SPEAKER_05: But this Gen AI movement is the actual industrialization movement of making more images, making more software, making more mock-ups, making it easier for us to access data quickly.And so I think with that, when you apply that back to software engineer agent, We can now, the same way we went from making one card eight to 10, we can 100x ourselves.And so me as a software engineer, I can work on 10 different problems at the same time or 30 different problems with the help of these agents.I think we're about to have the 1000x jump in software development.
SPEAKER_00: This Week in Startups is brought to you by Gusto is easy online payroll benefits and HR built for modern small businesses.Get three months free when you run your first payroll at gusto.com slash twist. Mantle, the AI-powered equity management platform designed for modern founders and operators.Get your first 12 months free at withmantle.com slash twist.And HubSpot.Join thousands of companies that are growing better with HubSpot for startups.Learn more and get extra benefits for being a Twist listener now at hubspot.com slash startups.
SPEAKER_04: All right, everybody, welcome back to this week in startups.It's your boy J. Cal here with my guy, Sonny Madra, GM of Brock Cloud.Go to console.groq.com and you can start playing with the fastest and cheapest inference in the world.Welcome back, Sonny.
SPEAKER_05: Good to be back, J. Cal.We missed you on the all-in pod.What happened?
SPEAKER_04: Well, yeah.I mean, a lot of people are asking me because I never miss, right?It's very, very rare that I miss an episode.
SPEAKER_05: Yeah.You had the Iron Man going.I think you had the Iron Man going.
SPEAKER_04: I mean, I'm pretty good, generally speaking.What happened was I was here in Austin, and I went to the Salt Lake, and they have this great bison rib, and I'm chewing on it.I mean, I wasn't chewing on the bone itself, but it's a chewy rib, and I heard like a crack, and I thought, maybe I just, you know, when your two teeth crack together or something, I didn't think much of it. It wasn't hurting.And then two days later, I'm flossing and the tooth kind of hangs halfway out.I'm like, holy cow.And so, you know, then a couple of days later, I find an emergency dentist.I get there.They're like, oh, you know, that root canal I had 20 years ago.You know, those kind of deteriorate over a couple of decades.
SPEAKER_01: Yeah, of course, of course.
SPEAKER_04: And, you know, it's basically it's cracked.And so then I was able to get emergency surgery.They extracted the root of the root canal and then put in, I guess, what they call an implant or something.And then I guess I got away three months.It heals and they put another crown on it or something.But yeah. you know, for three days, I don't like taking pain meds.And, you know, they gave me some pain meds and that screws up my sleep.And then the pain was screwing up my sleep.So I just like really was in a fog and I tried to go and I'm like looking at the docket and I'm like, I can't even read the docket and my mouth is swollen.
I'm now four days or five days past the surgery.But at the taping of all, and it was like whatever, 48 hours.And I was just loopy.And I was like, you know, I don't want to like disappoint the audience and be loopy.So yeah, I just took a, yeah, I took a day off. Which, of course, is always great because the MAGA people are like, take a permanent vacation, Jake Allen.And other people are like, I'm not even watching the show.So to my MAGA friends who absolutely are thrilled when I'm not on, they're like, oh, more time to talk about Ukraine and Biden, Biden, Biden.But it was a good episode, actually.So great job.
SPEAKER_05: I only made it about three quarters through.But yeah, it was fun.But it was missing the moderator.
SPEAKER_04: How are you doing?I know you've been traveling and growth continues to grow.
SPEAKER_05: Yeah, you know what?Maybe let's like, it was a big week this week.Let's do a little bit of news before we do demos.There's kind of like three important news stories that we want to highlight coming out of this week.And, you know, I had a kind of a tweet about it.So maybe we'll use that as like the grounding here. And this one's kind of related to us, but we won't talk about Grok here.But three major companies launched chips this week.And this is really important.And the three companies were Meta announced a new chip called MTIA, Meta Training and Inference Accelerator.
Google made some of their chips available to the public out of their conference their tpu v5 and then intel launched the gaudi 3 ai chips for mass production and q3 and um that's really interesting because these are three major companies that are basically out there you know putting new hardware into the ecosystem um and you know let's uh what are your thoughts when you when you when you read that what what do you make of that
SPEAKER_04: Yeah, so it does seem there's a number of things that occur when we have a boom like this.You know, there were people who've been investing in this technology all along and they've seen, hey, there is a path here and Nvidia comes to mind, right?Like they saw there was a path here, but they were working on video games, crypto, and then AI.And then of course, crypto kind of plummets, video games still growing on a pretty good cliff. But of course, AI then becomes a spoon.And when people see that, there are two drivers in basic human behavior.One is fear and one is greed.Right.Yeah.And so if you pause and you think about, you know, OK, you're Intel, you're Google, you're Amazon, whoever it happens to be.
