work here. I said, number one. I want the newsletter to give somebody tactical and practical advice every day. Okay, so then we got to what is that? Well, oh, I'm going to summarize the news. Are we gonna talk about deals? So, you know, all this commoditized news. Everything's a commodity now. The news goes by in your social feed. Funding announcements go by. You used to tune into a program like This Week in Startups 15 years ago to get that information. Now all that's commoditized. What's not commoditized today? Actual tactical and practical advice on how to grow your company. That's like hard to come by.

• 49:07Speaker 00

  So I said, Lon, please focus on every day finding the best tactical and practical story of the day. So Lon, tell us about this one here.

• 49:18Speaker 01

  So you can see that here's an example of a post that we did. Jason actually surfaced this tweet from Hector Serrano. He works at Whisperflow. He's a content guy. I think he's the head of content over at Whisperflow. And he released this really great, it went a little viral, actually. A lot of people were sharing this. He released this really good 15-minute discussion about UGC ads. And he really walked people through the, the whole sort of workflow that they use at Whisperflow for bringing in creators, training them to make UGC ads. And then he makes the case for a lot of founders are going to be inspired to try to massively scale up their UGC ad program right away. Like, oh, we brought in five creators. It started to convert pretty well. We got good ROI.

• 50:01Speaker 01

  Let's hire 100 creators tomorrow. And Hector makes a really interesting case for you want a smaller group of creators that you can really mentor, coach, work with carefully and closely. Make sure they're really making a lot of content that expresses the brand's voice and value. And it isn't just like identical, repeated, like a wall of content. And he really lays out the whole case, uses data, makes it interesting and compelling. So that's the kind of stuff that we're looking for. There's a lot of, I would call it like quote unquote inspirational founder content out there where it's like, drink lots of water and remember to meditate and here's like the three mindset.

• 50:38Speaker 00

  And follow your passion.

• 50:39Speaker 01

  Right. There's a lot of that stuff and some of it's useful. I don't want to knock people who post that kind of thing.

• 50:43Speaker 00

  It's not tactical.

• 50:44Speaker 01

  But it's not tactical and it's not practical. We're really looking for what are things I could put into action at my own company today after reading about it here? So like strategies that are more thoughtfully laid out workflows that you could sort of follow. That's the kinds of stuff that I'm looking for. I do, I often try to find things that are from that day or that are going viral that week. But

• 51:07Speaker 00

  Well, here's what's great about this one. This one was going viral, but if you're busy, you might miss it. So what we are is we're the curator, we're the filter for you. If you look at this tweet here, scroll up again. So if you are busy, which all founders are, Lon and my team are now looking for this content. For everyone that we put in the newsletter, and we're doing 250 a year. We are not putting 10 or 20 others. In other words, we've curated it. We've really thought about it. Then we've looked at what they shared and then we vetted it and put context around it so that you can quickly absorb it. So you don't have to do that work. So two levels of work are being done here for you, the founder.

• 51:50Speaker 00

  Or if you work at a big company and you're a growth person or you're a product person. We're trying to provide two levels of service. First, in the selection. So 25 different, you know, tactical and practical inspirational content get considered. The best one gets picked. That would be two hours of work for you, three hours of work for you. So we save you those two or three hours per day. Then when we organize it, we look for all the replies, all the people retweeting, all the people challenging it, all the context around it. That's another two or three hours of work. So each one of these represents, I think, four hours of work is what I would say.

• 52:30Speaker 01

  There's one other, there's one other thing I'd like to add on here. Jacob, pull up that other one that I just sent you with the railway guy. So this is one that's going viral right now, Jason. I found it yesterday. It's taking off today. And he lays out, I mean, it is, this is a really good one. He sort of lays out a lot of very simple techniques that dev tools and other companies that are making products for coders and developers can use to get LLMs to recommend their tool to agents more. Because what's happening so much today is that people aren't going to Google anymore. They're not even going to ChatGPT and asking, hey, what's the best tool to connect my app to this? Or to get my app online quickly, or whatever they want to do. They're going directly to an agent and they're saying, just do this.

• 53:14Speaker 01

  And they let the agent decide. And so what the team from Railway has figured out is how do we get agents to choose Railway more than some of our other competitors automatically? Without a human ever having to prompt it or ask. And he lays out a bunch of these techniques. It's very interesting and cool, but I saw a lot of people are sharing it. Not the original show knuckle up that it was on, but a lot of other people are sharing it. And then they're tagging on their tools and they're using it to promote their tools. So if they have an AI platform or a workflow app or they're doing AEO consulting, they're trying to sort of like

• 53:50Speaker 00

  What is AEO and explain the graphics and the design we did for the newsletter to this?

• 53:55Speaker 01

  AEO is answer engine optimization is what it stands for. It's basically the new version of SEO. SEO was how to optimize what you're building for search engines to find it. AEO is how to optimize what you're building for LLMs and agents to find it. So it's the same concept. And actually, it's interesting. A lot of the rules are sort of similar. But obviously there are other techniques that are just for agents, like the one that, that, you know, like examples that he's using is put all your documentation as markdown files because that's easier for agents to read. So they're more likely to read it than a webpage and a lot of stuff like that. As for the graphics that you asked, so we actually went, or Jason went in Claude Design

• 54:35Speaker 01

  and sort of came up with a granular sort of style guide, an artifact that you created in, in Claude. It was like the, here's what the look and the feel and the visual vibe of the newsletter should be. That doesn't, it, it won't map exactly over Substack because Substack's a little persnickety. They want you to use their fonts. They want you to use their backgrounds and whatever. But as you could see, I was able to come pretty close. I was able to get to

• 55:02Speaker 00

  So what I did was I said to the Claude Design, which is a really good product. I said, I'm building a new newsletter. I explained, I took a conversation I had with Lon from a Slack. I dumped it in there. I said, hey, we're doing this new newsletter. Give me three different looks. The idea being I want to kind of be like somebody's moleskin where they're writing notes. You know, if they were in a meeting or a talk. So here, it's a field guide. Now, when you get a moleskin, you have choices, right Lon? And people who are in the tech business have choices. So I said, hey, give me something where it's a moleskin and whatever. Came up with two or three different designs and one of them was a field guide, field notes and a grid.

