Cliff Weitzman：从全球100位顶尖CEO身上学到什么，以及为什么 Token 支出将超过薪资

1. Applying to 26 Colleges: Cliff's Volume of Work Philosophy

Ready to go? Cliff, I normally ask for great suggestions beforehand, and honestly, most of the time they’re like 6 out of 10, and I actually have to get really creative. I have to use so many references, and I love that. But then you sent me yours, and I was like, “Oh my God.” I actually sent them to the team, and I was like, “This is wild.” So, we have some of the most insane stories coming up.

First, thank you for making my life easier with the schedule. Second, thank you so much for joining me in person.

My pleasure. It’s great to be here.

2. Buying a 3x Leveraged Nvidia Option in 2022 Before the Market Caught On

Now, I do just want to start. At university, you applied to 26 colleges when everyone else applied to 8. Can you talk to me about this and the strategy behind that?

Yeah. I moved to the US when I was 13. I didn’t speak English, and my parents knew about 3 schools: Harvard, MIT, and Stanford. We had a college counselor in my high school, and she recommended applying to 6 schools. I was like, “Okay.” So, I started filling out the applications, and quickly I realized that even for kids with perfect SAT scores and perfect grades, you have about a 1-in-10 chance of getting in.

I was like, “Okay, it’s a lottery.” How do you win a lottery? You buy more tickets than everybody else. Most kids in my school didn’t need to do 7th period when they were a senior, but I had academic workshop, which is basically the class you go to if you’re on the short bus. I went to academic workshop every single day because I just couldn’t read very well.

I made it my goal, my job: instead of leaving school, I’d go back to academic workshop and apply to colleges. When I finished the normal application, I just kept going. So, for my main essay, instead of making 3 variations or 5 variations, I wrote 48 drafts. Then I had to make a second Common App account because I ended up applying to not 20 schools, which is the max, but 26.

In the end, I didn’t get into many schools that I wanted, but I got into a lot of schools I didn’t expect to get into. My overall principle in life is volume of work: more reps, more shots at goal, more free throws. That’s how you get unexpected outcomes on the bell curve.

One of my favorite sentences is actually from Alex Hormozi. I don’t know if you know Alex Hormozi, but he says, “Volume times leverage equals output.” I really think about that. I used to send hundreds of emails to sponsors for the show and had zero leverage. Now, with more leverage, the output is significantly more, but I always really think about that.

3. Building Speechify: Dyslexia, Deep Learning & 10 Million Books a Year

I was talking about this show with my girlfriend last night, and she was saying, “Did you create a text-to-speech app because it took you a long time to read?”

Yes. I’m super dyslexic. In 1st, 2nd, 3rd, and 4th grade, I couldn’t figure out how to read. I wanted to read fast, but I couldn’t read fast. So, we built a deep-learning-based text-to-speech model in 2015 that now reads more than 10 million books’ worth of words to people every year.

Wow.

So now, if you want to read fast but you can’t, you can read fast.

So, you’re a great guy.

Very great guy.

Great guy. And this was apparently the quote in the yearbooks of 40 of my 14-year-old friends in freshman year. That is what you wrote. How did that lead to Speechify?

Yeah. When I was 14 years old, freshman year of high school, I was very self-conscious about the fact that I was very bad at spelling. You get a yearbook, and typically people will sign the back of the yearbook and write some nice message about you. I didn’t want to misspell stuff in everybody’s yearbooks, so I came up with a strategy: I would write the least number of words possible, and I would write the same thing in everyone’s yearbooks.

“David, you are a great guy. Steve, you are a great guy. Bill, you are a great guy.” Then, after the last day of school, we went to a sleepover at one of my friends’ houses. He went upstairs to show his mom his yearbook, and I heard him scream from the top of the stairs, “What the hell, Cliff? I’m not gay?” I was like, “Oh my God, what have I done?”

The crazy thing is, 8 years later, I was graduating college, and I wrote a 30-page paper for one of my philosophy classes. I submitted it to a friend before fully submitting it, and they sent it back to me. I had 3 spelling mistakes, and I cried because, if you had told 14-year-old Cliff that he’d go to college, he wouldn’t have been sure about it. He definitely wouldn’t have expected himself to go to Brown, but he for sure would never have expected to be able to write a 30-page paper with 3 spelling mistakes.

The reason why I didn’t have that many is because I can hear the spelling mistakes with Speechify.

Dude, I was listening to you and Ali Abdaal. Great show. I heard that, in 2020, you made a list of the top 100 consumer subscription companies in the world by revenue, flew all over the world to meet the CEOs of each one of them, and then tried to learn from them.

4. Flying Around the World to Meet 100 Consumer Subscription CEOs

I really enjoyed the show, so, with the greatest of respect, I was kind of frustrated listening because you said that, and I was like, “Amazing. What did you learn?” We didn’t go into it in that conversation. What did you learn from flying around the world and meeting the top 100 consumer subscription companies?

Sure.

What did you learn?

So, first, basically, at every point in Speechify’s history, I have led every department except for AI: design, engineering, QA, and growth. My rule is that I will read 100 books about the topic, listen to them with Speechify, identify 100 experts, and go and have a conversation with them.

The first thing that I learned is that, if you send a good email, people will respond, no matter how amazing they are. I did Mike Krieger from Instagram, Evan Williams from Twitter, the founders of Plaid, the founders of 23andMe, the founders of Honey, the founders of Grammarly, and Vlad from Robinhood. All of those people became Speechify investors very early on, and they would all respond to the emails.

The second rule was, again, volume of work. If the CEO didn’t respond, I would message the CMO. If the CMO didn’t respond, I would message the head of growth. If they didn’t respond to my email, I would politely continue doing so until they either told me to stop—which happened less than 1% of the time—or agreed to go on a chat. I would also message them on Instagram, and I would ask people for introductions. I was just unlimitedly hungry to get the information.

The second thing I learned is that, once we would get on a Zoom call, we’d chat. There was one thing that always succeeded: I have listened to 100 books a year every year for the last 15 years, so I’d ask them, “What are your favorite books?” They’d tell me, and most likely I’d read at least one of them. So, we’d connect on the book.

Then I’d be like, “Oh my gosh, I’m going to be in Denmark this Sunday. Do you want to hang out?” if they were in Denmark. They’d be like, “Oh, sure.” Then I’d book a flight to Denmark and show up. I did this with Anton at Lovable recently. I’m still doing the same thing.

After I did this with the heads of growth, I did it with CTOs, and I did it with every type of person that I wanted to learn from. Then I’d go and watch how they bought Facebook ads, or I’d watch how they made content. From that, I figured out what would work.

The first thing that I learned is that growth is just an arbitrage game. You are competing with every single other person in the world who wants to get their product in front of users. That means your leverage points are either the content that you make or how you distribute that content.

When it comes to buying ads, there are always new tricks that you can come up with. At the time, adding polls to your Instagram Story was really helpful. Right now, whitelisting ads is really helpful. Volume of work—posting a ton of ads that you test—is really, really important.

There are different rules for SEO. There are different rules for how you structure your team when it comes to who the creative director is, who writes the scripts, how you do the editing, and so on. Every team does its own thing, but for every company there’s a different rule.

One person I learned from is the founder of Blinkist. A small thing he said as an aside that really helped me at the time was, “Don’t even bother spending money on any platform that’s not Meta until you reach $100,000 a month in spend on Meta.” Then I stopped spending on other places.

I also realized that a lot of the people who were senior in these companies were already rusty. They were not very strong individual practitioners of the craft. So, I would go down a couple of levels to find the person who was actually buying the ads, who was actually editing the content, and who was actually seeing what the CTA was and what the things were.

I made sure that everybody in our company, no matter how senior they got, was still not just a fat general sitting in the back, but a warrior who could take out their sword and actually buy the ads hands-on with a keyboard.

We had a situation where we hired a head of growth. 30 days in: nothing. 60 days in: nothing. I was like, “All right, man, what’s going on?” He’s like, “Well, we’re trying to hire the people.” I was like, “Okay, tell me what you’re doing for hiring.” “Well, I gave the head of HR the brief of what I’m looking for.” Bro, source the people yourself. Message the people yourself. Get on the calls yourself. You’ve got 7 days.

In those 7 days, I need to see you making ads in CapCut. I need to see you editing ads in Figma or Photoshop, whatever tool you want. I need to see you buying the ads. And if that doesn’t happen, it’s going to be very difficult to keep working together.

It reminds me very much of Keith Rabois: you want to have barrels, not ammunition. You want someone who can go from 1 to 10 on their own. They do the full stack. I don’t want, “I can put the video on CapCut, but I can’t do the editing. Someone else needs to do that.” It’s like, “I can take the video, I can edit it, and I can post it. I do all of it.”

We look for 3 things when we hire people: fire in the belly for the product, high loyalty to the team, and the ability to learn fast and ship to production. We have a principle even inside of engineering, which is that you’ve got to respond to messages within 60 seconds because you want to unblock people. Just inform them if you’re able to unblock them or not. If you can’t, that’s fine. They’re going to go learn it on their own.

We have a rule: if you ship a PR to a repo that your team is not responsible for, it gets fast-tracked, and that team has to respond very quickly because I want to encourage people to learn how to contribute to other codebases. The person in the company who learns the fastest is my brother Tyler, who leads our AI team. The whole principle inside of our AI engineering team is: just learn everything, and then you win.

5. Companies Have Bulking and Cutting Cycles

Okay, there are many things that I want to unpack here. You can tell from my incredibly athletic physique that I am a fan of weights and gym work. You said before that growth is about bulking and cutting cycles. I love this so much. Can you explain how you think about that?

Companies go through cutting and bulking cycles like bodybuilders do. There’s a reason why cutting and bulking makes sense: you can’t oscillate between the 2 in 1 week. You have to commit for 6 months to doing 1.

For Speechify, for example, it took us 4½ years to find real product-market fit. The product stayed exactly the same. The vision today is very much the same as it was in the very beginning. But in 2020, we started growing really fast. That was around the time when install-to-trial, trial-to-paid, and renewals all went up. People used the product a ton.

I tried to figure out how to pour fuel on the flame. The reason why companies succeed is that you have a person who puts their beating heart into the product and their will to power. That’s the thing that makes the thing successful. You can only do 1 of those things at a time.

The second you split your attention—even fundraising—you can’t fundraise and do sales at the same time. You’ve got to do 1. All I was thinking about at the time was growth, which is why it was a no-brainer to fly to all these places to meet people.

There was a point where I was actually very annoyed because I didn’t succeed in having the growth outcomes that I wanted. I started logging my time in 30-minute increments. Every day I’d have 48 units. I noticed that 30% of my time I spent on helping other people, and 30% of my time I spent learning SEO. So I decided I’m not allowed to learn anything from that point on. I’m only allowed to execute.

The second I saw myself reading an article, I would close the article and go do. The rule is you just have to go, and you have to learn, and you have to execute. You have to have that maniacal focus on the 1 thing. Then we grew a lot.

Now there was a focus on, “Okay, well, we’ve got to get the margins to be good.” That’s a cutting year, and there you’re focused on profits. The conclusion that I’ve had across the almost decade I’ve spent working on Speechify is that any Harvard MBA can cut costs. If you’re smart, you could do it. It takes a genius to grow revenue.

That’s why investors value exponential growth of revenue, even for companies that are not making a profit. But it’s very difficult to do both at the same time. You’re clicking the gas and clicking the brake at the same time.

Speechify just finished 4½ years of running profitably, and now we’re in a hypergrowth phase. It doesn’t matter whether you’re a company, an athlete, or a social media creator: you have bulking years and you have cutting years, and you need to decide which year it is for you.

In a hypergrowth phase, are we fully accepting that CAC will go up, and we’re happy with that?

No, not at all. If you think about CAC, you have to consider, “Okay, what is my LTV, and what’s the first payment period?” We like to think about the first payment period. Our product is $140.

I’m never willing to burn CAC for no reason. The blended CAC can go up, not the direct CAC. What does that mean? I’m okay spending $500,000 a month making 20,000 pieces of creative because that is a creative investment. I’m going to learn a format that is new that will unlock additional growth later.

I’m okay doing an experiment on Apple 11 or a specific new product that Google dropped for YouTube videos. We’re 1 of the 200 companies that’s provisioned to test ads on OpenAI. So we’re spending a ton of money on OpenAI right now, and we’re learning it faster than everybody else. By the time it rolls out to everybody else, we’ll already be really, really good at running ads on OpenAI.

You grow, you grow, you grow. You test new channels, you test new creative. I’m very happy to burn an unlimited amount of money on excellent engineers. I’m not okay just setting money on fire doing ads that are not attributable, spending money on a billboard where I don’t know whether a user comes from it.

6. How Speechify Thinks About Paid Ads & CAC

I think about everything as a test, and if the test succeeds, then I do more. If the test fails, that’s fine. The classic adage that you learn from CMOs—I learned this from Kelly at Netflix—is that they spent $1.7 billion on performance marketing in 2017. You’ll hear that Brex is also really good at this: “I know I’m wasting half my spend. I just don’t know which half.”