Meta probably has a fear. or I should say they see an opportunity having their own chips.What's the opportunity there?I think they're probably opportunistic.Hey, we have this data.If we have these chips, we can save money.So sure, we're backing up the Brinks truck.I think they're one of the biggest... customers of nvidia yeah you know hey this gives them a little bit of a hedge hey you know if we have to negotiate prices with nvidia you know we do have some of our own ships being deployed here and then for other folks it could just be complete fear that they're going to lose their business you know like intel right um and that other people are going to run away with it so There's usually those dual motivators.
And what's good for everybody is most of these projects will probably fail to be significant.But in aggregate, they're going to drive prices lower and the competition will make everybody sharper. So you could always see in one of those tweets, I guess it was Intel was claiming, you know, they're all going to make claims.Hey, we're better on this task.And friends, we're this percentage better.And that's what you really want to see in a thriving economy is competition.And here we go.Competition's here.
SPEAKER_05: Yeah, I'll just add like one more thing on that particular story.I think that holds true.You know, we don't really see any of these chips in production or external benchmarks just yet.So that's, you know, kind of TBD, which means that it is sort of like a, you know, maybe like you said, like an internal thing slash pricing pressure thing. The other thing that I thought that Google did at their conference was not only did they launch like an AI chip, which is all the rage these days, but they also made available a new kind of CPU, which is an ARM-based CPU, which is interesting because that's them kind of going more directly after the traditional compute stack with a chip that they've created.And so it's really fascinating how the future world of compute is playing out.
SPEAKER_04: Yeah, and if you look at someone like Facebook, they have really pushed open source hardware, commodity hardware, because they realized that the business they're in... They were the first, actually.
SPEAKER_05: They were the first back in the day.
SPEAKER_04: Yeah, so they've pushed very hard for decades now on just making servers commodities, data centers commodities, the infrastructure and data centers a commodity, because that serves them. Right.And I think Google participated in that as well.So you have the people who have a consumer application like Google search or Instagram, Facebook, a social network and the advertising businesses.They don't care about the hardware being commoditized.That's better for them.Now, NVIDIA may care very much.Grok may very care very much. And so this just makes everybody sharper and better at their jobs.And I think that competition is going to become very real.
But like you're saying, are these in anybody's hands yet?It takes years to get these things into people's hands, right?
SPEAKER_05: Yeah, this was a big project that, you know, Facebook started, you know, back in 2009, right?So it has been over a decade, which is incredible.
SPEAKER_04: Yeah, almost 15 years now.
SPEAKER_05: Wow, crazy.
SPEAKER_04: And that really has driven a lot of server-based and cloud-based computing is this commodification of the hardware sack, right?And so, yeah, great.Awesome.It's going to be great for the industry.And there's other problems that need to be solved.And then I guess we'll get into some of those now.
SPEAKER_05: Okay, switch gears.Two more quick news stories.So Adam Schiff.So this one is kind of near and dear to you, right?Because he has put something on the table to basically require people to disclose, right?It's called the Copyright Disclosure Act, and disclose where they got their content from. You know, this is an interesting one because it's very near and dear to you.And we've talked about this a lot.We have a demo coming up, which I think is an incredible demo, but definitely does cross into some of these lines.But what do you think about this law?
SPEAKER_04: I think this is sort of what you've been saying for a long time when it comes to... Yeah, I mean, I hate to say I told you so, but I told you so. Listen, as a founder, there are things I love doing, like building products or meeting with partners, hanging out with my team and dreaming up new ideas.And then there are chores that I don't want to do.I don't want to do HR.I don't want to do payroll.I don't want to deal with all that.So I use Gusto. Gusto is the best for payroll, for HR services, and for running a small business.It makes everything so much easier.Even a mid-sized business, man.
I get a lot of portfolio companies that are pretty sizable using Gusto because it is designed for you, the small business owner. And payroll is something you definitely do not want to mess up.You've got to get it right.And Gusto is going to make it perfect for you by calculating paychecks perfectly.Also, payroll taxes.You've got to get your taxes right.You can't make mistakes there.And you want to set up open enrollment.You want to be good to your people.Gusto handles onboarding, health insurance, 401k, time tracking, commuter benefits, offer letters.