• 55:43Speaker 00

  And so I forgot that sometimes when you buy a, a moleskin, you know, a notebook, it has grid paper in it.

• 55:51Speaker 01

  Graph paper. Yeah, yeah.

• 55:52Speaker 00

  Yeah. And graph paper is just lines, which makes it really easy for you to take notes. So then I told it, well, give me, I, I fed it the Substack, Lon. I gave it two or three articles. I said, make me a style guide.

• 56:06Speaker 00

  Which Lon, you will pull up in a second and give it to the team and we'll walk through it. I said, give me a style guide where it's like a profile, notes, a chart and standard things we might do over and over again and make standards for it. And one of them here is it's your next customer asks an agent first. This is an infographic that you might have in the newsletter business or in the magazine business previously. So here it's like, hey, it just walks you through a visualization of what the person Lon found is talking about. So this is another way we're going to provide you value in the newsletter, which is, hey, it says dev asks, how do I deploy this? Agent picks a vendor for them. Then the agent installs the, the MCP and the agent is live.

• 56:49Speaker 00

  And then it has the line through the homepage and it just gives you some notes here. Now, if we go to the next image, I don't know if there's a second image in this newsletter.

• 56:56Speaker 01

  Yeah, there's more.

• 56:58Speaker 00

  There we go. So here it says, is your product agent ready? And it made a checklist because that was one of the things I told it to make was a checklist. Ask the agents who they pick, give agents a door, shorten the path to live, write docs, a model can quote, yada, yada, yada. So we have these field notes now. This is a way for you, if you don't want to read blocks of dense text, to get into the notes. All of this is available for 250 a year, $1 per

• 57:28Speaker 01

  Basically.

• 57:28Speaker 00

  Edition. Yeah. So my concept here is for 250 bucks, I would advise all my founders to do this. That means if we can get to 1,000 people signing up for this, it'll be a slightly profitable business for us. But it'll be like we're collectively sharing the writers here at my company and that This Week in Startups. So the goal is for all of us to share the expense of this. The first 100 people who sign up will get it for a hundred dollars a year. So go sign up, TWIstartups.substack.com if you want this kind of content. And I want to have like a full-time or two full-time people on this. So we're also hiring

• 58:10Speaker 00

  business writers. We want business writers. We're looking for a writer who would obsess about this. Lon and I are building the standard and the templates. But I need a writer who wants to be in our office, work with founders and obsess over these 250 newsletters. So here we are. I'm back in the content game. I'm not trying to make this into a big business, to be honest. I just want to have enough resources here, like one or two full-time writers, pay them decently to have a great job, which is to help founders. If that's you, you can email Lon, L O N at launch.co. [email protected] if you want to write this every day with us. Or we want a person in Austin, sitting with our investment team.

• 58:54Speaker 00

  So that you can be adjacent to all the conversations. We meet with over 10,000 founders. We work with them in our programs. And so I'm trying to get a flywheel started here, Lon. These tactical practical things happen in the newsletter, but we want to invite these people to be guests, producer Jacob, on the program. So the Whisperflow person, who-

• 59:13Speaker 01

  Hector Serrano.

• 59:15Speaker 00

  Is Hector booked yet, Jacob, or not?

• 59:18Speaker 01

  I believe he is. Jacob and I have both met with Hector. Yeah.

• 59:22Speaker 00

  He's booked. Okay. So now people can start to understand the genius of what I'm doing here. This is, I'm trying to teach people management as well. To take people behind the scenes, I wanted these tactical practical talks. So I asked like different people in the organization, hey, can I get some tactical practical talks? And everybody failed. Nobody was doing it. I was losing my mind. So then I just said, Lon, my trusted right-hand guy, here's what I want you to do. Stop doing anything else you're working on. I want you to do this newsletter. It forced Lon to have one a day. The first couple, I'm not going to say were weak, but they weren't as strong as the last couple, which is what you want in a digital product.

• 59:58Speaker 01

  It took a little while to calibrate exactly what we were looking for. And I'll be honest with you, a lot of the legwork of this research is being done by my friend Grokbot. And it took me a while to calibrate him to be able to find the exact kind of content that we need. But the great thing about all of these AI tools that is so still incredible to me, once we'd written a few that we all agreed on and that we liked, just by showing those to Grokbot, that was what it took. And now it understands and can sniff these out much more easily without constant direction from me.

• 1:00:33Speaker 00

  So you have it bringing you those 20 choices a day, or 10 choices a day.

• 1:00:37Speaker 01

  Exactly.

• 1:00:38Speaker 00

  You start there, you start looking for it.

• 1:00:39Speaker 01

  Correct.