That’s okay. Do that, and then do the work to figure out which half you’re spending. But don’t just spend with increased CAC for the sake of growth. It always has to be an experiment you learn something from.

When you look back at the consumer subscription companies and their ad-buying patterns—you said, “I want to go in and I want to see how they buy ads”—who had the most atypical, weird process that stood out?

Greg at Ladder. Ladder is a health and fitness company. Perfect example: he’s the CEO of the company, but he buys the ads himself, and he found a lot of arbitrage plays that other people have not found. You just have to go and figure out what those plays are.

Do you think there are new plays to be figured out today?

Every single day. Every single day.

What is the play today that people don’t know?

Whitelisting ads on Instagram works really well.

So, what is whitelisting ads?

Let’s say you find a creator who is in your niche. For us, maybe it’s teachers. Maybe it’s someone with ADHD or dyslexia. Maybe it’s an AI creator. They make a video for you, and then you go, “Hey, make me 20 videos. I don’t need you to post them on your profile. I’ll run them as ads myself. If some of them are good, we can post them from your account and increase the spend.”

Users might not see it as a normal organic video, but you get to test all of those videos, see which one performs, and then have them post organically the one that actually is successful.

Is consistency not important? What I mean by consistency is consistency of personality.

Not at all, because you’re marketing for different demographics. You find this interesting: we test almost 1,000 AI-generated ads a day right now, on top of the roughly 8,000 organic creatives we make with humans every month.

One of the things we do is reskin. My sister Geffen leads our ads team, and we have a bunch of great ads with Geffen. Then we’ll turn her into a 60-year-old lady. We’ll turn her into a 40-year-old man. Then we’ll turn her into someone who’s African American or someone who’s Asian.

We’ll change the background from a coffee shop to a library to a farm, but a farm wouldn’t convert for our users. You just test all the variations. I know that for us, the highest-converting demographics would be people in their 30s and 40s in terms of paid, but we have crazy usage in the 18- to 28-year-old market.

People who are 60 and above need reading glasses, so they use Speechify a ton. Now we’re launching voice agents at symbol.voices.ai, and so that’s a different demographic.

All the initial ads for Speechify were me. It was my face. I would sit in my house, especially on Saturday or Sunday, with my phone and record. I’d be like, “As a medical student, I use Speechify to read while I’m working out. As a lawyer, I love working from home. You just need to read that contract and find that 1 gotcha, and Speechify can find that for you. You just need to listen asynchronously.”

I have dyslexia, and I would just play all the characters. It was my face, and it converted really, really well.

But I got maxed out. When Geffen joined, we hired other people to the team, and suddenly we found that women convert a lot better than men as audiences. Right now, I would say we have more women who use the product than men by a good margin. It’s just because our team happens to be very woman-focused. You just have to test everything.

On the video, before we move on, what do you use? I see everyone saying, “Oh, we do AI video.” We’ve seen a lot of that. Do you remember Icon, the Peter Thiel-funded one that said it did AI ads? What do you use for your AI ads?

Yeah. Because it’s an arbitrage game, if you use the tool that everybody else uses, typically you’re not going to win. So we just built our own platform.

There was a point several months ago where my girlfriend went to Disneyland, so I had 4 days free. In that time, I went to Skool—Alex Hormozi, shout-out—and I did a course for AI-generated video ads. Then I hired everybody from inside the community who was really good. So I built a team.

I hired 4 new engineers outside of our core team. In 1 week, I hired 10 people. Twenty-four seven, all I was thinking about was AI-generated content, and my goal was to get to 1,300 ads that we tested every single day.

We just built tools for it. Initially, I built it on n8n, and then all the n8n instances would time out, so we built our own platform and our own website that does really good AI-generated stuff and allows you to reskin. It automatically posts to Meta, TikTok, YouTube, and all the platforms, and it gives me all the data that I want.

I’m obsessed with Manus, so I’d get reporting for Manus every single day. After those 4 days, I continued to monitor it, but I assigned a leader for that project, and that project has done really well.

We don’t use any external tool. Obviously, we’ll use all the LLMs, but that changes on a weekly basis, right? likely Wan 2 was really good, and then Kling was really good, and then likely Seedance was really good, and now ChatGPT’s new image model. Then likely Nano Banana, obviously. You just constantly test and iterate, and you find flows that work. So we just built our own.

1,300 ads a day. A day. That’s wild. What do we look for in the outcomes? At the end of the day, I work for Speechify. I go, “Okay, Cliff, here’s 1,300 ads. What now?”

What you want to do is throw a lot of things at the wall. No matter how smart you are, you’re not going to know ahead of time which thing is going to convert.

You really don’t.

You have no idea. So first, you test. Once you test, you look for the following: in Meta, you can see the click-through rate on the ads, and you can see the CPA, cost per action, and learn what the CPM, or cost per 1,000 impressions, is.

If we see that the cost of acquiring a user and having them pay is less than a certain amount, we’re like, “Oh, very interesting. This is a standard deviation better than everything else. Let me put it into the main campaign. Now it’s going to compete against our best-performing ads ever.”

We let it breathe. If it does well, it gets more spend, and if it does less well, it gets less spend and we cut it. So it’s like evolution. You put it into the wild and see if it performs.

And so every day, you’ll take the top 10, the top 20?

We upload 1,300 to our testing campaigns. Of those, we’ll take the ones that win. It might be that nothing wins that day, in which case nothing gets launched. It might be that a lot wins that day, in which case they all get launched.

It’s like March Madness brackets. You just graduate to the next level of competition, where you can get more spend. It’s almost like duplicating the algorithm that TikTok and Instagram have anyway: when you post a video, 1,000 people see it, and then if it does well, 10,000 people see it, and if it does well, 100,000 people see it, and if it does well, 10 million people will see it. It’s the same thing with performance marketing ads, except you’re paying for every view.

Totally get that. You said you use Manus a lot.

A lot.

I’m just intrigued. I don’t know if you saw, actually, just this morning, the acquisition was blocked by China. That’s interesting. What do you use Manus for?

That is interesting.

When GPT-3 came out—actually, even GPT-2—my youngest brother Eric had neuroinflammation and psychosis, and he’s had it for a long time. It causes OCD and paranoia. My brother, my sister, and I got really good at using GPT-3 for medical stuff. When ChatGPT came out, we started using it even more.

My main use for Manus is medical things. I build, on average, a website or 2 a day. For example, I’ll make a document of all the neuroinflammation or Lyme doctors in the world, and then I’ll automatically message all of them. I’ll upload all of Eric’s medical records, and then I’ll build an interactive graph of them.

I might go and look at a company that’s similar to us. Speechify is basically a B2C version of ElevenLabs. I’ll find every article and every podcast, transcribe it, and map out their entire business. I’ll understand how their business works.

Or I’ll go, “Hey, we’re selling into education. Great.” I’ll make a website with a map of all the United States, every single school district in the United States, who the district administrator is for each school, and make it into a website. Then I’ll send it to our B2B team.

Why do I do it this way? Number 1, it makes it easy for me to understand things. Number 2, I can then plug that into another LLM, and it has access to the entire website that it can scrape. I’ll then put that into Claude, ChatGPT, or Gemini. That makes it really easy.

Manus makes it really easy to post automatically for you. It’s really great at deep research. For example, my girlfriend and I went to a Passover seder the other day, and she didn’t know too much about Judaism. So I wrote a 120-page document on the history of Judaism for her. That way, she didn’t get bored at seder, because she could listen to it with Speechify. She listened to it in 25 minutes in the car on the way there.

I’m constantly making things. The thing is, it’s very easy to use from my phone, and it’s very easy to use from my computer. I use Claude Code all the time, and Dispatch from Claude as well. Manus is just a little bit faster and a little bit easier to use.

Jarvis, which is Speechify’s B2C voice agent, is essentially like Manus but voice-first. Those are the places where I like Manus more than I like Claude Code or Gemini.

I do want to go back to the ad spend as well. You had 1 ad drive $3 million of revenue. Can you talk to me about that and the lessons learned from that?

Sure. The Rule of 100. I had gone and talked to all these people who are good at growth, and then I made a list of the top 100 creators. I talked to them, and I’d live in their houses. Then I was making those ads in my apartment.

Some of them were good, but I was stuck in a local maximum, and I wanted to break out to a global maximum. My conclusion was that the way I would do that was to make a list of the top 100 best-performing ads in history: every Super Bowl ad that was really funny, every ad that made me cry, and every ad that won an award.

I made a list of maybe 500 ads and then whittled them down to the top 100. I read the script for each one of them, and I rewrote the script to be about Speechify. Then I would up the ante on the ad.

The rule in videos typically is that you win or lose by the first 3 seconds. Do you have good retention in the first 3 seconds? But the rule I developed is that you win or lose by the first image of the video—the first 1,080 pixels.

I wanted the first 1,080 pixels to be as crazy as possible. The way to represent Speechify is with headphones, so I got bright red headphones, put them on my head, put on a suit, and went into a hot tub. The first shot is me with red headphones and a suit in a hot tub.

At the time, I had long hair, and I just threw the most random knickknacks behind me—weird photos, books, and stuff like that. I developed this character called Cliff Silver. Cliff Silver appears in many of our ads, and he’s a version of me. He’s very gullible and kind of dull, but he has great confidence.

7. Why 60-Second Response Times Are the Best Predictor of Engineering Success

The ad starts, “Hi, I’m Cliff, founder of Speechify. What is Speechify? Well, you give us stuff you need to read, and we read it to you instantly.” “Yeah, instantly. Are the voices any good?” “No. Our voices are great.” Then it goes into all these different scenes.

The video was really long, so I edited it and made almost 50 variations of that video, like my college applications. I tested all of them. One did pretty well, another did better, 20 didn’t perform at all, and 1 did amazing.

That YouTube ad basically launched the company to a whole new stratosphere and performed for about 3 years straight. That video got hundreds of millions of impressions on Instagram alone. Every year, I remake that video.

I did the nacho commercial from the Super Bowl. I did the Isaiah Mustafa Old Spice commercial. I literally got a horse for the ad. You just take it as far as you can.

Did it work?

Which one?

Did they work like you thought they would?

It’s a complete random distribution of what works and what doesn’t.

What did you think would really work that didn’t?

I thought that the Old Spice commercial would absolutely annihilate.

It got like 300,000 views. The one that I did not expect to work that worked amazingly was a rap song I wrote about my experience with dyslexia and ADHD, and that thing made us, like, $3 million on a rap song. You have no idea.

Do you believe in the power of repurposing? We did a video the other day that went super viral comparing American founders to French founders. Basically, American founders have nothing to say other than, “We’re a trillion-dollar company right now.” French founders have like $100 million in revenue, and they’re like— It went super viral, and the team is now repurposing it into like 10 different formats.

Oh, of course. That works amazingly. Formats are the thing that works, right? One of the things that works really well for MrBeast is ladders, right? The $1 hotel, $100 hotel, $1,000 hotel, million-dollar hotel. Competing where you have people from age 1 to 100 all competing against each other in the same thing. Like, “I got trapped on an island,” blah, blah, blah. He does the same formats but changes the view and how it works.

American founder versus British founder is a perfect format, and then you switch it to French versus Ghanaian, whatever you want. And, of course, repurposing works. You take the same piece of content and put it on TikTok, Instagram, and YouTube, and you change the hook. Especially in performance marketing, it doesn’t matter how many times you post because you’re not going to damage your brand by posting something that will not perform organically, because it’s behind the wall anyway. It only gets served; it lives in the shadows. So just test 800 variations. Why not?

Do you give a fuck about conversion? And I know that sounds like—

We have TikTok, and we get lots of views. Conversion back to the core podcast and core product is very little. Would you say then, “Not worth it. Don’t bother”?

Yes.

Fascinating.

Right. Only do things that result in conversion. We have more than 50 million people who use Speechify now, the organic B2C product that makes it really easy to read. If you don’t download the app, start a trial, and convert to paid, there are 2 options. The first one is, “I don’t care.” The second one is, “I actually do genuinely care that we solve dyslexia and ADHD.”

There are 2 billion people in the world who have either ADHD, low vision, autism, anxiety, a concussion, or who are second-language learners. Our philosophy is that we roll out the red carpet for the user. We give them the best experience possible. If you can’t pay, that’s fine. You still have a great experience.

I think about that as our act of charity. We acquire the user, and they can use the product for free. Amazing. But that’s not what’s going to cover the ad spend. The ad spend is covered by the people who pay. If you don’t get people who actually convert on your product, there’s no point in doing the work that you’re doing.

Do you have a separate growth team for ad testing, new-channel testing, and iterations? Is that all in marketing, or is that a separate growth team?

It’s all in one. It’s not split. There are subteams inside of growth, but they all report to the same people.

When you look at channel composition, what does yours look like today?