And they even give you access to HR experts.So Gusto takes all of this off your hand and lets you focus on important stuff.Your product and your customers.It's super easy to set up and get started.And if you're moving from another provider, Gusto will transfer all your data for you.Here's your call to action.Because you're a Twist listener and you're part of the family, you're going to get three months free.Incredibly generous.Totally unnecessary.Thank you so much to our friends at gusto.com slash twist.
You must go to Gusto, again, gusto.com slash TWIST to get free months free.Thank you, Gusto team.You can't just take people's work and build a derivative product out of it without their permission and without their consent, without, you know, paying them.And so here we are.The industry has stolen content. and they've explicitly stolen it, and they knew they were stealing it.And so opening AI, it's come out, has stolen tons of content, they built scrapers to do this.And their concept that it's fair use is laughable.You as the person who created this IP, you have the right to create the derivative products from that IP.And the fair use doctrine is very narrow, and its purpose is very clear.
educational to make society better, right?A small portion of the original work and not interfering with that person's ability to monetize it.You can look up the four part test.There's little exceptions.So, you know, if you're doing an educational project, you get a lot more leeway.An example of that might be if I wanted to do, you and I decided we're going to do a critical analysis of Blade Runner.We could sit here and play clips from Blade Runner. But we can't create Blade Runner 2049.And so people will use like a really silly argument like, oh, well, this is how humans learn and the computers learn differently.None of that really matters.
The way intellectual property works is you have some amount of ownership of it.There are fair use exceptions for education and using small amounts of content.But that doesn't give you the right to create the derivative product and exploit future products unless you have a license. So what I like about this bill is it just says, hey, tell us what you used.And that seems more than fair.If you used my content to build something, I should know.And so what I hope happens here, and I'll just end on this, is I hope that New York Times gets an injunction against OpenAI and they have to pull their product from the market.And I think the New York Times, Dow Jones, Disney... Yeah, I think they should all get together and gang up on open AI at this point right now to protect the future.And this idea that, oh, it can't be done and technology always wins.
Tell that to Napster.Tell that to Scour.You can stop this.You must be coordinated in order to stop it.The music industry is a highly coordinated group. What has to happen is journalists, publishers, and content creators on the internet have to join that group.And as a united front, they should get together and they should not stop with just a settlement with OpenAI.They should force them to remove the product from market And they should force them to start the product over again.Because if you stole all that content and you made the model, and then you're saying, oh, well, now we'll leave you out of it.
You've already created the value.And $100 billion in value has been created.And I think $2 billion or $3 billion in shares have been sold.So essentially what's happened is they've taken the value of the New York Times and other publications.They've built a company.That company has perceived value in the market to the point at which the most sophisticated investors in the world are buying it. At a $90 billion or $100 billion valuation, that money, a portion of that belongs to the New York Times and whoever else they scraped.And so, you know, hopefully this time the content industry will act in unison and not fold.That's my belief on it.
SPEAKER_05: I think you and I are a little bit different sides here.I wouldn't take that open AI should go down, but I do think, you know, a lot of those organizations struggle and they should get at least some share of the revenue or something like that, that being generated while everything else is being settled.But that bill is going to be interesting.And, you know, let's put another bet, J-Cal. Because, you know, since we're, you know, we're on our way to 100,000 bets, we're probably past it, but who knows?This is just take a side.Does this bill pass or does this bill, you know,
SPEAKER_04: Yeah, or what version of it passes.I think that a bill will get passed.I don't know if it'll be this one, but I do believe a bill to do this will get passed because it's going to take 10 years to litigate this stuff, or 5-10 years for this to go through the court system.
SPEAKER_05: But NAPSA didn't take that long.
SPEAKER_04: Well, they were a private company sued into oblivion.So, you know.Yeah.So, I mean, here, I think OpenAI, they have a lot of resources.And so part of this is they can fight for a long time, whereas Napster didn't really have the ability to fight for even a long time.Even YouTube.The reason YouTube was sold to Google was because... So they could fight the lawsuit.And that took years and years and years.And so... you know i do believe that there will have to be some new laws um around a training specifically because we're sitting here and we keep debating is it legal to train on other people's data um i guarantee you it's not legal um and i guarantee you it will be found to not be legal to use other people's content to train a new intelligence that you can then go exploit that work and so well maybe we'll come back to this bet we have to think about it a bit more okay we have
Just think about how to frame it.And it's not that I want to see open AI go away.I do believe the technology is valid.But if they could build this technology without the content, they would have.And all the papers that you and I have read and discussed say the more data you put into it, the better the output. And that's obvious to anybody who uses these, that it's stealing other people's content in order to give the answers.So if it's giving answers based on content, it found that the New York Times or YouTube videos or whatever web content or books, and they know they stole it.They did it covertly and they're trying to cover it up.And I think they need to basically redo all the models.They'll have to start over.