• 1:00:40Speaker 00

  So if you have one of these, if you've written one of these, please email Lon. So now what I've done is Lon writes it. That's our first level of vetting of the idea. Y'all give us feedback on it. Then we're going to have the best of those. So out of those five, we do every week. I want the best one, Jacob, to come on the program here and do a video. Then I want the video to go back in the newsletter. So we can ask Lon and I specific follow-ups to the person. Then the best of those, let's say the best one every month of the 20 we do, will come and spend an hour or two with our founders at Founder University, either in Japan, Saudi, or the US, or virtually, or the Launch Accelerator,

• 1:01:20Speaker 00

  and give a bespoke version of the talk where they take questions and answer, and we give our founders a little bit of a tool. So I'm just taking everybody into the mind of a CEO here. I'm thinking like a CEO. Now we've got great content, but I want you all to subscribe to this. And then I came up with a really creative idea. In the newsletter, if you give $1,000 a year as a top contributor, what are we calling it?

• 1:01:48Speaker 01

  A founding member, I believe is what we're calling.

• 1:01:52Speaker 00

  A founding member. So, you know, on Substack, you can pick what level. So he's picked founding member, $1,000. For the entire year, we're going to put at the bottom of every newsletter, founding members. And I think we put one, right? Is there a founding member of Founder University?

• 1:02:07Speaker 01

  I believe, yes. Scroll all the way to the bottom of that post. Let's go take a look. Oh, no, I don't, I don't think I have the footer here.

• 1:02:14Speaker 00

  Okay, go here.

• 1:02:15Speaker 01

  It must be on the email.

• 1:02:16Speaker 00

  A little more, a little more. Okay. So at the bottom of each one of these, it needs to be here.

• 1:02:21Speaker 01

  Okay, I'll figure it out.

• 1:02:22Speaker 00

  We're going to put founding members. The first one is Founder University, right? I paid $1,000. Has Founder University have the link there. So you get 20 characters, and you could just put your link at the bottom of this. So if you think the people are reading this content are people you want to reach, for but $1,000 a year, you can support the newsletter. And, you know, if you've got a product, like a SaaS product, or you're a consultant, you could put yourself into there. If you get 10, one click a month, and you convert of those 12 clicks, you know, one a year to a customer, it pays for the $1,000 50 times over. If you charge $50,000 a year, you get the idea. So we're trying to figure out ways to make the newsletter sustainable, so we can hire great people,

• 1:03:04Speaker 00

  make great content for the small number of people who want this thing, and make it sustainable. That's what I've been thinking about when it comes to humans, Lon. Anybody's going to be able to create these bots. I want to hire humans to be on top of them. I'll talk about our strategy for this week in Venture Capital Newsletter and this week in AI Newsletter on a future episode. But I want to take you behind how I think as a content producer. Lon, any questions here for me?

• 1:03:31Speaker 01

  Uh, no.

• 1:03:32Speaker 00

  Or thoughts on our content creation process? You and I have been working together on content for a long time.

• 1:03:36Speaker 01

  Yeah.

• 1:03:37Speaker 00

  The goal is to get this sustainable.

• 1:03:38Speaker 01

  Right. And yeah, I mean, I feel like we were able to do this pretty quickly because we worked together so many times. And I feel like I have a good sense of where you were going for this. I do also, before we move on, I want to shout out, Jason mentioned the business writer positions. They are up right now. You can email me, [email protected], feel free.

• 1:03:59Speaker 00

  That's what I want people to do.

• 1:04:00Speaker 01

  But you can also go to careers.launch.co. Jacob, let's pull them up. All the business writer jobs that we were just talking about are there, along with all sorts of other potential positions that you can apply for right now. I mean, it is the AI revolution, but we are still hiring humans at launch. We've got everything from senior, that would be sort of 10 years of experience in writing, business journalism, writing about venture capital, writing about finance. And then there's the business writer, that would be more like five to seven years of experience. And then there's junior, if you are just getting started, but you are super, super passionate about tech and startups and entrepreneurship, then we want to hear from you and we want to talk to you about starting your career.

• 1:04:40Speaker 00

  So let me explain the strategy here. I told Lon I want two full-time writers. But because we have myself and Lon, two people with taste, who know how to write, yada, yada, we can get away with two juniors. We could get away with two seniors. We could get away with two business writer analysts, or any combination of these, because we know how to mentor people. So when you're hiring, what I recommend to people is they get dogmatic. We only want a senior person. We only want a junior person. We only want the person in between junior and senior, which is just the title without the words junior and senior. So as a playbook, I'm just giving another tactical practical piece of advice here. What I like to do is have three different positions. So here, like a junior out of school, I think

• 1:05:25Speaker 00

  is a 60K position.

• 1:05:27Speaker 01

  Correct.

• 1:05:28Speaker 00

  And I think the senior writer, we put at 90.

• 1:05:29Speaker 01

  Correct.

• 1:05:30Speaker 00

  And then the business writer analyst, we put at 70, I'm assuming.

• 1:05:33Speaker 01

  I believe it's 75?

• 1:05:34Speaker 00

  75. 75, right. So 60, 75, 90. Now, when somebody applies for this position, when we tell them what it pays, we don't have to have a negotiation over salary. We know which one they applied for. We know by looking at their LinkedIn, how many years of experience they have. We're not going to give people the senior writer position for 90K unless they have 10 years of experience and they applied through that one. So if you applied for the junior, and then you ask for the senior salary, that's not going to work. This is one of the things you have to deal with when hiring is salary bands. So be very specific about it, both internal, external. You know what I always have, Lon, is I have people

• 1:06:16Speaker 00

  on the team say, oh, this person applied, they want 20K more than the job they applied for. And we say, well, wait, why did they apply for the job? No, they applied for that job. That's what that job pays. And, well, it's, everybody takes a shot. I mean, if I was applying for a job, I would say, yeah, I want 10K more than whatever you posted. And I would try to negotiate my best. But then you can say to the person, well, you applied for this position, it had this salary, that's what this job pays.