Word of mouth is still the biggest way we get people, right? It’s someone telling their friend, “Hey, I know you like podcasts and audiobooks, and I can see that you’re reading at a normal reading speed. Don’t you want to read 3x faster? Use Speechify to make AI-generated audiobooks.” That’s the biggest one.

We crush it on Apple. We won the design award last year, and we got featured by them like 14 times last year. SEO is huge for us—10%, 15% now. 15% of our users come organically from ChatGPT.

Inside of performance marketing, Meta is still the best winner. Google is the most profitable. YouTube is pretty good, especially on desktop. Apple 11 is interesting. TikTok is interesting, but they’re not as big. And you have this great halo effect. The more that you get the message out there in different ways, the more you sow the seeds that result in people talking about it and telling each other about it.

8. Why OpenAI Ads Will Be Massive

You said OpenAI Ads, and you said you were one of the first 200 on OpenAI Ads. What are the biggest lessons from OpenAI Ads?

You just have to test, like everything else.

Do you think it will be huge?

Of course. Yeah, massive. Massive, massive, because it knows so much about you. It’s the same thing as Meta, right? We talked about this earlier. The reason why Meta ads are so good is because they know what you message on Instagram and what you message on WhatsApp. As a result, they can target really, really well, so people convert.

OpenAI knows everything about your history, what you’re interested in, and what’s inside of your psyche. It’s exactly the same way that Google has built such an amazing business, because there’s relatively high intent when you talk to it. The ads convert.

Now, it has extremely high CPMs. But if you find content that converts, the CPMs don’t matter. That’s something that people really don’t realize. It’s okay to pay a high CPM as long as people convert and as long as you have attribution. They just launched the SDK for tracking, so that’s really important.

Would you buy Meta today?

9. Why Cliff Would Buy Meta Today

Absolutely. Yeah, I think Meta is one of the most underrated stocks right now.

Why?

10. QA Is the Most Valuable Skill in an AI World

Number 1, Mark Zuckerberg is an absolute savage. He will hire people if they’re right, no matter what. Number 2, I think that in a world where engineering is commoditized and design, for the most part, is commoditized, the most important thing becomes excellent QA. We can talk about that later. And customer acquisition.

Google has pressure on it from all the LLMs, but Meta doesn’t. The only real pressure on Meta now is TikTok, and TikTok has a lot of issues now that it’s been separated, yada yada. Zuck has also proven that he is the number 1 acquirer in the world. No one acquires as well as Meta.

The leadership team is incredible. Chris Cox is amazing. Alex Schultz is amazing. And I think that, also, in an AI world, data is everything. They have more proprietary data than everybody else. They haven’t figured out the model side yet, but I really do believe that they will.

I look at all the people who are vibe-coding new products right now. Where do you think they’re going to get their users?

You said design is commoditized in some respects. Everyone talks about taste. It’s the new word of 2026, which is, “Oh, no, the value is in taste.” Can you help me understand that? Is that—

No, no, it’s not. Everything is in taste as well. When I’m saying design, I mean the actual grind of sitting in Photoshop, sitting in Figma, sitting in Illustrator, sitting in Canva, and making things pixel-perfect.

We’re already at the stage where most of my leadership team is sending their emails from inside of IDEs, like Claude Code terminals. When we want to launch a new product, we definitely don’t assign a designer in the beginning. We say, “Just use the default Claude Code stuff,” and then we’ll bring designers in afterward, and they’ll make it a lot better.

We definitely do use design for user experience, interfaces, and stuff like that. That matters a lot. Design is a really important part of Speechify. But the hammer-and-chisel part of design is not as important. Therefore, the thing that’s left is taste.

Very few people have taste. I have taste in some things, but in other things I have bad taste, which is why we hire really great designers at Speechify to make up for my shortcomings. Taste really, really matters, even if it’s just telling you, “Pick that one, not that one.”

I’m down 40% on my Figma position. Hold or sell?

Hold. Dylan Field is a savage. Anything Dylan Field does, you should go for.

So, you think they survive in an Anthropic and AI-design world?

I do. I also think that Figma is a relatively antifragile company, just because it’s a really good team. They’re excellent at sales. They have really good taste. Figma itself is a great product because Dylan coded the original thing. I think it was in C/C++. It was like the first product that would render vector-based architecture in a browser.

They can do the same thing. Today, everything is moving really fast. Do I bet on an old incumbent that’s really bad at innovating, or do I bet on the Figma team? I would bet on the Figma team.

The product they end up with might not be the product that they have today, but I think it’s a great company. I think they’re actually benefiting from being a publicly traded company, and I think that they’ll continue to rise and do really great.

Remember, Figma traded at 40x revenue, and now they’re at 6x revenue. It’s less that it’s not a good company and more that the market is making a correction. That’s fine.

Claude Code or Cursor internally—what does everyone use engineering-wise?

Claude Code. We were huge adopters of Cursor about 2 years ago. I think I got the Anthropic team plan in December, and it actually drives me crazy that not everybody has moved over to Claude Code.

11. Why Speechify Is Moving From Cursor to Claude Code

A lot of people will use Claude Code from inside the terminal in Cursor. I prefer that they use Claude Code inside of the Claude Code IDE. I think it’s more powerful, but it’s fine to use it inside of Cursor as well.

The most important thing is that you have to use it to the max. I’m always surprised that there are some limitations that you need to enable.

It took people 3 months to ask for some of the features to be enabled, and it’s like, “Why did it take you 3 months to ask me about this?” I’m pushing people constantly to hit limits. We’re spending more, and we’re getting to the point where soon we’re going to spend more in tokens than we spend on actual salaries.

You’re getting to the point where you spend more in tokens than you do in salaries?

Next year, I expect we’ll spend more in tokens than we’ll spend on salaries.

Is that atypical?

It’s atypical right now, but I don’t think it’ll be atypical in the long term.

You don’t think the majority of great, high-quality companies—

We’ll get there in 3 years. You just need to be a really strong early adopter.

Across all functions or just engineering?

No, across engineering. I would like to be spending more tokens on our growth team and creative. We’re not there yet. We need a lot more testing mechanisms to be put in place for that to make sense.

Are they AI-native? I know that sounds really VC-wanky, but I am pushing people to use AI internally, which honestly is kind of frustrating, especially on growth stuff. So, we do clips here where basically they interview me for an hour and then turn it into 100 clips.

But you know what actually sucks? I’m in the same top, in the same studio, and when you look at the grids, it just looks boring.

Yeah, it’s very easy for you to change that already.

It is, but I’m the one who’s pushing them. My point is, I’m dragging my team into an AI world.

So, I can tell you what we do. Remember, Speechify is a voice AI ecosystem with 50 million users and 1.1 million five-star reviews across platforms. We need to educate even our users on how to use AI.

Their first agent was a text-to-speech tool that reads to you. The second one was a speech-to-text tool that does dictation. The third one was a podcast generator where you can be in the podcast as well as 2 new hosts.

Then we launched teacher agents that will ask you questions about the document and then teach you back the content gap that you were missing. Now we just launched Jarvis, which is Speechify for Work. It lets you do productivity, and then the enterprise Speechify Voice AI.

We’re constantly educating our own users on how to use AI, and what we found is it takes about 100 iterations on a feature before the feature gets adopted. But once it gets adopted, it gets adopted really well. So, you need to do the same education for your B2C users as you need to do internally.

Internally, I am constantly screaming from the rooftops that people have to use Claude Code. I’m more extreme about it than needs to be because I need to move people from all the way over here on the right to all the way over here on the left, and I’m okay if they meet me in the middle.

So, I’m just extreme about it. I’m like, “If you don’t spend 1,000 credits a day, I’m disappointed in you. I need to see that happen.” Send me screenshots, and I will get people on a Zoom call, screen-share the entire engineering team, and show them how I use it.

Then I’ll go, “Tim, you’re really good at this. Show us what you do. Tyler, you’re amazing at this. Show us what you do. Simon, show us what you do.” We’re just constantly demoing. You’ve got to make sure that people are aware of how much other people are using it.

Then I’m like, “If you are not using this tool right now, it’s like a chat in everyone’s Slack. Comment and explain why. It’s going to be very difficult to continue working together if you don’t use it and you don’t explain to me why.”

So, they have to explain to me why, and then I will call them and convince them to use it. I’m like, “You have 24 hours. Send me a Loom video with what you used AI to make.”

What single thing do you use Claude—well, Anthropic or Claude—for that has been most impactful?

12. How AI Helped Cliff Diagnose His Dad's Cancer

The personal side is medical stuff. We talked a little about my brother Eric, who has Lyme neuroinflammation and psychosis. We did a proteomics map of all of his genome, put it on Supabase, wrote an MCP for it, and then had it compared to every single article that had ever been published about anomalies in the genome.

We found 4 anomalies in how his genes were expressed. That was an amazing breakthrough for us.

The second thing that we did is my dad has had prostate cancer on and off for the last 8 years. There was a point where his PSA continued to go up, which meant it came back. We just didn’t know where it was in the body.

The next thing you need to do is what’s called a PSMA scan. The doctor refused to do it because he was like, “Listen, the cancer is too small. What we need to do is sit and wait.”

I was like, “That sounds like the worst advice ever. If we know he has cancer, waiting for another 3 years is a terrible thing to do.” He’s like, “Well, we don’t know where it is.”

So, what I started to do is I went on Kaggle, which is the largest online competition for data scientists. I started hiring people on Kaggle to help me diagnose my dad’s cancer from the radiology reports.

It has a lot of open-source repositories on data for what’s called time series of PSMA PET scans for cancer. I came up with 4 theories. The first one is called tipping theory, which is basically where you look at the level above and the level below where you think the cancer is.

You make the data really noisy, and you see if you can figure out where it is. We started making models for that.

Then I basically took 5 different approaches inside of this research. From talking to Gemini, Grok, and ChatGPT, I realized that there was a machine that could help us.

At UCSF, there’s a thing called the China Basin that does a normal PSMA PET scan, and it will scan a voxel of 4 mm in diameter. But there’s another machine called the uEXPLORER that can scan a voxel of 2 mm in diameter.

There were 3 of them in the United States, and 1 of them was at UC Davis. I messaged the guy who was responsible for that lab, and he agreed to test my dad.

The doctor, who I put a lot of pressure on to give my dad the exam, had scheduled it for the end of January. In December, we got the scan in this higher-quality machine, and we found it. It was in his left seminal vesicle.

I would never have had the confidence to push the doctor so hard to get the scan if I didn’t have LLMs open during the conversation with the doctor. I definitely would not have found another machine or gotten my dad scanned.

13. Building an Internal AI-First Engineering Culture

So, by January 7, he had the left seminal vesicle excised. Before he was even supposed to have the second scan, he had surgery. He was good.

Are you quite skeptical of the state of modern medicine? What I mean by that is, when you were a kid, you kind of just believed everything that doctors told you. The more I grow up, the more doctors tell me, “Oh, don’t worry. The scans were kind of fine.”

I pay for my health care here, and it’s a lot. I’m like, “Kind of fine isn’t great. Actually, send me the PDF, and I’d love to go through it myself.” I put it into ChatGPT, and I’m like, “Kind of fine, yes, but there’s a lot more you could have told me to optimize.”

Are you actually quite skeptical of the state of modern medicine?

3 years into my dad having prostate cancer, we had a meeting with a doctor who was very renowned. He was like, “Yeah, your dad’s fine. We’re just going to sit and monitor.”

I was like, “Screw that. That’s the same thing as us being a JV soccer team. No, I want to be coached like I’m Real Madrid or Barcelona. Give me—tell me exactly what to do. Do you want me to wake up at midnight and see the moon? Whatever you want me to do, I will do that thing. I just want my dad to live longer.”

The problem is not the doctors. The problem is the system, where doctors have to treat statistics, and they need to make sure they don’t get sued. In your case and my case, because we’re educating ourselves and we’re willing to take the risk ourselves, we can do things that result in better outcomes.

They’re a little bit more risky, and that’s fine. My contention is you should just learn as much as possible because the information is accessible to you. Then you are the quarterback of your own medical experience, or that of your family members.

That was so many things that we’re doing about token maxing and spending more on tokens than we are on people, and then we went on a side tangent there.

I think one thing that was so interesting for me was going through some of the notes. You said adversity quotient matters more than IQ or EQ. I don’t think most people know what adversity quotient is.

I had to expand upon it in my searches. What are the leading signals for you of a candidate/person that is going to be successful at Speechify?

People talk a lot about IQ, intelligence quotient, or EQ, emotional quotient, but the most important one is AQ, adversity quotient. How good are you when things get really tough?

I think about it like a Jeep that’s going through the savanna. I don’t want the Jeep that goes the fastest. I want the Jeep that’s not going to have a flat tire, the one that’s going to stick with it the longest.

There’s a heuristic I use for this, which is sometimes you look at an engineer and they’re willing to grapple with a really hard problem for 8 hours straight. Most people, if it’s really hard, quit after 30 minutes.