And I think that's what... The New York Times and Wall Street Journal, Dow Jones, Disney, they should all Viacom, all the major content creators, they should get together and they should, as a group, sue OpenAI into oblivion and make them pull the product from the market and then make OpenAI get permission and then have them start over.That's what I think should happen.Got it.
SPEAKER_05: Got it.
SPEAKER_04: And that's my advice to content injuries.You're going to lose your entire business.They're going to lose their entire business.Just like Google took over the entire ad business.
SPEAKER_05: It's a perfect segue.You know, Jayka, you always manage to do it.I didn't even set it up this way.But there is a new service that just got launched this week that's backed by tech and music heavyweights called Udeo. U-D-I-O.And I'm going to pull it up here.You can see here, it's a new service that's backed by tech and music heavyweights.And it has powerful capabilities.And I'm going to go right into it, J. Cal.Yes.
And I did this song in the style of Dire Straits about New York City and big money.And J. Cal, I'm going to play this for you.Okay, here we go.And I want the raw reaction.
SPEAKER_02: Okay, I'm going to get it.He makes it on his own.
SPEAKER_01: And drive on to where the skyscrapers soar.Yeah, oh, earn more, earn more.
SPEAKER_04: I mean, it's so impressive.It's not exactly Dire Straits.That was more like, I forgot the person.Against the Wind?Who sings Against the Wind?Tom Petty?Against the Wind.No, Against the Wind, Bob Seger and the Silver Bullet Band.Oh, yeah, yeah, yeah.A little bit more Bob Seger stuff.
But the fact is, if you played me, if that song was playing on the radio and you and I were driving down the road, I would not know it was AI.Yeah. I might think it was cheesy when I listened to the lyrics, but we are now past the uncanny valley.This is absolutely of the quality in which it would pass the Turing test.In other words, you would think a musician, a human musician had written it, but that does sound so much like Bob Seger that I'm surprised it didn't get the Dire Straits piece.And I'm excited about this product to exist because I do think... We might get some Rolling Stones songs or some Dire Straits songs or Bob Marley songs or Jimi Hendrix riffs that, you know, I know it sounds sacrilegious to people that you might enjoy as much as the originals.Look, business leaders face a maze of tasks today.We all know that creating and managing your company's ownership shouldn't add to your stress.Well, meet Mantle.
This is the AI-powered equity management platform for modern founders and operators.It's going to simplify your strategy and save you a ton of time.Mantle's been built from the ground up by founders for founders, and they've spent 12 years building and scaling successful companies themselves, and they've seen every mistake in the book, and they've solved for it.You can model your price round, you can update your equity documents, you're going to understand your dilution, and it's designed for ease of use across all stakeholders.Powered by Mantle AI Assistant.Man, it's fast. For example, you just drop in your term sheet and you watch the platform generate a pro forma cap table for you in seconds.This used to take, oh my God, you would ask your attorney, it'd take a week and then it was wrong.And now it just gets done instantly.
And this will allow you to focus on the things you need to focus on and not worry about your cap table.So here's a call to action.Visit w-i-t-h-m-a-n-t-l-e.com slash twist to get your first 12 months free. And you're going to lock in an exclusive rate of $100 a month after your first 12 months.That's withmantle.com slash twist for your first 12 months free.See why hundreds of founders are switching to Mantle right now.
SPEAKER_05: You missed the beginning there because I think I was playing it before.So I'm just going to do the beginning of it as well, the first 15 seconds, because I really thought it was great.
SPEAKER_01: All the folks around him say he's too bold But he takes it and makes it on his own man.
SPEAKER_04: I mean, it's, um, yeah.I mean, it sounds a little like counting crows, maybe.Okay.And crows in there.So yeah, I, I, I like it a lot.
SPEAKER_05: Yeah.And so look, they've done an incredible job here.Um, of all the ones we've tried and like, they don't, you know, have a, like a sort of like you can put in style of, you know, a band and, um, Um, they've really made it work.They've really put it together and it's super, super impressive.Um, and you know, it has a bunch of X steep mind folks that started it, um, you know, backed by one of the Instagram co-founders and they 16 seats.
SPEAKER_04: So when you type in the name of it, though, it should, if you say I want it in the, so is Dire Straits licensed their music to them?Is Bob Seger licensed to them?Do they have the rights to people's music?It's just trained on, you know, Columbia Records and Bob Dylan's catalog.
SPEAKER_05: Will.i.am is involved.And, you know, he's kind of been at the intersection of tech and music for a few years, right?And there are many years now, but. In the article that I pulled up, they didn't say they didn't confirm or deny if it's been trained on copyright music without permission.