• 1:06:41Speaker 01

  Yeah.

• 1:06:42Speaker 00

  And I don't think you can ask anymore what you got paid at the last job, depending on what state you're in. So you do need to understand on an HR basis, like when you're hiring for a position now, you can't actually anchor based on what they're making anymore.

• 1:06:55Speaker 01

  Yeah. I don't think you were ever supposed, I was always told not to ask that.

• 1:06:58Speaker 00

  You could, you used to be able to. I think now it's actually forbidden. So this is another thing, like on a tactical practical basis. I had somebody at a company, I'm not going to get more specific than that, who was like, this person has a thick accent. And I was like, whoa, timeout.

• 1:07:18Speaker 01

  You're definitely not allowed to do that.

• 1:07:20Speaker 00

  Well, actually.

• 1:07:22Speaker 01

  Really?

• 1:07:23Speaker 00

  For every one of these hiring positions and what you're allowed to ask for, there's a question you can ask that doesn't break the rules, that some HR specialist has fine-tuned to keep you out of hot water. And then there's the version that gets you in hot water. So I believe, one of them was explaining to me, you can't ask where somebody lives, right? So you can't say like, oh, you live over here. Oh, that's on the other side of the tracks. That's this type of community. That's this level of rent, whatever it is. It could be ethnicity. So this wouldn't apply for business, obviously, but it might apply for, I don't know, you're hiring for retail.

• 1:08:03Speaker 01

  Right.

• 1:08:04Speaker 00

  And you said, I don't want people from Bed-Stuy. I want people from Bay Ridge, Brooklyn. You know, you could, you could kind of sort people based on that. I used to get people who wouldn't, because my last name was very ethnic, Calacanis, they didn't know what that was. They don't know if that's Middle Eastern, Greek, Turkey.

• 1:08:20Speaker 01

  Right, yeah, yeah.

• 1:08:21Speaker 00

  So I would get people who would get them. So I always put Jason McCabe Calacanis, my mother's maiden name, which is my middle name. I always included that in my resume. Because when I was coming into the workforce in the 80s and 90s, people did hire based on ethnicity. I hate to tell you folks this, but Irish people hired Irish people. Italian people would see an Italian name, they would hire it. Yada, yada. So it turns out if the job is to be doing public speaking and you can't communicate, you can hire people based on proficiency in speaking. But you couldn't hire somebody just based on accent.

• 1:09:01Speaker 01

  Yeah.

• 1:09:02Speaker 00

  If that makes sense, right?

• 1:09:03Speaker 01

  If they're clear and they're a good communicator, I don't think you're allowed to be like, well, your voice sounds funny to me. I don't.

• 1:09:10Speaker 00

  You can hire based on ability to communicate.

• 1:09:12Speaker 01

  Right.

• 1:09:13Speaker 00

  Cannot hire based on accent.

• 1:09:15Speaker 01

  If you can't speak clearly and fluently, that's obviously a different situation, for sure.

• 1:09:20Speaker 00

  So literally, if you were to write, hiring this person, if you were to write in your notes, this person's accent, and you use the word accent versus the person's ability to communicate, you know, in a keynote, I would give a seven of 10, like.

• 1:09:39Speaker 01

  You would just say communication.

• 1:09:40Speaker 00

  So these are very nuanced issues, and you need to go through hiring training if you're gonna do this. Anyway, these are just some tips that I would share with people of how I look at hiring folks and also building a team or building an editorial product and why. So it's like, this is sort of my tactical practical. I like the idea of building, in today's world, a human layer on top of the AI, because anybody could create the bot lawn has. So then you have to say, well, what can we do? Well, we can do the infographics, the packaging and the curation. So if the, whatever the AI presents, a human looking at whatever it presented could do four hours of work and make it better.

• 1:10:21Speaker 01

  Yes.

• 1:10:21Speaker 00

  And make it more applicable and ask the questions you might ask when reading.

• 1:10:25Speaker 01

  That's what I always say.

• 1:10:25Speaker 00

  So that's what we're doing.

• 1:10:26Speaker 01

  I feel like all the AI discussions is always this very binary, like use it or don't. Like, if you use it, it makes you dumber. If you don't, you stay smart. And it's like, that's really not the truth at all. It's really, you can choose to use AI in the laziest way possible, just ask Claude for a thing, copy and paste it in and say, that's your work. Or you can collaborate and work alongside AI. And it's really no different from doing research at the library, using Google or any other tool. You don't just take what the AI does and say, this is my work now. You use it as your research level or as your collaboration level or proofreading or knocking ideas around with or digging deeper. It should always be a conversation, never, I ask for a thing, you give me a thing

• 1:11:10Speaker 01

  and then I just take it and post it somewhere.

• 1:11:12Speaker 00

  Not, at least not at this point. And that was like, even with the design of this, I pushed it three or four times and then I had it make a style guide. Then I had it take an existing article and make supporting graphics for it. So when you're making the supporting graphics for it, what is the prompt you give it now? Because you have the artifact of the style guide. So how did you make the most recent ones? Did you tell it, I want this specific graphic? Or did you say, read the article and make me some graphics that's?