But the things that really move the company are on the other side of 5 hours of grappling with a hard problem. You kind of have to be obsessed to do that. The way that I find that is I look for projects that people actually did and published to production. Let's say you went and spent 4 years learning computer science and we have a call together. My goal in that call is not to not hire you. Nothing will make me happier than hiring you.

I just need to get you to give me the information that proves to me that I can underwrite hiring you. The whole point of that call is for me to find out how awesome you are. I want to hear about your side projects that you've built. I want you to screen-share and show me the things that you've made. But if you haven't made them, you're just not willing to deal with the little bit annoying thing that requires you to register an Apple Developer account and put an app on the App Store.

The little annoying things that require you to set up a Cloudflare instance and make sure that your website sticks around 3 years after you published it—I need to see that those things happened. Then, ideally, I want to see something novel that you created. You don't need to be a genius. You just need to be willing to do the thing. The last thing I'll say is I have a friend, Valentin Perez, who has a great line, which is, “A little bit of slope makes up for a lot of y-intercept.” It doesn't matter where you start. It matters how fast you grow. The growth rate has to do directly with how much pain you're willing to bear.

I think that this is exactly the same when I look for founders that I want to invest in. As awful as it seems, there are certain things we look for. One of them is that struggles in parental relationships often signify very successful founders, sadly. If you were an investor today looking for adversity quotient or struggle duration in a founder, what would you advise me to ask questions about and dig deep on with founders?

There's a great question from YC, which is, “What is a non-computer system you've hacked to your advantage?” It's a really good one because I'm looking for the story that shows me your ability to, A, build a system; B, learn something; and then C, have an inhuman amount of desire and will and grit. So I'm then looking for the rest.

Are there any mega red flags on the flip side? For me, I hate the people that are like, “Founding GTM advisor.” And it's like, no, no, no. You're just a GTM hire that was one of the first. You're not a founder; everything needs to be a founding role.

Yeah, there are a lot. The first one is any situation where you can discover dishonesty. What I really don't like is when people give me a stat and say, “Oh, I raised the stat XYZ.” Then the more questions you ask, the more you realize they changed the definition of what the denominator was. That's one thing that I really dislike.

The second thing is people who talk too much. The amount of signal per sentence that I get needs to be high, and if it's not, I don't want you wasting the time with the rest of the people on the team. We're all about execution here. Sometimes I'll meet someone who I think is really talented. They might even be really technical, but they just talk so much, and I'm like, “You're just going to annoy everybody else and get us off track.” I want someone who's hyper-focused on exactly one goal, the thing that really matters.

The last thing is you don't want people who are too comfortable. You want people who will be loyal. I always ask people, “What made you interested in interviewing with Speechify?” If the answer is bad, we shouldn't hire that person because there's no differentiating reason why they're here and not at another company, and they might leave.

What is a bad answer? “I want to earn more money.” Is that a bad answer?

Terrible answer. I never would hire that person, not in a million years.

Why?

Because someone else can come and offer you more money, and then you'll leave after we spent a year coaching you up and making you amazing.

And a good answer is?

“I have a kid with dyslexia. I have a niece with ADHD. I use the product myself every single day.” Those are great answers. I read the blog post by Tyler Weitzman on how he skipped 4.5 years of math in high school and built 47 iPhone apps when he was there, and then he went and did his coterm at Stanford. I really liked his explanation of how he shipped the first deep-learning model that worked on the CPU of an iPhone. So I'd love to work with him. That's a great answer.

Just do a little bit of research about the company. I want to see a reason why people are there.

Yeah, that's a better answer. I agree. The cash one was always flimsy. You said that about Tyler and ADHD, or earlier you said about dyslexia, that the best founders are dyslexic or have ADHD. Unpack that one.

Michael Dell has this great analogy: when you're running a company, it's like you're driving a car, and you want the car to fit the driver. You want the company to fit the person who's running the company.

If you have dyslexia and you're 9 years old and you suck at the one thing you're supposed to be good at, which is reading, there are 2 options. The first one is you think the world is stacked against you and school sucks, and you go down this really dark path. 40% of people who are incarcerated have dyslexia, even though 17% of the population has it.

The second option is you keep trying, and you try again and you try again and you try again, and you develop this belief that it's okay to fail and that even in things that seem impossible, if you try enough, you'll succeed. So 40% of billionaires also have dyslexia.

It ends up being the case that you coach yourself that it's okay to fail, and that's exactly what startups are. You're stacking the odds against you. You're doing something that feels impossible, but you're developing this AQ to never give up.

People with ADHD are definitely not slow, and they're not lazy, and they're not stupid. Their mind just works differently. We actually find that for Speechify users, if your brain works at 600 words per minute but you only read at 200 words per minute, of course you'll get distracted. If you start listening at 600 words per minute, you love that and you stay engaged.

Running a company requires you to be responsible for and understand the engineering, the design, the quality assurance, and the go-to-market. All these different things are stimulating you at the same time. Being ADHD is actually very good for being a CEO.

The problem is, how do you get from 0 to 1 to 10? The 10 to 100 is an amazing place to be ADHD. But you have to go from 0 to 100, where you're doing all the IC work yourself, when you need to stay focused, and that's really difficult. It's great to be dyslexic or ADHD as the person who is running a large company, and it's not great as the individual person who needs to build the company.

I'm being a terrible interviewer in this because I'm just jumping around so much, but I'm so enjoying the spontaneity of the discussion. You mentioned the multiple different disciplines that you have to adopt in the early stages of a company. What role does not exist today that you think will be incredibly commonplace in 5 years' time?

Coach. One thing that we do a lot at Speechify is coach people on how to use AI better. Force them to use AI better. We're still hiring people at a crazy rate. Last year, we had 178,000 people who applied for the open engineering positions at Speechify. We have an asynchronous technical challenge, and 19,800 people finished that challenge. So we're very good at hiring good people.

Still, we focus much more on coaching the people who are inside the company. This is something that has been my battle internally, actually convincing our leaders to do, because they have a lot of things that they need to do. It almost feels like a waste of time to go and spend time teaching other people, but you have to do it.

I have a friend, Austin Ray, at Ramp, and his entire job is teaching the organization how to use AI tools. We almost had Austin join Speechify, and it was one of my biggest mistakes that we didn't make that happen.

Why didn't you make it happen?

We hired him as a consultant. We worked with him for a while. I had super-high conviction on him. I couldn't get to the level of conviction I needed from other people, and I made a huge mistake: because I didn't get a resounding “We have to hire him” from everybody else, I didn't do it.

I should have trusted my gut and done everything to give the sexiest offer possible, because he was also obviously a very expensive hire. So there's a cost to it. We should have just given him the offer, because it was clear he was such Speechify DNA.

Do you want to hear something nuts? I've been investing for 10 years. Basically, what I've learned is I can't decide the difference between a 7 out of 10 and a 10 out of 10. But I'm very good at guessing between a 1 out of 10 and a 7 out of 10. So basically, anything that hits a 7 out of 10 in terms of founder quality, I should invest. Right? If I'd invested in every founder that I've met, I'd be 17x more successful. Wow. 17x more successful.

I would have had 2.5% of Deel. I'd have had 2% of ElevenLabs. I'd have a seed check in Flight School, a seed check in Vanta. And I knew these guys were great, but I wasn't sure they were 10 out of 10.

And the honest truth is the market. I saw shipping, I saw SOC 2 compliance, I saw payroll.

Oh, this is really interesting, huh?

I have the same philosophy about hiring. If you can really trust someone, you can afford to pay them significantly more. And it’s actually just: hire the person for more money. If they don’t work out, you part ways. It’s fine. You should always be trigger-happy to hire really good people if you see that they’re really good. Just ignore the fact that they’re expensive. Just make it work.

Also, when you get spidey senses in the interview process, listen. It will only get worse post-hire. What I mean by that is, when someone—you just get the feeling that they really actually want to be an investor, not in finance, trust me, when they get on the finance team of the investment firm, they’ll be pushing to be on the investment team, not on the finance team. Listen to the early spidey senses.

You said before, “Why do we never hire from Google?”

Yeah.

Why?

So, Speechify is an applied AI research lab that focuses on voice agents. There are 50 million people who use the product. We hire all over the world, and the benefit of having a lot of B2C users is people know about the product.

When we try to hire from large organizations, we do have very similar DNA to Amazon. Typically, we found—this is my personal experience—when we hire folks from Google, they come from a place of a ton of abundance. So the hunger is not the same hunger. And you need to coach them out of essentially bad habits. And that can infest your culture in a bad way.

The other thing is that this is true in general of hiring in the Bay Area. The loyalty is a lot lower than you would get from hiring in Europe, because in Europe, people really value a good company and a good job. It’s just not the same thing in New York or San Francisco. Sometimes you find diamonds in the rough, don’t get me wrong, and you should definitely hire those people, but by and large, you’re taking a big risk.

For us, we really like to invest in people. It’s extremely important to us. We’ve only had 1 real case of attrition at Speechify of someone who worked with us for a long time, and that person went to work for Elon Musk directly. Okay, makes sense. But generally, I don’t want people whom we coach and who understand the context of the company to leave. We find that the people who come from companies like Google are more likely to do that.

What do you do to retain people in a way that others don’t?

A lot. I’ll give you an example from the early days of Speechify. We had an amazing teammate. I loved him. He took time off school to work for us, and he told me that he wanted to leave because he wanted to have more friends. There were not a lot of friends in the Bay Area for him, and he wanted to go back to where he was going. Okay, no worries.

The way I found that out is he said he was going to leave. I was like, “Why?” And he didn’t really know. So I sat him down, put my phone in his hand with the Notes app opened, and I was like, “Write the 10 top goals you have in life.” It became very clear that it was all around friendships. I was like, “Okay. Can you describe to me what you’re looking for in your friends?” He described what he looks for in his friends.

“Cool. My assignment for you today: don’t do any Speechify work. Give me, by the end of the day, a Google Sheet with 50 people in the Bay Area that you want to be friends with.”

He gave me Thiel Fellows and people who won the award for being the best student at Stanford and Berkeley and whatever. Then I DMed all those people on Instagram, and I was like, “Hey, we have dinners on my patio every Friday. I know some of you; I don’t know others of you. You’re all invited this Friday and next Friday.”

Of the 50 people, 35 showed up. He was there, and he made a ton of friends. Then he stayed. So often it’ll be stuff like that.

Early Speechify, the person who’s now our COO—we couldn’t get him a visa. We went to 15 lawyers; no one would do the visa thing. So in the end, I read the immigration law myself, and I figured out how to get him a green card directly. It took a lot of work, but we got it. After that, we became really good at immigration.

There was a person who worked at Speechify, a really, really amazing engineer, and he was based in Ukraine. He was going to leave. This was right around when the war was about to start. I flew to Ukraine. I ignored the U.S. restriction that you cannot go to Ukraine. Literally, to board the plane, I had to sign a form saying the airline is not responsible for what happens beyond this point.

I worked things out with him there for 3 days. I ran a hackathon with other people there. We sorted out the problem, and he stayed.

14. Are You Worried About AI Layoffs?

Typically, this is not the case at Speechify. In most companies, this is true: people don’t quit companies, they quit managers. That has not happened at Speechify. Typically, if there’s an issue, we solve it very fast. But people typically leave not for money. They leave for another thing in their life that’s not satisfying them, because they’re, for whatever reason, not satisfied. Figure out what’s not satisfying and solve that thing for the person. Sometimes it’s money, in which case, just solve it.

Are you worried by the number of layoffs that we’re seeing?

Of course, I’m super worried. I think the one thing that we think about—we don’t think about enough—is, in a world where we create more value than ever before and the Gini index gets really messed up and value accrues to the top 0.1% and not to the rest of the population, what happens? How do you upskill people? I think about that a lot, and I don’t know the answer.

If you look at this graduating class of computer science students, there are basically no jobs. You really need to make your own side projects to look impressive. I don’t have a great answer for it other than more people should be entrepreneurs.

Do you have a bent towards younger hires today more than ever, given their proclivity to be more AI-native?

No. I have a bent towards older hires.

Huh.

Because they read textbooks about software engineering. When they see what’s coming out of the LLM output, they understand it. Not all of them are adopters of AI, but some of them are. So you can hire people who are both grizzled veterans and good at AI.

If I see a cracked young engineer who’s great at Claude Code, of course I’m going to hire that person on the spot. There’s no question. Regardless, a prerequisite is that we have this entrance exam.

There’s this gigantic code repo that you need to read very fast. You can’t read it with your eyes. You have to use AI. You have to use LLMs to ship all the features, and then you get a test at the end: how much of the code repo did you break? Then you get 30 minutes to fix what you broke.

Our functional engineering challenge is not just a challenge of how good you are at TypeScript and Python, but how you use LLM coding agents. That’s key. If you don’t pass that, there’s no way that we can work together.

And the best are… You never know.

You just have to try with everybody. You never know. Sometimes people tell you that they’re great at AI, and then you give them a challenge, and they suck. You just have to test.