SPEAKER_04: That means it has been.Yeah.All these people are guilty and it's all going to come up.You might as well all fess it up right now because you're going to get busted because you sent a scraper out and the scraper scraped it.There's a tracker of the scraper.Unless people who are in these LLMs are going back and erasing all the code they did and then telling their employees to lie, the employees are going to know. And there's some employee who had a conversation, just like when that woman from OpenAI with Sora was like, yeah, I don't know exactly what it was trained on.And it's like, I think she's the CTO.
SPEAKER_05: The CTO, yeah, yeah, yeah.
SPEAKER_04: If you don't know, that's like going into the back of the restaurant and you're like, who cooked this steak?And there's like 10 chefs and they're like, we don't know. It's like, but you're sitting there cooking a steak in front of us.
SPEAKER_05: There's the head chef.There's the head chef.
SPEAKER_04: The head chef is literally cooking a steak and the head chef is like, she's like, I don't know who cooks steak here.It's like she's flipping the a stake over as she's saying it.It's like, well, you know, you're lying.And, you know, the question I have is I don't believe the music industry has the right to this.In other words, if you're Columbia records and you have a relationship with Bob Dylan to publish his music, do you have the right to create virtual Bob Dylan?I don't, believe you do and so i think there's going to be another layer of lawsuits which is bob dylan's going to be able to sue columbia records if they try to license his lyrics to one of these services and his voice to make future music and i think bob we should not allow record labels to own people's entire likeness and ability to write the next set of songs without their permission
SPEAKER_05: The Rolling Stone article is interesting.They definitely talk about, they don't say whether they did or not.They do say they have AI to detect if a song sounds too similar to an existing piece of work.And so this is the New York Times lawsuit kind of thing where they got it to spit out the exact words of the article. And so maybe that's the, you know, you remember that famous lawsuit with Vanilla Ice and Ice Ice Baby?And then he took that, you know, beat from, I think it was like a Sting song or wasn't it?Or something like that.Well, no, no, that was Queen and David Bowie because that was under pressure.Yeah, you're right, under pressure.Yeah, yeah, sorry.
Yeah, I got that wrong.And he said, oh, no, I changed the beat by one, right?Yeah.And then he got in trouble because he didn't change it enough.
SPEAKER_04: let's put things into two different categories.There's a human being who's inspired by another human being to create something.Yeah.Vanilla ice.Right.
SPEAKER_01: Yeah.
SPEAKER_04: So whatever you think of him as an artist, he has a collective amount of music he listened to his over his lifetime and he created a new song and either he unconsciously channeled that or he explicitly copied it and edited a little bit.Okay.So now that one human has created one piece of output based on some number of pieces of input.Okay.That feels like a certain scale and then society can debate that. the output there and did he steal it or not?Right.And how close is it?And there's a test for this, which is, is the audience confused?
So if I were to create, if I were to create a star Wars film and I called myself, you know, Darth J. Howe, and I made a whole, uh, you know, character around myself as a Sith Lord, et cetera.And I had a lightsaber and I had, you know, two droids who, you know, went around with me and made funny jokes and were, you know, and I was up against, you know, a Jedi.And then I had a little teacher who was like Yoda, you know, like the audience would be very confused by it.And they would assume that this was in some way Star Wars.And so now we look at wholesale, right? So now you have two examples there, Ice Baby and then me creating my own little short film.In both cases, you can sit there and discuss, hey, how confused is the audience?Now imagine you take every film ever created, every song ever created, every lyric ever written, every story in the New York Times, every book ever published by Harper. in the library and then you say now we can create all output at all time and i as the person who created this llm get all that economic value these are profoundly different things and so for people to use the analogy of ice ice baby is just so ridiculous because you have but one artist making but one song and the courts and and society can you know debate that taking everything ever created
and putting it into an LLM, which is what Sam Altman and OpenAI is doing.This is what they want to do.This is what Google wants to do.They want to take all the content in the world ever created, and they want to get the economic value from it.That is completely different. than writing a short story inspired by Stephen King that might use one of his characters.And even in that case, you probably would have to settle and license, you know, the content or license like the hip hop industry did.We have a great analogy because the hip hop industry has been sampling songs forever and they just pay a small licensing fee.And the people who have their music licensed by some hip hop artist, which has a tradition of doing these things, they love it.Yeah.
Like for the love of God, please license it more.I I'm sleeping and you take something from queen and under pressure and I get paid a royalty.
SPEAKER_03: Yeah.