• 1:11:38Speaker 01

  As always, with me, it is a collaborative and iterative process. So I give it the rough draft that I wrote. I give it the full Substack page that I'm working on. I give that to Claude and I'd say, don't make anything. Give me your five to seven ideas for graphics or ideas for infographics that you think could accompany this. And then I go through that list myself with my knowledge and taste. And I usually have to tweak them, like it has an idea. And I'm like, that's not gonna, that's gonna be a little thin for an infographic. What if we made it this? Or like, most of that text that you were reading off, like that's written in the blue pencil, that's a collaborative effort between me and Claude. It's making suggestions, I'm changing or tweaking them. And then, you know, after two or three back and forths, we come up with a design that we like.

• 1:12:23Speaker 01

  And that's when I have Claude actually make those graphics that I want. So as always, I think, you need to always be the human in the loop, no matter what the decision is. It's great.

• 1:12:33Speaker 00

  I could tell it wasn't just spitting those out, because I looked at the graphic and I was like, this actually makes sense. And so this is where slop versus tool. Slop versus taste. So you have on one end of the spectrum, you have taste, and the other one, you have slop. Slop is, you didn't review it, you didn't iterate. Taste is, you thought about who's gonna consume it, including yourself. And you said, what is aesthetically or interesting or pleasing to me and/or the reader? And then you gave it iterations to do and said, that's not good enough, let's do that. So because you don't see what's on the cutting room floor, to make anything great in the world, 90% of what you created has to be thrown away.

• 1:13:17Speaker 00

  And you're keeping the top 10%. That's the secret, I think, between slop and no.

• 1:13:22Speaker 01

  Absolutely.

• 1:13:22Speaker 00

  Okay, we got a little bit of a news, let's do one or two news stories.

• 1:13:24Speaker 01

  Yeah, I do, I do wanna talk about this. So Tesla last night unveiled the long awaited next Roadster. Dramatic event at the SpaceX tech site not far from here in McGregor, Texas. The vehicle, as we're seeing on screen, performed a mid-air flip. It jumped over several other cars and trucks from the Tesla fleet and it accelerates from zero to 60 miles per hour in under a second. The car had no driver during this demonstration. Elon Musk said this was be, "To make sure we don't kill anyone," is the quote. He also noted that success was not guaranteed. He ensured the viewers watching this live were at a safe distance in case something went wrong. But it didn't.

• 1:14:04Speaker 01

  The car has 10 nozzles along with an ultra-high-pressure air tank that propels it through the air. That's what you're seeing right there. It also allows it to sort of slow down and reposition itself to come in for a graceful landing. Elon described this as the inevitable result of pairing the world's best car company and the world's best rocket company. Word is that this new Roadster will go into production in 2028. I did wanna bring up this back and forth he had with Marques Brownlee. This was back in 2019, when he was already talking about, the new Roadster is gonna do something like this and it's an image of a flying car. And then Marques responded, "I feel like you're not joking." And Elon said very matter-of-factly, like, "No, I am, I am not joking."

• 1:14:49Speaker 01

  There it is. So he's had this on his mind for a long time, Jason. The last night was a long time coming for Elon Musk.

• 1:14:57Speaker 00

  Yeah, I mean, he left the breadcrumbs, as I said on the pod here. I had seen the demo, he had showed it to me a couple of months ago. And so I've been sitting on this. And when he showed it to me, and when you watch it there, it looks like AI.

• 1:15:11Speaker 01

  It looks fantastic.

• 1:15:11Speaker 00

  That does not look real. And I said to him, is that real? When he showed me a demo run of this, he said, "It's real." I said, "It does not look real." He goes, "I'm gonna do, this is why I need people to come in person."

• 1:15:24Speaker 01

  No, if people weren't physically there.

• 1:15:27Speaker 00

  Yeah.

• 1:15:27Speaker 01

  Show the loop.

• 1:15:28Speaker 00

  So I guess.

• 1:15:29Speaker 01

  Show the loop and flip one, Jacob, while we're talking about this.

• 1:15:32Speaker 00

  And so what that one is doing is, there is suction on the bottom of the car. And the suction on the bottom of the car keeps it gripped to the ground. That's made so when it makes turns. And so here it is, it drives up. And what you'll notice here is, it drives up and it just stops halfway. Then it comes back down. And then I guess this is the one where it actually takes off. And that really doesn't look real. What it does there is.

• 1:15:57Speaker 01

  I saw this 20 times and thought this was, somebody made this with AI before I actually realized this is.

• 1:16:02Speaker 00

  Well, here's what's happening. There are fans on the bottom of it that are sucking it into the ground. So therefore, it doesn't fall.

• 1:16:10Speaker 01

  Right.

• 1:16:10Speaker 00

  And then at the second it's coming off the edge, then the blasters come in and flip it and land it. You would, as impressive as this is, we've been watching him do this with giant rockets for a while now.

• 1:16:22Speaker 01

  Right.

• 1:16:22Speaker 00

  So with accelerometers, I guess, and knowing, you know, how the thing is balanced, quadcopters have been doing this with drones, right? Keeping them perfectly balanced.

• 1:16:32Speaker 01

  Sure.

• 1:16:33Speaker 00

  And if you go and you hit a quadcopter, it rebalances, right? So there are videos of them, people hitting it with a broom handle and it rebalances. Those have existed for 20 years. You have him landing rockets. So just knowing with an accelerometer, like where the thing is in space and then knowing how to make tweaks, whether it's with those propellers on a quadcopter or, you know, whatever it is, a drone, a commercial drone, or with a car with these, you know, different air pumps blowing air. Apparently, we're at the point where our computer can

• 1:17:07Speaker 01

  Yeah.

• 1:17:08Speaker 00

  make it very hard for these things to crash.