Sometimes people have never used AI before, and you force them to sit with it for 48 hours, and by the end of the 48 hours, they’re phenomenal.

You said before, 20 minutes ago, that 60-second response times are the best predictor for any engineer’s success. What do you mean by that? And what should we know?

Speechify is a remote-first team. There are 45 AI engineers on our research team. There are about 150 product-facing software engineers, and we have people in 36 countries now—200 people in total. Everybody’s really good at engineering, so communication’s become really important.

I can’t have a situation where you’re blocking somebody else. If someone else is blocking you, number 1, it’s your responsibility to inform them that they are your bottleneck and inform them multiple times a day that they are your bottleneck.

But we also really believe in people learning. If that person cannot unblock you, you should go and learn how to unblock yourself right now. But if that person doesn’t give you an answer, you’re stuck. You don’t know whether you should invest the time to unblock yourself or wait for that person.

Another principle that we have is managing up. If your team leader needs something from you, obviously you need to respond, but you should update them anyway about what you’re doing. In the morning, the first thing that you should do is send your leader, “Hey, here are my top 3 goals for the day.” They shouldn’t respond unless you’re going in the wrong direction, in which case they’ll adjust your direction.

If someone is messaging you to get unblocked, it is beholden on you to give them a response. That response can be, “Hey, busy right now. I can give you the response by 4:00 p.m.,” or, “Hey, won’t be able to get the answer to you until 3 days from now.” Just give me a timeline or give me the answer.

If you’re the person asking and you didn’t get the response, call them on the phone. I don’t know why people don’t use WhatsApp calls more. Just call them on the phone. They’ll give you the answer, and then move on.

Young people are scared to call.

Everybody—yeah, young people are very scared to call. I think that at this point, people above 45 are willing to call and people below 45 are not, because everybody’s skill has atrophied.

Just get the answer and move on. We actually had a viral video the other day because I said that if you don't respond to a message within a 3-hour period, I call it the Titanic rule. The movie Titanic is 3 hours long, and everyone should be able to go to the cinema with their families.

If you don't respond within 3 hours, it can be, “Hey, I'm busy. I'm at a football game.” Cool, fine. That's a good answer. But you need to respond within 3 hours, even on weekends. I got decimated for it. Decimated.

Wrong culture.

It's fine. Obviously, if you're asleep, don't respond. But even if you're with the kids, you could tell me, “I'm not free right now.”

Or tell me, “I'm not free right now.”

Yeah, dropping the kids off at school—back later.

Yeah, just inform me.

Fine. Totally agree. The reason is we are so privileged to live in a world where you can work from your house. You can work remotely, you can take vacation, you can do whatever you want, be anywhere in the world. You don't even need to work 9:00 to 5:00.

You want to work at 3:00 p.m. on a Monday or at midnight on a Saturday? That's fine, too. Take all of Tuesday off, but just get the work done that you said you were going to do and make sure you have a responsibility to your team because it is a living organism. Help the organism win.

You need a fire under your ass that says everything needs to have been shipped yesterday. Speed is the number 1 strategic advantage for any company, especially today. If you're not adding to the speed, you're by definition distracting from the speed. It means this culture is wrong, and either you fix that or you're not in the right culture.

What can I do today to increase the speed of my company? Is that daily deadlines or goals?

No, deadlines are challenging. Goals are also challenging. You definitely said goals. But a dream is a goal without a deadline, so always add a deadline. That goes on a calendar with multiple people on the invite.

“I'm going to ship this to production by 3:00 p.m. on Tuesday.” Invite Rohan, Simon, Cliff, Pankaj, so everybody knows. Sometimes we have calendar invites that start with, “This is not a meeting. It's a deadline.”

If that day, the second you know it's not going to be released by that time, you're responsible for updating everybody else: “Hey, it's not going to be done by then. Here's why. I'm going to do it 24 hours later.” That's fine, but just update us.

When you launch it, send me a screen recording or a Loom video showing me that it's done so I can give you feedback. I'd love to get a TestFlight link, too, but ideally, just send me a screen recording to begin.

That's number 1. The other thing to increase speed is to show it yourself. When people message you, message back. When you need something, call people. If you are moving fast and it's clear that you are hungry and you need it to happen now, everybody will adapt to the way that you behave.

What did you do too fast that you later regretted in a major way?

I think you did an episode of Market Share Shen[?], who was like, “I don't introspect.”

Yeah.

Because you don't want to learn the wrong lesson. I'm trying to think about it. I can't come up with something that was objectively a mistake.

The closest thing I can think of is that we had 2 people we hired from Apple and Snapchat—very legitimate people—and they messed up the culture of the company. They would leave at 5:30 p.m. They didn't live and breathe the company, and in the end, both of them quit after 3 months.

We got very lucky that they quit, but I didn't let them go within the first 30 days. That was a mistake. I was young. I didn't know what good meant. Later, I learned. I was begging and borrowing for people to work for my company at the time. We didn't have any leverage. It was, “Thank you for working for me,” at the time, at least.

I was thinking about this going to bed last night. I've never, ever seen a young founder not fall for big logos.

Right.

Every single one.

Yeah, yeah, yeah.

It's just the canonical path. You will learn the lesson. The thing you learn as you get older—I'm sure you find this, and you're a tiny bit older than me now—is that humans are kind of the same. We make the same mistakes. We do it continuously. Would you be long Snap?

No.

Why?

1, we try to run ads on Snap all the time, and we never succeed. So if you're a rep on the growth team at Snap, send me an email and teach me how to make Snap work. I haven't succeeded after trying.

2, there is a portion of the population that uses Snapchat as their main form of communication over iMessage. They message on Snapchat. I don't think that the way they're dealing with their cash is fastidious today. I think they're overpaying creators left, right, and center. I just think that the lunch is completely eaten by Meta at this point.

We were talking about calls versus messages. Slack messages kill your company.

Yeah.

I need [inaudible] to me on Slack at a [inaudible].

Oof. Most long Slack messages should have been a phone call. Just call. You'll get the answer immediately, and then the person can ask you follow-up questions. That's so easy.

All your engineers who are supposed to be coding are spending time reading Slack. I'd rather you put that in the pull request commit message. There's just no reason. There are things that you do need to communicate on Slack, but you should just call.

The other thing is, we have this internal voice agent. The world is getting eaten by voice agents now. It's so much faster to communicate by voice than to communicate by text, period. Text is great for asynchronous communication, but if you try to move fast, you don't want to go asynchronous. You want to move now.

You said about the Apple and Snapchat hires, “They quit, but I should have got rid of them earlier,” or whatever. Normally, you do a performance review and put them on a performance improvement plan. Why are performance reviews a waste of time?

Total waste of time.

Why?

The problem is that you didn't communicate to them ahead of time what their goal was. Typically, the reason why people end up on a PIP is because they didn't ship something to production.

I have a philosophy that if you made something amazing but it's not in production, it's a waste of time. You get no credit unless it's in production. The analogy I always give someone is: you're a farmer and you're a milk delivery guy. You have the best milk in the world, and you put it in a bottle, but you didn't put it on my doorstep. You put it down the street, and now the milk is spoiled. There's literally no points for all this milk that you created.

It's like you're a football player and you didn't put the ball in the goal. You put it right before the line. You didn't cross the end zone. No credit.

The problem that the leader has in that situation is they should have informed the person ahead of time that they had not shipped the thing to production. All you do is get on the call and go, “Hey, Matthew. Hey, Harry, listen. We've been working together for 30 days. You still haven't shipped anything to production. We're a startup, and the only time that I as a CEO get recognized is when things go to production.”

“I need you to put your thing in production in the next 72 hours. If you can't, reduce the scope of the feature. If it has 8 features, only focus on 1, but ship that thing to production.”

Get it through the pull request. That doesn't mean you submit the pull request; it means whoever needs to approve your pull request, get on the phone with them and get them to review it live right now. Then you go and QA it. If you're not allowed to QA it, call the QAs and get them to approve it right now.

If there's an issue, then you go fix it. Just make all the timelines faster. Work on that feature until it is in production. Then I know that you're an actual outcome owner, not just an engineer.

If you communicate that, the 72 hours end, and you still didn't send the feature to production, you don't need a performance review.

Huh.

The person is just not an outcome owner. Then you need to decide: are you okay with not having an outcome owner in the company? That's up to you.

Now, you said about QA there. QA is now the most valuable skill in the world.

Totally.

Why is that, and why are most people bad at it?

All right. In a world where software engineering is commoditized and design is commoditized, if you try to build stuff with Claude Code, even if you give it all the tools in the world, it will not succeed in QA-ing itself to perfection. Definitely not on different devices and different phones, when the Wi-Fi goes down, and all that stuff. You still need a human to do that.

I'm amazed to see, even with very good software engineers and designers, how often they fail to put themselves in the user's shoes and just test edge cases. I have a friend who works directly for Elon. Elon's the best QA; he'll always find your bugs. It's super embarrassing.

I find the same thing with Peter Thiel. I sometimes find more bugs than the actual QA team does. I do because it's in production live. I see it. So then I call people immediately or send them a screen recording. I'm like, “Why do we have this bug?”

Either there's a good explanation for why the bug is there, in which case I'm cool—got it. Or it's just a huge lift to fix this, and we'll fix it later.

Maybe we'll tweak the scopes to bury it a little bit. Or it's actually relatively simple to fix. Let's fix it. So, let's say you built a feature with one shot of a great prompt on Claude Code, and now it's on your phone. The next 48 hours is just you finding the bugs and making an indelible piece of software that does not break. And then you ship that thing to production. That's what separates a product from a great product: QA.

This is the other thing that we do. Let's say you have a person who's not an amazing outcome owner. They just can't, right? People are like that. Hey, that's okay. That's a personality trait. But if they're still a good software engineer, they will be good at QA. So, before letting them go, I'd actually rather put them in QA for a little bit and see how well they do there. And if they do well there, that tells me something, and we should not get rid of that person. We should just coach them properly. But if they're also bad at QA, then you should let them go.

So much of company interaction and assessing talent is done in meetings, in some respects.

Mhm.

Why did you ban meetings?

We hate meetings. Number 1, I like having an asynchronous culture. I don't want you to not be able to take your kid to soccer practice, and I don't want you to not be able to go hang out with your wife and watch a movie. If there's a meeting on your calendar and it's 60 minutes long, 30 minutes before that, you still need to be home. Thirty minutes after that, you still need to be home. You can't think about the project that you're doing. It's better if I just call you for 3 minutes or if I just text you. There's no reason for meetings.

The other thing I hate about meetings is situations where 1 person is talking and 5 people are listening. That should have just been an email, or that could have been a Slack message. Who needs this? Nobody needs this. You're just creating slowness for absolutely no reason.

The places where meetings become a core of a company are companies that don't know their direction, and so the meetings are used to try to find the direction. In that case, just go work out with someone from your team or a group and chat about it. You're doing something else at the same time that's building togetherness.

The other thing I really dislike is performance reviews. We used to have a situation where leaders—we don't have managers; we have leaders. I don't like the fat general sitting in the back saying, "Take that hill." I want the guy who's running up the hill first, engaging the enemy first. You should be the strongest individual contributor on your team if you're leading that team. If you need a people manager, we'll give you a people manager to manage the communication. But I want you mainly hands-on keyboard, coding, designing, talking to users.

We had leaders who were spending 8 hours a week in performance management meetings with their 8 direct reports. I was like, "This is stupid. We're going to cut that to 15 minutes instead of 60 minutes, and we're going to do it every other week because there literally is no reason for doing that." Just have a habit where your team sends you their top 3 priorities every single morning. You don't respond unless the direction is bad, in which case you tell them, "This is the correct direction to go."

Do you like HR?

If by HR you mean recruiting, I love recruiting. If by HR you mean dealing with internal politics and issues, we, for the most part, have succeeded in sidestepping it so far.

Amen. Thank you. I tweeted it recently, and basically the only people who shat on me were people who said, "Recruiting's really important." And I completely agree, but I think it's so important that I take talent out of HR. For me, talent is its own independent function, and HR is management. There's acquisition, and then there's management. And bluntly, I don't think great people need that much management. Vacation monitoring is not something that great people need.

The way I think about it is, if we hire you, I'm not going to be your manager. You need to manage me. I need you to tell me where I need to unblock you, and if I know that, I will go and unblock you.

We had this a lot in the beginning of the company. I would just hire a ton of people, and then I'd give them the goal. If they got it, I'd give them the next one, and if they got it, I'd give them the next one. But some people would just stop communicating. They actually just stopped working at Speechify because they didn't keep communicating. You just need to be interested in the problem and think about it all the time and work toward it and send demos, and then you make progress. And that energy is the thing that I'm hiring for.

Our leadership team is stacked because everybody is a savage. They think about the product constantly. Rohan just launched this great thing. He thought about it in the shower and launched it. You just want to be in a situation where your team pushes as hard as you do. That's it.