SPEAKER_04: Great.So, but the, you know, the technology industry doesn't even want to give that inch. That just shows you, you know, I don't want to use a really colorful piece of language here, but it just shows you how unethical our industry is.Okay.Okay.
SPEAKER_05: Well, look, you know, it was nice to see some industry folks involved.Yes.Hopefully they find a path to it.It's really good.That's great.
SPEAKER_04: Yeah. Oh, I give this an A. I give it an A. Yeah, just straight up.It's great.And if you type in, I mean, at a minimum, if you type in Dire Straits, it should say, you know, we don't have permission to use Dire Straits' music and like this in here.You can describe other things about music you like, but it shouldn't tell you that it doesn't have the rights to it.And if it does have the rights to it, it should say we have the rights to it and explain that.And this is what I think. What's great about the bill that Adam Schiff is proposing is it's just looking for disclosure. just be honest about what you're doing here.And our industry could do, the reason our industry is not being honest is because they're liars and they're thieves.
If you weren't stealing and you weren't lying, you would tell the truth.We didn't train on this.When people don't tell you in our industry what they've trained on, they're liars and they've stolen it.So just that's the ultimate tell.Because if they didn't steal it, they'd say, no, no, no, we trained it on only stuff we've licensed.You really think that these super intelligent people at OpenAI don't know what they trained it on?Really? These are the smartest people in the world.
SPEAKER_05: Let me just give you the opposite side, just for the sake of argument.When you let something loose onto the internet, like a crawler, and you made a crawler and you had it go look at videos everywhere.Now, the question you could say is like, hey, did you hit YouTube?That you should know because you shouldn't have been there or not, right?
SPEAKER_04: Yeah, it's a URL.You could just look at the log files.
SPEAKER_05: But what if there's YouTube videos embedded on pages, which they are all over the place, and it watches those?
SPEAKER_04: It's literally in the code.It's literally in the HTML code.You would see it there.It would say, embed YouTube.Let's be honest.You're actually arguing my side of it, which is these are highly technically sophisticated people.They know what an embed is.They actually would know bad content.So all you have to do to bust the OpenAI... stealing is to look at what they excluded and ask them why did you exclude this and if you said to them hey why did you exclude this and they'd say oh those are just like we happen to stumble upon with our crawler or some random it's like porn sites most likely right yeah and they said oh no we didn't want to have porn sites in here we didn't think that was really good content or
Although this was a different language and we were doing an English language once, so we didn't need Japanese, so we excluded it.Or this was just machine language.This was just code.So we left out programming code because we're not doing code in here.We're doing whatever content.This is an image database.And so you could just ask them, what did they include?What did they not include?And the crawlers break... the crawlers are optimized like google crawler is optimized they're optimizing it all day long there are there are probably thousands of people in the last year who have optimized the google crawler yeah i would say probably a thousand people have sat there and written code to make that crawler more efficient so to pretend that they just like oh we just set it loose on the internet and we don't
No, what came back is the height of arrogance would be the height of arrogance for them to say, right?Hey, everybody, I am obsessed with AI right now.You know that every Tuesday we do a bunch of AI demos and we're trying to figure out on this very program, how do we take all this AI potential and make it a reality?So when we're running our startups, our startups are more efficient and we get more done with less and we delight our customers.I just read a report by our friends at HubSpot And they're talking about all the different ways to use AI to improve specifically sales marketing and customer support.Now, if you're running an organization, you know that you live and die by yourselves and that marketing drives yourselves and that customer support is how you keep your customers.These are three of the most important pillars in your startup.And the team at HubSpot did something really interesting.They surveyed a thousand early stage founders.
They asked them, how are you using AI to do sales, to do marketing, to do customer support? First, using AI to segment your customer list.What a great idea.Second, personalizing everything with AI from your website content to SMS to email.Third, how do you encourage your team to get creative with AI?And fourth, how do you balance between investing in AI-trained experts, right, and developing your current team?Now, you're going to read this report and you're going to get all of this information.These weapons for massive growth, WMGs, are just waiting for you.If you're in your podcast player, check out the link. in the episode description.
And if you want to save 70% on HubSpot for startups program, head over to HubSpot.com slash startups.That's right.HubSpot loves startups.They were a startup themselves.Visit HubSpot.com slash startups to see what discounts you qualify for and start using their powerful solutions and pricing that will break the bank.HubSpot.com slash startups.Yeah, I mean, the whole idea is because something's open on the internet, Sonny, doesn't mean it's yours. It does not mean it's yours.My car is parked on the street here.It's on a public street.