• 1:17:10Speaker 01

  Right. It's just like, you know, the space shuttle's gotta hit the atmosphere at like exactly the right angle so it doesn't burn up and they have to know exactly the right place.

• 1:17:19Speaker 00

  You have to work out.

• 1:17:20Speaker 01

  Right, so it's, I guess it's kind of that same technology, but it does seem unreal.

• 1:17:24Speaker 00

  It's actually instructive that you use the space shuttle one. It didn't work out for the space shuttle. I think they, they didn't have the technology to do it properly and those tiles were compromised, etc. So here we are, 40 years later. We're kind of figuring these things out. Now, one, I was thinking, how practical is this? This is not street legal. He made a, he explained to people it's not street legal. It is street legal, I think, to have the vacuum suck it down and have it be able to corner well. So there'll be three versions of this. Everybody wants to know if I'm gonna get one. Of course, I'm gonna get one at some point. It's gonna be available in two more years. This thing's taking forever. I would have rather the Roadster have a two, three, four, and this would be version four.

• 1:18:07Speaker 00

  And I could have been buying a Roadster all this time. But I guess this product is a bit of an obsession for Elon and the team. And they keep wanting to make it look, or make the capabilities amazing. I just wanna drive a two-seater convertible. So can we just get one of these out the door? I like the version he showed. This is a new design as well. The design is gorgeous.

• 1:18:31Speaker 01

  It's very sleek, yeah, it's very futuristic.

• 1:18:33Speaker 00

  Yeah, it's, I mean, the car design is getting incredible. But all I want is, I just want an FSD, two-seater, gorgeous Corvette-like, Ferrari-like car. Don't need it to fly.

• 1:18:46Speaker 01

  Sorry, you're only flying cars here.

• 1:18:49Speaker 00

  Only flying cars now.

• 1:18:51Speaker 01

  How does half the speed of sound? We haven't talked about that one either. He claims it's gonna be able to go at half a Mach.

• 1:18:58Speaker 00

  Yes, which I think you, as he joked, it's gonna rip your retina off.

• 1:19:02Speaker 01

  Yeah.

• 1:19:02Speaker 00

  Already, I set all my Teslas. So we have a couple of them that are like the high-powered ones, I think, in the fleet of Teslas I have. And we put them all on chill. Why? People in the backseat, the kids can get carsick. These things are so good. So I always downgrade. I pay for the fastest version of the Tesla and I purposely downgrade it to chill mode. So that people in the car don't get carsick.

• 1:19:28Speaker 01

  It's intense.

• 1:19:29Speaker 00

  So that's the other thing. These cars are now so fast, the pickup is so ridiculous in an electric car. But I'll be honest, the only thing that matters right now in car companies is FSD. Being able to check your phone, being able to check the Nick score, do a text, change a podcast. Whatever it is, that's more important to people than anything is having the car drive other places. What this really symbolizes more than anything is the merger of Tesla and SpaceX, which I don't have any inside information on. But in my newsfeed, it said it's gonna be announced and done by the end of this year.

• 1:20:12Speaker 00

  Which would lead somebody to believe, should you buy Tesla? Or should you buy SpaceX shares if they're gonna merge? Is there some arbitrage that's gonna happen there? Either way, it's gonna simplify Elon's life to have it be one company. Cause right now, there's the fab that I think the two companies are collaborating on. This is a collab between the two companies. And then they have tons of data centers for Tesla. And Tesla's doing real world AI. Then you have Grok doing LLMs and Optimus is at Tesla. But then you have Grok as part of SpaceX. SpaceX does manufacturing. All this stuff has to be under the ticker symbol E L O N. And that's the next shoe to drop from what I read online. I don't have any inside information.

• 1:20:57Speaker 00

  Yeah. Like I did for having seen this demo.

• 1:21:00Speaker 01

  Just like Stark Industries, Musk Industries. I feel like it's.

• 1:21:04Speaker 00

  That's the obvious next thing. Alright, what else we got on the agenda here?

• 1:21:07Speaker 01

  Alright, well, we should.

• 1:21:08Speaker 00

  There was a couple of big news stories I didn't get a chance to comment on.

• 1:21:10Speaker 01

  Yeah, we should. Oh, well, we could talk about Muse Spark Open. We could talk about some of the ones that you didn't get to talk about, like there's the ChatGPT lawsuits. There's the bugging, the spying story.

• 1:21:22Speaker 00

  Oh, that's an interesting.

• 1:21:23Speaker 01

  Yeah, let's do that one.

• 1:21:24Speaker 00

  Those last two sound good.

• 1:21:25Speaker 01

  Earlier this week, the FBI arrested OpenAI engineer Abdulrahman Al Fazan for allegedly hiding a recording device in an executive conference room. Prosecutors say the bug was wrapped in toilet paper behind a TV. And instructions for designing it were found on his wife's ChatGPT account. Al Fazan is a Saudi national. Prosecutors say he had booked a one-way flight back to Saudi Arabia for October 12th. The Wall Street Journal reports that OpenAI was already training staff to spot honeypots and spies who use romance, Jason, to get their secrets. They are told not to wear their OpenAI gear in public and to be wary of friendly strangers. I mean, if you work at OpenAI and someone comes up to you in a bar and they're being a little too friendly to you,

• 1:22:09Speaker 01

  should you now presume that person is an international spy trying to get at your models?