Do you think everyone can be a doer, or do you think it's inbuilt in you? So, you do think everyone can be a doer. You don't think it's inbuilt in you?

No, I think everybody can be a doer. I have this concept where I like awakening people. I meet people, and Steve Jobs has this great quote: "The world around us was built by people no smarter than you and I."

You live your life, you want to try not to bump into the world too much, but you could do things. Someone just needs to tell you that you can do things sometimes. And the second you see it, especially if you're next to someone who's a really high-quality operator, you're like, "Oh, yeah, I can do those things."

There's a great question that I like: Who's the first person who believed in you more than you believed in yourself? That's my favorite thing to do. I'll meet someone, and I'll be like, "This is such a talented person," and I can see how amazing they are. I just get to the point that I believe in their ability more than they believe in their ability. And so, I just give them more responsibility and more opportunity, and they almost always rise to the occasion.

Sometimes they don't, in which case we figure out how to help them get to that place. But yes, everybody can become a doer.

You said you were hiring a lot of people quite quickly at one point. You never raised a lot of money and did a big, splashy round. Can you just help me understand what the fundraising profile of this business looked like? Because it's such a weird business.

We chatted before, and, in all honesty, it was weird for me going into meeting you. But we went for a walk around the park because I was like, "I don't know. I don't know. My friends are VCs—all of them, obviously—and there's nothing online." Help me understand this.

Yeah. So, I started Speechify when I was in college. I was 21 years old. For 4½ years, it didn't grow really fast. I just iterated on it until it did, and then it started growing really fast. And then I did this blitz of finding all the people who could teach me.

4½ years in, how much revenue were you making?

By 4½ years in, yeah, we were making more than $5 million at that point. And so we started growing really fast from the $5 million upward. A ton of the people who I talked to were the early investors: Dylan Field from Figma, Mike Krieger from Instagram, Ev Williams from Twitter, the Zillow founders, Audible founders, Grammarly founders, Robinhood founders, Honey founders, 23andMe founders, Plaid founders, the founders of Medium and Brex, Richard Branson, and Gary Cohn, who led all of Goldman Sachs. Charlie Songhurst was my very first investor.

Charlie Songhurst was your first investor?

Investor, yeah. And so we started scaling the business like that. We just found product-market fit, and the business started to grow like crazy. And what I realized is, if you find lightning in a bottle, you don't want to tell other people about it. Speechify has 1.1 million five-star reviews. We have 94% of the B2C voice agent market. We read more to people than any other AI product in the world.

How much money have you raised?

We don't share it publicly.

You don't? You actually don't say it?

No.

Does your team know?

No.

Wow.

The team knows, but, yeah, no, we don't.

And you don't say your valuation publicly?

No. There are only 3 reasons people typically share these numbers. Number 1, they want to bring more users in. We get a ton of users. We get almost 1 million users every single week. There's no reason to.

Number 2, for hiring: 178,000 people applied to open engineering positions at Speechify last year. We don't need to make splashes about fundraising to hire. And number 3 is to get more investors in, but there are 10 investors in my inbox every week who are wonderful people who want to invest in the company. That's not what we need either.

What I need is to learn how to get more users. We just launched the enterprise business this week. That's a new part of my business. What I want to learn—

An enterprise sales team?

Now, that is exactly what we're doing now.

But front-deployed engineers, yeah.

Again, it was so funny. I was researching an old PitchBook, which is—you pay $50,000 a seat. You get data on everyone. There’s nothing on you. It came out with news articles from 3 years ago, and I was like, “Fucking hell, this is impossible to do.”

Avoid doing news as well. We just don’t like it. I want users to come to Speechify because they want to use voice agents or because they want to read their PDFs, not because I’m looking for competitors.

You know what’s interesting? I’m going to talk about 2 people now because I had Jake Paul on the show recently, and he was like, “Relevance is everything. Relevance is everything.” I really like Jake and Logan, but the comments were like, “Less relevant now.” You were like, “Logan actually doing WWE makes sense.” Would you buy or sell the stock of Logan Paul?

Oh, 100%. Super hard. So, what would you do?

I don’t have a relationship with Jake. I know Logan super, super well. He’s one of my favorite people. Logan went super hard on Vine. He was the biggest person on Vine, and then Instagram, and then Facebook, and then YouTube.

This is the concept we talked about, which is bulking and cutting cycles for companies. It also applies to creators. At that point, he was just dialed in, right? It’s like you’re cutting: you’re just creating value. You’re not learning. You’re not doing anything. You milked everything you could milk from YouTube, and he was like, “Cool. All right, what’s the next challenge?”

They both got into boxing. They’ve done really well in boxing—really, really well. Then Logan got the seat in the WWE. Now, people forget this, but WWE is a top-20 YouTube channel, in addition to having TV and Netflix and everything like that.

When I was 30, I was really concerned that I hadn’t achieved the goals I had in life—not even close. So, I got this cabin on a river in the UK, and for 4 days in a row, all I did was make a list of the top 200 people who had accumulated the most power to impact the world by the age of 35.

One of the common themes I saw is that people sometimes have this exponential growth between the ages of 29 and 35. One of the things that happens is they’re like a virus: they attach themselves to a larger organization from within, they leverage that organization, and they use that organization to help them grow.

The WWE is what launched The Rock. Logan is now accumulating all these amazing fans from the WWE that he could not have accumulated on his own. In addition, obviously, he did a great job with likely KSI launching Prime—a legitimate business in a very, very short period of time.

Both Logan and Jake now have their fund. Jake was, like, the meme of Sora. Great thing to do. I think they’re now in this bulking mode where they’re exploring a bunch of different areas and seeing what takes. Prime took, WWE took. There are all these different things that they’re trying.

Like him or hate him, part of the whole philosophy is being not likable but still generating attention. Logan’s a really, really smart guy in my personal experience. I also think that there are very few people who are playing the game like he is. He’s a very unique situation. Is it a company? Is it a person? Is it a social media company? What’s happening here?

I think he’s created a lot of leverage for himself, which is really interesting.

Would you apply that same philosophy to me, then, as being a virus? But if I were to do that, that would mean joining Andreessen Horowitz or joining Sequoia.

But you can’t leverage Andreessen Horowitz or Sequoia. You cannot redirect the snake’s head to be going in the direction that you want. Maybe you could. What I would do is join Union Square Ventures. I would join an institution that has a really storied name but doesn’t have particularly strong leadership, and I would become the leadership for that organization.

Over just building it yourself and being a Thrive?

You can also just build it yourself. The other thing that you could do is find another way of doing it. What’s the best example? The podcast All-In.

Yeah.

You can have a side chapter to your life. Let’s say you now became a fourth or fifth host on the All-In podcast. That would be pretty good for your career. It wouldn’t take that much time from you, but it would help grow you really fast. It would help grow them really fast, too.

Let’s say you went on Shark Tank or Dragons’ Den. That would be amazing for you. A huge existing institution takes very little of your time, but you’d be the loudest voice in the room with the most interesting things to say, and you’d get a ton more views.

By the way, a great example of it: maybe you write a book, “How I Became a VC by the Age of 28.” Then that book is available in all the airports and newsstands, and you’re using Penguin Random House as your launching pad. Figure out what is a launching pad for you.

For you, what’s interesting is you started, what, 10 years ago? 15 years ago? Podcasting was not a thing at the time. You were so early. You rode a wave that came up with you.

The question is, can you go now to a different wave that is already big and then, because of the strategic position you are in, grow with it? TV is a great option. YouTube is an interesting one.

MrBeast is another of these characters. You slept at MrBeast’s house for 3 weeks. What did you learn?

I first met Jimmy at one of my friends’ houses. I had 8 people working on the leadership team at Speechify at the time. I had 100 people. He had 100 people. He had 8 leaders.

15. What Cliff Learned Spending Time With Mr Beast

We just ended up talking for 3 hours, and I was supposed to fly to New York. He was like, “Don’t fly to New York. Come back with me to North Carolina. I want you to meet the rest of my team.” I was like, “All right, sounds pretty cool.”

So, I flew there, and I just ended up staying. We worked out. I would see how he filmed videos, and I learned a ton.

The first thing: I once saw him stop a shoot with 100 people on set just to essentially repaint the door pink. No genius can figure out what’s going to convert in ads. Jimmy knows what’s going to convert in organic content. He has the algorithm in his head.

He has spent, between the age of 13 and today, studying what converts, and he’s dialed in on human psychology. It’s insane.

The second thing is the formats. Their videos are so data-driven, and they figure out, “Okay, here’s a format that works. How do we up the ante on this video? How do we make it even more amazing?”

The next one is that the content that converts the best is content that doesn’t even require language, because that works on every single continent in the world. You want the simplest concepts to be in the video.

For example, there’s this ad for a waterproofing company, which is a very complicated concept. The best ad is a bucket. You make a hole in the bucket. The water’s coming out of the bucket. You put their material and duct tape on the bucket, and it stops coming out.

What are the concepts? Bucket, hole, duct tape, water. Everybody understands these concepts, so make things simple.

For us, PDF is a complicated concept. A book is not a complicated concept. Our best-performing ads include books, not PDFs.

He’s great at hiring. He hires a ton. He tests people, and then if they don’t work, they don’t work. But he’ll always try and try and try and try and try.

He’s maybe the most ambitious person that I know—an insane level of ambition as an operator, as a businessperson, and as a creator. I’ll give him his flowers: he’s done a really good job.

His views are at their lowest point in several years. Why would you be bullish on Jimmy moving forward?

Not in several years. Last year, they had their highest views, but this year they’re having a little bit of a challenge.

One reason why is because we talked about platforms. YouTube was the number-one place for a long time. Now, the best place to be a creator is to be a streamer, partly because streams get clipped, and then those clips can get watched 7 billion times in a week on TikTok, Instagram Reels, and all these other places.

He’s already started big-time with streaming. Last year, he also took on a huge project, which was Beast Games. They did really well on Beast Games.

They also spent a ton of time with Feastables, and they have 2 other big projects that they’re running right now. I would say that Jimmy, for the last 24 months, or even more, has been operating as an entrepreneur, building the company and building the products even more than building the content, while making the largest unscripted show in the history of the world in terms of the number of people participating.

He’s taking much bigger swings. I’d say that, similar to Logan, Jimmy went through a period of bulking, and now they’re in a period of cutting and figuring out, “Okay, here is the next phase of the business. Let’s figure it out.” They could grow even bigger.

I have a segment of this show, which is wild stories, because you gave me so many options. Another one that I loved was, “What led us to buying a 3x-leveraged option on NVIDIA in 2022 before the market caught on?” I was reading this like, “What?” Can you talk to me about that?

It was not me; it was my brother Tyler. It was the case that we were very excited about A100 GPUs back in the day. H100 GPUs were coming out, and they were 3 times faster than anything before.

And we spent so much time trying to spend millions of dollars buying H100 GPUs, and no one would let us buy them. We were like, “What the hell is going on here?” There was a presentation that Jensen made as part of the earnings call, and it was just this amazing example of how they could render a world. We were like, “This is a completely different level of technology.”

The benefit was that we just had context. It became very clear that the market was mispricing NVIDIA. Tyler had even higher conviction than me. I just bought a bunch of NVIDIA stock directly. Tyler bought a 3x-levered option. That’s now like 36x up. Tyler leads our AI engineering team; he just had the context.

My undergrad was renewable energy engineering. I put 1/3 of my money into Tesla in 2015 because I had context. I had done a capstone in photovoltaics engineering. I just understood what was going on. I think at the end of the day, Charlie Munger and Warren Buffett have this philosophy of the 20-punch-card investment idea. Everybody would be a much better investor if they only had 20 opportunities, but when you find the opportunity, lever as hard as you can into it in a responsible way.

Do you know what the biggest lesson I have from 10 years of investing is? You have to diversify to concentrate. What I mean by that is, I should have invested in everything, and I should have done the work to find the Alan Changs, the Adam Foroughis, and the Cliff Weitzmans. You have to be broad to have the luxury to really know.

Yeah, and then triple down when that person is in your corner. But the other thing is, can you go deep into an area that other people are missing? If you could do that well, you will find the honeypots.

Would you still say NVIDIA is a buy today?

I mean, look, I don’t know of a company in the history of the world that has made as much profit as NVIDIA today.

Do you think it’ll be a $10 trillion company?

You just need to look at the multiple of the revenue and profits. What is it trading at now? $3 trillion?

Yeah.

What’s the—

More. More—$4.2 trillion.

$4.2 trillion. What’s the multiple on revenue?

I think 35.

35 is reasonable, given how much they’re growing. So, yeah, I think they have a lot more juice.

One thing that I’m spending more and more time on is energy, naturally.

Yes.

16. Where to Invest in Energy for the AI Era

How do you think about energy requirements from AI, given that we’re now at 1% of energy utilization going toward AI? That will likely increase 10 to 12 times in the next years. How do you think about energy requirements, where you would be investing, and how you would advise me?