You can walk up to it.You can sit on the hood of my car.It doesn't mean you own my car and you take it.Yeah.Yeah.And that's the position that open AI is taken.We trained on the open internet.Yeah.Society's open.You know, my front door might be open right now.
You walk in, doesn't mean you can steal from my house.
SPEAKER_05: Yeah.No, it's interesting times where they are for sure.
SPEAKER_04: I feel like I'm... Why do I sound like a lunatic trying to defend copyright?It's so funny in our industry how people trying to defend basic... But you know why?Because many people in our industry...
SPEAKER_05: Many people we know have benefited, right, from the growth of these companies, whether directly involved or indirectly involved, right?And look, the flip side is we don't really pay to use Google.We don't pay to use YouTube.We don't pay to use any of these services.And so the company has to sit there and say, well, we have to do some of these things because that's how we make our business work, right?And so... It's a really interesting time.The whole notion of training a model and the arguments to be said, well, that's just like a person going and watching all the videos of guitar players on YouTube and then all of a sudden creating a style of their own that's in the style of Jimi Hendrix.Do you owe anything to Jimi Hendrix?Probably not.
But it's different if you're doing it as a commercial entity. But if you went and become famous, it's fine.That's the thing I can't really, you know, no one could go to an artist and say, hey, you were inspired by watching YouTube and you got here and you now owe all these people royalties because you've become a, you know, a multi-platinum or, you know, however the category is.
SPEAKER_04: There is an example of this.If you were to go and take all of Dire Straits records and create a Dire Straits cover band And then go on the road.And there's one called the dire straits experience.And I wanted to see them because I would, they literally do the entire album of alchemy and they call it re alchemy, which is my favorite album of all time.And I've really wanted to see, but they're only touring Europe.And I just, that hasn't been in sync. but they have to pay a licensing fee to do that.So there is literally an analogy here, which is if you do a cover band, you have to pay fees.Every time you perform, you're paying licensing fees through royalty organizations.
If you wanted to create a version like I used before of Star Wars, if you wanted to make a Marvel comic, if I wanted to take one of the X-Men and make my own series based on Cyclops, I would need permission, right? And so you could be inspired by, I think we all know what that is.And then when you take the entire corpus, that should give you an indication that if you took the entire corpus, yeah, maybe you are stealing.Yeah.Yeah.Why would you not get permission?That's the other thing is like, why wouldn't you get permission? This is like a reasonable thing to do.
SPEAKER_05: I'm going to show you this.I tried to do this one.Maybe I missed it.
SPEAKER_02: What did you give a letter grade, by the way, to this?
SPEAKER_05: I'm going to give it an A, but I noticed, I just want to share one thing because I want to give these guys some credit.
SPEAKER_02: UDIO.com.
SPEAKER_05: I'm proud of them.Yeah. And I was doing this because I thought we would be able to, we could get this song here.I said, a song in the style of Dire Straits, Sultans of Swing, about a friend that has really passionate copyright protection.Oh, no.Oh, no.And when I did create, it did give me an error a second ago saying, you can't do Dire Straits.It did?
SPEAKER_04: Yeah.
SPEAKER_05: Fascinating.Yeah.So.
SPEAKER_04: So maybe they are thinking about these things.
SPEAKER_05: Let's play this one real quick.Guardians of the score.
SPEAKER_04: Guardians of the score.Here we go.Okay. it's hilarious i mean it's hilarious i mean right now it's super funny and i think um there's gonna be great art created with this that's what i'll say i do think we're gonna see some great things come out of it so i don't want to say i don't want this product in the world yeah i just want to see it be fair so if you're going to do this for dire straits it would not kill the person who is making this piece of software, it would not kill them to get permission from Dire Straits, and it would not kill them to not have permission from Dire Straits and have every other artist.So like, remember... what's her name?Um, Taylor Swift didn't want to be on Spotify at some point and she didn't have her music on there.And I was like, yeah, that's totally fine.I mean, and Daniel Lack was like, well, no, you know, and he kind of got into it on Twitter at some point.
And I was like, no artists who don't want to be on your platform don't need to be there.It's their choice.So let them not be there.And you know, you have to cut a better deal with them if they're a better artist and it would not kill this service or any other service to get permission to to pay and to pay a term that that person felt was great.The Rolling Stones might say, yeah, well, the Beatles held out for a long time, if you remember, to be on iTunes.That was like a big deal when the Beatles finally decided.It's their art.It's their content.
SPEAKER_05: They had like a whole ad campaign about it.
SPEAKER_04: Yes.And you know what?As it should be.It's their content.They get to negotiate it.And you know what?All these people are hypocrites because if you stole the... the Google algorithm and made a new search engine and you sold their index, you could be sure they would be all over you.Absolutely.