• 1:22:15Speaker 00

  I have been in this situation myself over the last 10 years. Where people have shown more interest in me than obviously I deserve. And this could be me being delusional or not. But there was an instance when I was overseas in another country where somebody who was very clearly a Russian, by their accent and a Russian, very clearly like, you know, many inches taller than me, much more attractive than I am, was showing way too much interest in me than. And I'm a married guy, like I'm not interested, obviously, but you know, not exactly proposition, but everything up to that.

• 1:22:57Speaker 00

  Wanting to spend time, asking to have a drink, asking, you know, where I was staying. It was just obvious to me. I've got my guard up for this kind of thing, just naturally. As somebody who's a target, you know, and anybody in the tech industry who has access to these models, if they are gonna change the world, then obviously you're a target as well. This kind of training's been going on for a long time. They have really good counterintelligence, I think, going on at OpenAI. Because they put a camera, once they found this device, they left it. Cause they knew the person would retrieve it. They put a camera behind the TV and got an actual shot of the person, which you can pull up.

• 1:23:39Speaker 00

  There's a picture of the person going behind the TV set in the boardroom to go collect the device with a shocking look on their face, which is either them just being guilty by default. Hey, guys, you still have this on, the same thing on the screen, but I asked you to reference the other thing. So just take that one down and go do the other one. So you'll see this picture of the person where they're ready to go, like to retrieve this. The countermeasure, they put a video camera or a tiny pin camera to get the person on camera retrieving it. I don't know that this Saudi person is doing this on behalf of the Saudi nationals or behalf.

• 1:24:19Speaker 00

  I think this is on behalf of people who are investors in the company or who are trading the stock. I don't think this is international espionage. I think this is financial or corporate espionage, which is slightly different.

• 1:24:31Speaker 01

  It's interesting. I mean, that was my, that was gonna be my first question to you was, what do you think they're actually trying to get on tape in this OpenAI conference room? That, you know, would be so valuable that it's worth sending this guy in there?

• 1:24:46Speaker 00

  I think this guy worked at a venture firm, according to reports. I think people are trading the stock.

• 1:24:53Speaker 01

  Oh.

• 1:24:53Speaker 00

  If I'm being honest.

• 1:24:54Speaker 01

  I see.

• 1:24:55Speaker 00

  So if you were trying to make money and you were putting a billion dollars or $10 billion into stock. I'm guessing here, but I don't know that in the conference room, the new models coming out is something that the Saudi government needs or the UAE would need or the UK would need. I don't think a government needs to know this kind of stuff. What they, somebody who was trading shares in the company would need to know.

• 1:25:20Speaker 01

  Right. That's it. That makes more sense. Cause I was like, you'd want the documentation or something if you were trying to like distill the model. Like what's a meeting gonna do for you?

• 1:25:29Speaker 00

  Nothing. The meeting's gonna do nothing. What you would wanna do is compromise somebody's laptop who was pushing code. That would be the thing you would want. But either way, this keeps happening at big companies, Chinese, Russian, you know, Saudi, whatever. And by the way, we're doing 100 times this on the offensive. So you can be certain at DeepSeek, you know, and Alibaba and TikTok, ByteDance. We have our plants in those places doing the exact same thing here. So, but also like the, this feels very amateurish. The person's spouse was looking this up on their OpenAI account and they.

• 1:26:13Speaker 01

  Yeah, how to look, how to make a recording device.

• 1:26:16Speaker 00

  So they didn't need to get a judge's, you know, information, a judge's warrant to go look at her, the spouse's OpenAI account. They just were like, well, who's this person's spouse? Who's around them? The OpenAI security team was able to look at their own OpenAI chat logs and say, well, who's the person's spouse? Let's look at what they're searching for. So this should let you know, if you're putting anything into OpenAI, Claude or Gemini or any of these services, it's recorded.

• 1:26:44Speaker 01

  Oh, yeah. Everybody should really know that. Like, I know that a lot of, you'll hear people say like, they're kind of freaked out about everything OpenAI or knows about them. Or Claude remembered that I went on vacations. Like they know everything, everything you enter in there.

• 1:26:59Speaker 00

  It's not only that, it's recorded.

• 1:27:00Speaker 01

  Yeah.

• 1:27:01Speaker 00

  The whole chat transcript is there and there's no way, I think, to get rid of it. I'll be honest. I think they just store everything for all time.

• 1:27:07Speaker 01

  I'm sure.

• 1:27:08Speaker 00

  Anyway, it's like an interesting story. It's interesting. They caught them. Yeah.

• 1:27:14Speaker 01

  It has got me thinking about my own security. It is a funny picture too. This guy like, oh, he looks like he knows he's been busted doing corporate espionage.

• 1:27:22Speaker 00

  Well, I think when he was looking in there at it, I think he might've seen the camera.

• 1:27:26Speaker 01

  Right. He like.

• 1:27:27Speaker 00

  That's what I'm wondering is if he saw the camera looking at himself.

• 1:27:29Speaker 01

  And he's realizing like, oh no.

• 1:27:31Speaker 00

  If they put a pinhole camera in there.

• 1:27:33Speaker 01

  I should have.

• 1:27:34Speaker 00

  Did he see it? But he's looking at something else, but, or maybe he's just really nervous. Like, I would be.

• 1:27:39Speaker 01

  I would be, yeah. I would be if I was doing corporate espionage too, yeah.

• 1:27:44Speaker 00

  Crazy. I think somebody's making a trade. That's my only guess.

• 1:27:47Speaker 01

  That's no, that makes a lot of that. That would be the sort of thing you would actually overhear in that room as opposed to like, how are they making their models? Like, they're not gonna sit there and

• 1:27:55Speaker 00

  Who's buying shares? When are they going public? What's in the S1?