One thing that I was surprised by is hydro. All the dams that can be dammed have been dammed. Hydropower is the most efficient and easiest way to create energy. We just don’t have the ability to invest in hydro more.

So, then it leaves you with fossil fuels, solar, or nuclear. I do think that we’re going to figure out nuclear fusion. I think that nuclear fission is also pretty great, but nuclear fusion is coming. We just don’t know where it is or when it’s coming.

One of my biggest side projects that I really want to do is build a mini fusion reactor in my garage. I just haven’t had a garage for a long time, but when I have one, that’s one of the first things I want to do. Not one that creates energy, just one that is contained, but I think that we will figure it out.

That’s very exciting, but in the meantime, while we haven’t figured it out, I’m very much a solar maximalist. I’m all in on solar. The theoretical limit for solar, at least when I was in school, was something like 47% or 49% efficiency, and we’re getting very close to it.

The cost of solar, typically, is mostly borne in the installation and manufacturing, not in the cells, and China has done an amazing job bringing the cost of solar down a lot. It’s just such a great investment amortized over time.

We just have not installed enough solar in the United States to the degree that we should. China is outdoing us tremendously, and there’s so much, essentially, public land that nobody wants that could have solar installations. There’s just a lot of regulation and zoning that stops it. This is the easiest way for us to solve the energy problem right now.

If I were you, I would look for either A, those types of amazing founders that you discussed who are working in the fusion space, or those types of amazing founders that you discussed who are working in the solar space and have some go-to-market that is amazing.

Let’s talk for a second about SolarCity. The reason why I invested in Tesla way back in the day was because I found Elon Musk from a Wikipedia article about the co-CEOs of SolarCity. I was like, “Who is this guy walking with Obama, with rockets in the background of his Wikipedia?” So, I read the Wikipedia and then I found more information.

SolarCity was not a technology company. They were a finance company that had convinced—I think it was J.P. Morgan—to amortize the value of a solar panel over 30 years, and they got them to underwrite the value of that. Therefore, they could loan money to consumers to install solar panels on their houses, and they became the largest installer of solar panels for consumers in the United States by far.

The innovation doesn’t always have to be technological. It could be a go-to-market innovation around finance, customer acquisition, or something else. So, I would look for those honeypots.

What do you know about product-market fit that most founders don’t know or think about?

What do I know about product-market fit that most founders don’t know or think about? The first one is the importance of iterations. When we started Speechify, I was living in Palo Alto. I was a solo founder at the time, and I just had a bunch of interns. I tried so hard to hire, and I found it really difficult. The main problem was people were not using the product.

So, I would walk onto Stanford’s campus every single day with a roll of duct tape, and I’d put it over my mouth, and I’d give my phone to users, and I’d see where they’d click. They’d always click on the wrong buttons. I’d go back, fix it, and try it again. I was constantly embarrassed by the fact that my user experience was not good enough until we got it to be really good. It took a long time.

Then what we did was add, in the bottom-right corner of the app, a big red button that said, “Help/message us.” If you opened it, it didn’t go to Intercom, and it didn’t go to email; it went to my personal text message, like my iMessage. The nice thing about that is, if it was 3:00 a.m. and you messaged me, I’d respond back. You didn’t need to enable notifications because I was messaging you. If it was not 3:00 a.m. but the middle of the day, I would call you. I’d solve that problem on the spot.

Every day, I would export the people who had churned, given that it was 30 days since they downloaded, and I would call them because I had their email. That gave me the ability to call them via FaceTime audio, and I’d figure out why they churned. That took a ton of energy because it’s like getting kicked in the teeth every single time someone tells you your baby is ugly. But it allowed me to solve the problems.

If I were starting a solar company, I would do the same thing. Every single time I’d have an installation, I would personally call the person and say, “How did it go?” I would follow up—or at this point, I’d have a voice agent do it. Just follow up with users constantly. It’s the normal YC model: just talk to users.

The other thing is, just because users don’t use the thing that you made doesn’t mean that you have the wrong direction. It means the implementation is wrong, but the idea could still be correct.

The whole Henry Ford idea is, “You can have any color that you want as long as it’s black,” or, “I would have given people a faster horse.” The Steve Jobs idea of this is one thing that Brian Chesky and Dylan Field are very right about: A/B tests are abdicating decision-making to the user.

Sometimes you want to just have a very clear vision, go for a very big goal, and work toward that goal until it’s in the market. Then, when it lands, sometimes people think, “If you build it, they will come.” They will not come. You have to build it, and then you have to convince them to join you. Then you have to iterate on the product, sometimes for years, until the fit is correct.

So, it goes back to AQ. The most important thing for product-market fit is not repositioning your product or doing anything else. Just stay in the game. Keep making iterations until it fits like a glove, and then you’ll find success.

It’s so funny. We actually have a very strong product perspective on YouTube versus Spotify. On Spotify, we do not offer video. You cannot watch this on Spotify.

Why is that?

You can watch it on YouTube, and we have shorts across IG, TikTok, and Instagram. But you can’t. Why? Because the quality of the edit you can make on an audio-only file versus a video file is infinitely better.

If you want a very curt, precise edit, which is significantly shorter—about 10 minutes shorter—that’s a lot shorter than what you listen to on audio versus on YouTube, where you have the full video.

Wow. Yeah, I like that. That’s very opinionated. It’s very taste-based.

By the way, the CEO of Spotify is actively annoyed at that.

So, why don’t you use AI to make the frame transitions work, and then the edit can be the same? Why don’t you take the audio clip—

AI is not there to make the frame transitions work.

I like the taste.

So, hey, the CEO is not Daniel Ek anymore. Who is it? Gustav Söderström? So, Gustav Söderström, Harry is asking for a solution. Please give it to him.

Well, I would love it if one should be able to upload the video file and then it should sync to the audio file.

Yeah. The thing that it does do is it automatically plays the video, and then you click Switch to Audio. That means when you switch between different apps, you don't get the annoying video thing in the corner.

That's kind of solving a different style of problem.

Different problem. Correct.

But it'd be great if you could have 2 separate files attached, and you could choose one cut or not cut.

Absolutely.

Super interesting one. Another one that I just have to ask is: you got hacked, you found the hacker, and then you hired him.

Yes.

And he lived with you for a year.

Yes.

What?

We had a security vulnerability that did not expose user data but could cost us a lot of money. Essentially, someone could use our models for free. I was aware of this. It was basically a trade-off between that and launching new features on the AI model. I was like, “No, no, launch the new feature.”

I was like, “No, no, launch the new feature. The worst that can happen is they can abuse us.” Then it happened again, and then we got hacked. I deserved it. But it's okay, because the amount that we lost was a couple hundred thousand dollars, but it wasn't crazy. Anyway, this happened, and I was like, “God damn it.”

I went on this search to figure out who hacked us. I found the person. The attack came from a school domain, so I ended up messaging the provost of the school. We got on a Zoom call. I was like, “Do you want to tell me what happened here?”

They fessed up to it. They were a kid; they were young; they didn't know that it—I mean, I don't know how they didn't know, but they didn't know this was not okay. I was like, “I see.” Then I asked them to stop, they stopped, and they took down the thing that they did. I was like, “To be honest with you, this is pretty impressive.”

This is exactly what I would have loved to have done at your age if I didn't know that it was not okay. But I was not ready to hire them for Speechify. I was like, “This person's got it. They got it.” So I hired them to do personal projects for me—little side projects, like optimization of my email, Drive, and all these small apps that I wanted to improve my life that didn't exist.

So all the early voice projects at Speechify—the very first version of Jarvis that I could call on the phone and talk to—he built for me. I had him in the core team, and then I would text him, “Hey, can you do this? Here's a cool repo I found.” He would just do it. We built the translation and all these things. Eventually, he came and lived with me, first in LA and then in Miami.

At that point, we built enough trust that he joined the core team and worked on real projects. But my philosophy is, if you find someone that's talented, you better work with them. Most people, if they get hacked, think, “That person is my enemy. They are evil. I do not want to interface with them ever again.”

You'll see in world history that the most successful leaders are the people who end up having some confrontation with someone, but recognize that person is amazing and find a way to still work together. So that's my philosophy in general.

I do have to ask just a couple more. One is: we're seeing insane revenue ramps from consumer AI companies, and it's distorting our minds in a lot of ways. When we're looking at insane revenue ramps from consumer AI companies, you mentioned churn earlier. What do we not know, think about, or analyze when it comes to churn, usage, and retention? Advise me.

The main thing to look for is inference costs burning those companies alive. In the early days of ChatGPT, they were just burning billions of dollars in inference costs. We had exactly that revenue ramp in 2022 because we were just early. We started doing this in 2015. Dyslexia.

We'd built models, and they were great, but once users probably started to use Speechify, it stopped being a business and started to be a charity. The whole focus of the 40-person AI engineering team at Speechify at the time—and all the GPUs that we had bought and the data center that we set up—was to reduce inference costs. We got it to the point of single-digit dollars for 1 million characters. If you go to any other model, it'll be anywhere from $30 to $100 per 1 million characters. So it's 2 orders of magnitude cheaper.

What is ElevenLabs?

For the standard model, it's anywhere from $70 to $100, and for the V3 Fast model, it's around $30. So literally, more than 10x cheaper.

That was just our whole focus for 2 and 1/2 years. We solved it. And that's why the company's been profitable for so long. So when you look at a company that's hyper-growing like that, you just have to ask yourself: are they just giving away free money?

Let's take Cursor, for example, which is a really great example. Cursor's one of my favorite examples ever because the market demanded that tool. Every software engineer who was coding inside of a conversation with ChatGPT wanted an integration into their IDE.

So all they did was take 2 separate tools, fork the open-source repository for VS Code, and add Claude on it at the time. All the electrons on the right side were dying to go to the left side. The second you put a bar of metal on top of it, of course you had the flow of money.

That flow of money, as we all know, went straight through Cursor and to Anthropic. That is actually great because they didn't lose real money in the process. They just upped the revenue that did not become profit. But very often, you'll have a company that raises $50 million, and they're just burning the $50 million alive.

So make sure that there's a plan for how to reduce costs with time. However, here's the most important thing: any Harvard MBA can figure out how to cut costs. It just requires being smart. It takes a genius to increase revenue.

So still, if that company has an amazing growth lever, as long as there's some differentiator—they have better design, they have community—community is so important—they have community around it—then that's probably a pretty good bet.

Are we about to see churn rates hit companies in a way that we haven't expected?

Not if it's a product that you really need. If you have a disability and the tool—if you are blind and you need glasses, you will never churn from glasses. If you need a wheelchair, you're not going to churn from your wheelchair, in the same way that people don't churn from Speechify.

But if it's a frivolous use case that's not important, then yes. A meme maker, sure. But if it's a tool that you need to survive, then no.

Does Apple buy WhisperFlow?

No, Apple launches a competitor.

Really? Why are they so incapable?

Apple bought the Siri team way back in the day. Apple has this philosophy of on-device security, and therefore they're not using models that run on the cloud by and large. The keyboard does not give you access to the full RAM. Therefore, it's very difficult to run either an API call that does text-to-speech to speech-to-text or something that connects to the app.

Wispr Flow has done an amazing job. I'm super proud of that team. Such great marketing, everything. Obviously, you can do the same things with Speechify, but they just did such a good job with the marketing. Organic, incredible. Great growth story.

Here's the thing, right? It's very difficult, like you said, to be AI-first, AI-native, audio-first, AI-first, mobile-first. Apple has built such an amazing ecosystem that by and large it's kind of become lazy, right? This is the cycle—the circle of life—for startups.

Look, even Apple makes so much money off the App Store. I can't even get an API to get a receipt for a user that makes a purchase. It's so difficult to work with the App Store at this point. And we love Apple; it's made our business so much money. But it's just hard. I'm very excited about the transition of the new CEO at Apple. Let's see what comes out of that.

Why are you so excited? Why are you so excited?

I think that the new CEO is more of an engineer. I'm excited to hypothetically have more support for good engineering tools for people to build excellent things for the platform. Tim Cook is like a god when it comes to supply chain. No one comes even close.

I think that Apple just won that race, right? AirPods were absolutely a huge win. Apple Watch was a huge win. The costs for producing iPhones are amazing. Have you ever heard of Apple having a supply shortage? The first time I ever heard of one was when Mac Minis were out of stock. But they always had the product to sell. Amazing, amazing company.

By the way, AirPods made more money than Spotify and almost every other audio company put together every single year for the last 7 years. The amount of listening is absolutely unhinged, since podcasts and audiobooks have 40% penetration in most markets today. In markets where they don't, like Mexico, they're growing exponentially. So you have picked the correct career. Speechify is just AI-generated podcasts, AI-generated audiobooks.

So, the TAM is really big. But Apple drove that with AirPods, so they won. They locked in the supply-chain side and, obviously, the brand side. Everybody trusts them.

Okay, time to go back from cutting mode into bulking mode. That's it. Dude, we can do a quick-fire because otherwise I could talk to you all day. I'm going to mix it up a little bit: you can buy OpenAI at 850, or you can buy Anthropic at 850.