• 1:27:59Speaker 01

  Trying to get a little insider.

• 1:28:01Speaker 00

  Is discussed in that room.

• 1:28:02Speaker 01

  Yeah.

• 1:28:03Speaker 00

  Yeah. Okay. There you have it, folks. All right. What else? Anything else on the docket I should?

• 1:28:07Speaker 01

  Yeah, we'll do, we'll do this one more. Eight states are suing OpenAI, speaking of House of Altman. They've each filed separately. They're arguing that the chatbot has insufficient safeguards to protect children and other vulnerable users. Seniors also get a shout out there. The states, Arkansas, Arizona, Idaho, Montana, Iowa, Kansas, West Virginia, and Nebraska brought the cases. The suits say OpenAI is marketing ChatGPT as safe for kids under 13, and that it plainly is not. Some examples in the suits, ChatGPT, they accuse of coaching kids to commit suicide or to harm others, and showing flagrant disregard, that's a quote, for minor safety. Here's Iowa AG Brenna Bird.

• 1:28:50Speaker 01

  The company told users all across the country that ChatGPT is safe. It is not. Brenna and the other AGs want OpenAI to disgorge money they earned while minors are being harmed, and stop making deceptive safety claims. There are no set damages. Numbers, there aren't numbers in play yet. OpenAI responded that it has built age-specific protections, and wants rules that will apply to the whole industry, not just one company.

• 1:29:16Speaker 00

  Yeah, Character AI just went out of business as well. I thought you would have that on the docket. That company

• 1:29:24Speaker 00

  was targeting users with characters. It has been very controversial in schools. And I think it's abhorrent to put kids on these tools if they don't have massive controls in place. It's kind of like YouTube. I was, you know, one of my daughters using YouTube many years ago, and it was on, they were doing Disney, doing the Disney princess phase of their childhood. And then all of a sudden, I see these Disney princesses smoking cigarettes, and cursing, and, you know, being completely inappropriate. And somebody had made a parody of Disney princesses gone wrong. Put it on YouTube, which is completely valid for adults as a form of entertainment, I guess.

• 1:30:06Speaker 00

  But not for kids. So they came up with YouTube Kids. There needs to be ChatGPT for kids, obviously. And Character AI, I think, doing parasocial relationships like this? Is it parasocial the right term?

• 1:30:20Speaker 01

  Parasocial would be like

• 1:30:21Speaker 00

  For this type of relationship?

• 1:30:22Speaker 01

  The people who watch this podcast and they feel like they are our friends and we all know each other.

• 1:30:27Speaker 00

  That would be parasocial.

• 1:30:28Speaker 01

  That parasocial thing. And then if we were

• 1:30:30Speaker 00

  What's the term for parasocial with an AI?

• 1:30:33Speaker 01

  That's a great question. I don't know if we have one, but we should.

• 1:30:36Speaker 00

  We need a term for parasocial relationships with an AI. But let's call it AI relationships. I think AI relationships are evil. I believe making them and giving them to kids is, like, high-level evil. I think making them and giving them to adults is abhorrent, on my scale. Like, I don't think that should be a business we should pursue at this point in time. Humans should have relationships with other humans. Building these tools to, like, give them to humans instead of having human relationships, and trying to promote the her-like relationship from the film Her, or any of these things, I feel it's abhorrent at a minimum. And with kids, or like seniors, it might even be evil.

• 1:31:22Speaker 00

  I, I just see very little upside to it.

• 1:31:26Speaker 01

  We've been talking about this a lot through, like, sort of the whole AI era. I, I feel like this is actually underlines a lot of why Americans are so distrustful towards AI. The original things that we were promoting AI for was like, you can, it can be your boyfriend or your girlfriend, or like, it's gonna be your therapist, like, the worst stuff. Like, the stuff that it's not actually good at. There's a million, we talk about it on the show all day, every day. There's a million things that AI is good at. It's a great researcher, it's a great writing coach, it's great, it's great at taking all this data, like, I use it for my, you know, streaming calendar, it tells me what to watch on Netflix, but it's not a good therapist, and it's not a good friend. And we sold Americans on that, like, first. And so now I get why some people don't trust it,

• 1:32:10Speaker 01

  even when it's like Notebook LM, and it's obviously very useful, because that's what we originally marketed it to, it was like, well, it's gonna be your new best friend. And it's like, it's not a good best friend. It doesn't

• 1:32:20Speaker 00

  Yeah, AI companionship, no good.

• 1:32:22Speaker 01

  I don't think it knows how to be fair.

• 1:32:23Speaker 00

  Artificial intimacy, AI attachment, very bad. I think it's all very bad. I, I think these attorney generals have a point. I don't think this stuff should be used by kids. It's on OpenAI to make sure kids aren't using it. And they should just have a kid's version that is severely gated.

• 1:32:43Speaker 01

  We talked to, Sue Kim is her name from that company, Brilliant, and they have made like the

• 1:32:47Speaker 00

  Yes, we're investors.

• 1:32:48Speaker 01

  They made the AI tutor. I didn't even know you were an investor. There you go, I'm bringing it up anyway. That, that should be how ChatGPT for kids automatically works. Like, it doesn't give you the answer. If you're working on homework or something with it, it just automatically reverts to like a tutor, kind of like walk you through it, help educate you. Like, as soon as a kid logs in, we know they're in ChatGPT Kids, we know how we want it to interact with them. It should just obviously be like that.

• 1:33:15Speaker 00

  There you have it, folks. Another This Week in Startups is in the can. We'll see you next week. Bye-bye.