Oh, damn. I like to be contrarian when it comes to this. Most people in my seat would say Anthropic right now because I love Claude Code. Oh my God, I would marry Claude Code. Very hard decision. To be contrarian, I would say OpenAI.

Why?

I often look at the founders, and though I get that Sam Altman rubs a lot of people the wrong way and comes off slimy, I think he's really good as a company operator. Really good.

So, I buy that. What would they say for me?

Sure. No one has raised as much money as Sam Altman in the history of Silicon Valley. That's completely ridiculous.

True.

The recruiting for OpenAI was absolutely insane. You get Greg Brockman, you get Ilya Sutskever.

Terrible. Terrible.

Greg Brockman?

Terrible. Terrible. This was years and years ago. This was 10 years ago. The latest recruits in the last 5 years? Mega bomb.

Well, okay. You had the whole coup, and the internal team did support Sam, right? Also, when it comes to enterprise deals, he's so good at it that it looks like black magic and looks like fraud, right?

Here's the analogy I give people, and again, I want to give them credit because they did it first. There's a lot of AQ in that company because OpenAI actually started in 2015. ChatGPT didn't come out until 2022.

The way I like it is, you had Columbus going to the richest man in the world at the time, the King of Spain, and saying, “Hey, I want to get to India”—AGI. On the way, he found the United States, America—ChatGPT. That's exactly what happened. Sam and Ilya and Greg and Elon went and got a ton of money from essentially Bill Gates, right? Microsoft. They tried their best to get to AGI and found ChatGPT along the way. Then they made it into a really great experience and product.

Anthropic is running circles around them right now, especially in the coding arena. Let's see what happens. I right now use Anthropic for everything, and I do not use ChatGPT for almost anything. One of my least favorite things in the world is ChatGPT clickbaiting me at the end of its responses. I've literally written a prompt, like a shortcut on my computer, to tell it off every time it does it, to get it to stop.

But I think that they're in a very good position—not in the tech community, but outside of the tech community. Their brand is pretty good outside of the tech community. I think that they have a ton of cash. I think they have a pretty strong team. I prefer to be contrarian than say the obvious answer. But Anthropic is so good. So good. Either give me more options, or give me an opportunity to be contrarian here.

Is Grok going to be a meaningful competitor?

Good question. Yes. Yes, it will. Same thing: Elon is a savage. There's a rule that everybody should know, which is, never bet against Elon. You just shouldn't do it.

What have you changed your mind on in the last 12 months?

Last year, MrBeast—Jimmy—was like, “Hey, I want to do this video where someone starts off with no money, no phone, no computer in the middle of America, and they need to start a business. I think that you would be a hilarious person to do it.” I was like, “Jimmy, I don't have time for this.” And he's like, “I'll do this many Speechify integrations.” I was like, “Okay, that's actually a really good deal. Let's do it.”

So, I spent 11 days sleeping in a Walmart parking lot in the middle of America. It took me 3 days to make enough money to get an iPhone and 5 days to make enough money to have a laptop. I started working at a company, worked at a pizza restaurant, worked at a gym, and the video essentially fell apart. There were a lot of challenges. It didn't work.

I was really upset about it. Typically, I take 100% responsibility—200% responsibility—for everything I do, and I couldn't figure out what I could have done more to make the video work. It took me a couple of months to figure it out, but eventually I did. I became resentful. I thought that part of the production team had said they'd do this and then they didn't, and then they'd said they'd do this and then they didn't. I forgot to be grateful for the fact that I was doing something so cool that could inspire hundreds of millions of people.

What I realized is, I often tell myself that my greatest advantage is that I have this AQ thing. I'm really good at keeping going when other people stop. I put more reps in than everybody else. But that's actually not my biggest advantage. My biggest advantage is I'm typically very grateful and I'm really optimistic.

In that moment, when things got really tough—around day 9, when I was very hungry—I forgot to be grateful and I forgot to be optimistic. I think that one of the core inputs for allowing someone to have super-strong AQ is that they can find the goodness in every situation. They can find a way to be grateful about things.

When I got into coding, I would misspell everything. An assignment that would take a normal person 4 hours would take me 17 hours. I had a notebook, and every day I would write before I went to sleep and right as I woke up, “I'm good at computer science. I'm good at computer science. I'm good at computer science.”

I found it really challenging, but I really appreciated that I got to learn because I wanted to. The thing I changed my mind on is that the most important thing is, even when things are shitty, you've got to find something to be grateful for.

Do you worry about wealth inequality in a world of AI?

All the time. Constantly. Constantly. Especially after that experience, I was generous with people who were living on the street before. Now I'm way more. I just buy people phones because that's the thing that I wanted. I'll go into the next convenience store, buy any Android that's there, and give it to them.

Because you just saw how difficult it was starting from zero.

Not having a phone is insane. It's literally madness. I came into the project with a full meal beforehand, and I have a lot of random skills. Eventually, I did succeed in getting an iPhone, but it was really hard.

If you don't have a phone, you don't have Google Maps. You literally can't navigate. You don't have Wi-Fi. You can't put a phone number down to get a job. You can't even drive for Grubhub or Uber Eats or do Deliveroo because you don't have a phone. Phones get stolen in those situations, too.

First of all, my personal opinion is that the government should give a free Android to everybody who wants one. You should be able to go to the Social Security office and just get a phone. It's a basic human right, in my opinion.

The second thing is, the Gini index is now getting really messed up. All the value is accruing to essentially buffoons like me who built a consumer subscription product or an AI product that gets a ton of people to pay for it, becomes a standard in their life, and, as long as you figure out the inference costs, you just print profit. That doesn't make any sense. It's too few people getting too much leverage, not to mention the hyperscalers. What happens to everybody else?

The solution is, don't be a Luddite about it. Embrace using these tools because, number one, you're going to get more entrepreneurs, which is an amazing thing. It's easier than ever to be an entrepreneur, which is, by the way, why I'm really long Meta.

In a world where it's easy to do engineering and easy to do design, more things launch and go to market, and it becomes the most important thing. The second thing is, how do you reskill people?

I do believe that you can get people who are old and grizzled and are veterans to learn how to use AI, but most people don't naturally do that. That's an unnatural behavior. Go back to the principle of a little bit of slope makes up for a lot of y-intercept. People just need to be awakened, and you need to convince them that it's fun to learn. You need to learn, but not everybody will do that.

That's the whole point of having a government system: there should be a way to re-educate people. Especially now, as you see K–12 kind of falling apart and universities falling apart, I think there will come a wave of people going back to school after the age of 40. But school will be different. I don't know what that looks like, but I think that this is the most important topic that we're not thinking about.

Would you suggest your children go to university?

I had the best time in university of anyone I've ever met. My cheeks started to hurt from smiling so much 3 months in. I found my people.

I was so lucky to go to a place where I had an open curriculum. Half my classes were classes I didn't have the prerequisites for. I got to make my own major, and I met so many people who have stayed my friends for life. I would just have two dinners every single night, and I'd sit with a table of 6 people I didn't know. I'd do that again. Actually, I ended up doing that again.

So, yes, I would, because it was the best experience of my life. But I would not tell them that the university is the thing that's responsible for their education. You are the quarterback of your own education.

You need to be constantly reading. The reason I'm the person that I am today is because, from the age of 13 till today, I have listened to 100 audiobooks every single year.

I will say that in the last 24 months, half of the content that I’ve consumed has been deep-research outputs from Speechify. I can go, “Hey, give me a comparative analysis of Catholicism versus Protestantism versus Mormonism versus Islam,” and then I listen to that 200-page document.

I’ll be like, “Hey, I saw this cool company called Etched. Explain what this company does. Hey, explain to me how to build a fusion nuclear reactor.” I’m just constantly listening to these things. My workflow is that I’ll ask a question, get back 100 pages, and listen to it on Speechify while I’m walking around. Everybody should constantly be doing this.

My thought is, when I have kids—which I’m excited to have—I want the teacher agents inside of Speechify to be amazing. They could skip all the classes in school. My brother and I skipped basically most of the math classes when we were in high school because we got bored of normal math.

Math is the only class that’s easy to skip because you just take the test, and if you pass it, you’re good. Everything else is subjective. But if you have an agent that asks you all the questions about the American Revolution and all the questions about Greek history, and it shows that you’re very accurate, that’s a signal that you can move on.

What purchase under $1,000 has been the most impactful?

There is a protein powder on Amazon called EarthChimp, which is pea-based protein powder. You take that and a 32-ounce shaker instead of the 16-ounce one. I have three scoops every morning and every evening, and I don’t need to worry about food because I get the amount of protein I need every single day automatically.

I’d say half my body at this point is pea-based protein.

And you—sorry, you don’t eat food in the day?

I do eat food in the day, but I travel a lot, so I have that 80 g of protein in the morning and 80 g of protein before I go to sleep, and I know I’m good for food.

80 g each?

Yeah.

Isn’t the absorption rate something like 20 g per hour?

Yeah—no. You see that in some studies, and by and large, that’s not true. If you talk to people who are bodybuilders, it’s just not reflected. It’s a very small optimization.

So you can ingest 80 and it’s not futile?

Yeah. You’ll fart more.

You fart like a trooper.

You would if you ate, yeah.

Yeah. With 80 and 80, dude? Dude, I don’t want to be next to you on a flight.

Yeah.

What’s the workout routine? And any tips for doing it while traveling? Because when I travel, I’m—

Okay, so a great new hack that I have is to take photos of yourself and upload them to ChatGPT. Ask it, “What body-fat percentage am I, and what do I need to improve?” It’s so good at giving you information now. Then ask it for a workout routine to target the parts of your body that you think you want to improve, and it’ll just give you the workout routine. It’s really good.

Most of my sport is tricking, parkour, and gymnastics, so I’ll go to a gym wherever I’m going. There’s a great one here in East London called Over Gravity. I’ll go train, and I’ll find the best athlete in the gym and go up to him and say, “Hey, what’s your name?”

“John.”

“John, I’m Cliff. Nice to meet you. You seem amazing. Can I hire you to coach me one-on-one? How much would that cost?”

“I don’t know, £75,” whatever.

From that point on, I’ll go train with them. Gymnastics I just love to do, so I’ll do it for 3 hours in a row. I love having somebody else who’s better than me watching me and giving me tips. It doesn’t matter whether it’s a sport I’m trying to learn, guitar, or whatever—it’ll hire someone to watch me do the thing and coach me to be better.

When it comes to lifting, obviously compound movements are great. DEXA scans are incredible, so you can find the breakdown between fat, muscle, and bone density in your body, and figure out exactly what your body-fat percentage is. It’s just all macros.

If you’re my size or your size, you want to have about 1 g of protein per pound of body weight. You want to have about 50 to 80 g of fat every single day, and then you cut or you bulk. If I cut, I eat 130 g of carbs a day. If I’m bulking, I’m eating around 260 to 270 g of carbs a day. Your body’s just a science equation. You can change it.

Have you ever done fasting?

I have done fasting for Yom Kippur, but I haven’t done it for health reasons. I’ve tried. It’s just very hard.

A bit depressing, I have to say.

Yeah.

And one meal gets you by for the day?

Oh, well, that is depressing. That is one of the worst things ever.

Okay, final one. What excites you most about the next 10 years? When you look forward, what do you think, “I’m pumped about this”?

I think humanity is essentially going through a new form of evolution. It took us millions of years. I mean, if you look at evolution, humans are what—the last minute on the clock?

We’re getting to a point where evolution is not a matter of your physiology, but of your brain’s ability to take in information faster, understand that information, and use it in a different way. I love fantasy audiobooks, and sometimes there’s this thing called a familiar in fantasy audiobooks—an animal, like a dire wolf, that can connect with you mentally and do things.

Basically, I feel the same way about my computer, my phone, and all the tools that I get to use. My team, by the way—it’s like Speechify is one living organism around the world. You can just achieve more.

Even if I just look at what happened to my dad with his experience with cancer and my brother, who’s still ill, we’re going to figure it out. I just have this Promethean view that we can solve everything.

When you talk to a doctor, you can very easily educate yourself to be sophisticated. When you don’t know how to solve your AC, you just take a video, and ChatGPT tells you how to set the AC, how to work your stove, and how to set up a fusion reactor in your backyard. You can learn everything. You can do stuff and build stuff more easily than ever before, and our kids are going to grow up with this mentality ingrained.

My dream for the world is that everybody believes they have a higher level of agency than we do today. Hopefully, that will be the case. The negative side of that is that people think they have less agency because they abdicate all the decision-making to AI. You need to have people be in the driver’s seat and use the tools instead of having the tools use them.

I think Sam Altman said it very well when he said older people use it as a replacement for Google, and younger people use it as their operating system for life. As you said there, “How do I make an AC work? How do I get better shoulder density?” Or whatever that is.

Cliff, this has been fantastic. Thank you so much for putting up with my wandering questions, but you’ve been amazing.

Thanks for having me.