SPEAKER_05: Fascinating times.All right, let's go on.JCal, you're really going to like this next one.
SPEAKER_02: Okay, I love it.
SPEAKER_05: It strikes me as like a JCal daily.So I think this one might make it into the daily workflow of JCal.It's called Aqua. And what it is, like sort of like a note-taking app that doesn't just do straight transcription.Okay.And so it lets you talk to it, but then as you're making the edits, it will fix what you're saying.And so you could kind of, let's say you're trying to formulate a thought, and then you're like, oh, no, no, go back and change it.So let's give it a try here, okay?
SPEAKER_01: Mm-hmm.
SPEAKER_05: All right.You know, I had this meeting with Jason, and we were talking about copyright protection.Actually, I could start all over.But really, we had this meeting, and it was about buying new GPU chips.And I was trying to convince Jason that Grok chips were the best chips.No, no, no, just start all over again.
SPEAKER_04: So it understands just start all over again.It might understand like when you use notepad or using the transcriber, a new paragraph, that kind of thing.
SPEAKER_05: Exactly.But not just that.Right.And so here's the example.You can say, remember that meeting?It's on Thursday, Friday, Friday.And so it knows that you meant Friday and you don't have to say start all over again. Or like, you could say imagination is more important than knowledge.Albert Einstein put quotes on that.He knows how to go back and do that.
SPEAKER_04: Right.So this is like, it's what I like about this is transcribing has worked pretty well.And we're finally starting to see the gains from it.And people use it.I see people all the time composing text messages by hitting the microphone.It transfers what they're saying.But you do want to bold something.You want to make something a link.You want to make a new paragraph. And somebody coming up with a way to do that without ever having to touch the keyboard is great.
So yes, I give this a B plus.It sounds like a really smart tool that needs to be made.If you could start talking to say, you know what, can we make a bullet point list out of that and see how that looks and then make it a table as well.And then show me the table.
SPEAKER_05: Let's do it.So let's do it.Let's give it a, let's give it a try here.Hold on.Let me bring it back up.
SPEAKER_04: Yeah, just say, you know, my favorite foods are, and give it a bunch of foods.And then say, can you put that in a table with the calorie count and see what happens?
SPEAKER_05: All right.I want to make some pizza tonight.The items I need are dough, pepperoni, cheese, pizza sauce, and peppers, onions, tomatoes.Actually, can you just put those items in a bulleted list?
SPEAKER_04: Okay.
SPEAKER_03: Yeah, that's pretty good.
SPEAKER_04: That was pretty sweet, I have to say.So the LLM was listening, and for people who are watching, it went from a paragraph just to a bulleted list.I mean, that was pretty nice, I have to say.Dictate, edit, and transform using natural language.I give it a B+.Yeah.
SPEAKER_05: oh, I thought you were going to upgrade it after the bulleted list.
SPEAKER_04: No, I mean, I think this is what I expected it to do.I feel like this is something that will be a notepad.And so, you know, as a startup, you know, my advice to with Aqua is to really start to think about, you know, what's defensible here and who they're um market is because i kind of feel like this is something that um notepad grammarly are going to be able to add pretty quickly uh but i do think that they've just you know they got like um a couple of uh months ahead of them like they're six months ahead of them so i do wonder you know who this is for i'm not i'm not going to stop using grammarly or my notepad to use this but i'm i'm going to play with it i will play with okay
SPEAKER_05: Okay, I'm going to give this an A. And I think what they should go all out on is, and like I said, you know, they should go all out on just dominating this, like becoming the best app that does it, like sort of in those early days of apps, because I do think, look, they have a lead, and I think they could create an experience, which The main reason I don't like using voice to text, which I notice a lot of people do now, you know, these errors happen.Then you're like sitting there editing it.You're like, I wish I just typed this out.But if I wanted to get a thought out and kind of the way that this allows it, this would be really, really powerful.You know, I'm going to try to put it in my daily workflow.
SPEAKER_04: Yeah, give it a shot.I mean, I think Grammarly works really well for this, and that's why I started with Grammarly.
SPEAKER_05: How do you use Grammarly?
SPEAKER_04: So I will talk into the Grammarly app, or I will talk into the Grammarly webpage, or I will use the Grammarly keyboard on my... phone and so yeah you know grammarly is just very good at cleaning up your pros and giving you some options and ideas of how to make it better and so i pay for it for my entire team yeah same here i think it's great i find my whole team is like it's less embarrassing the idea of being embarrassed by something somebody sent because they didn't use grammarly is kind of ending now which is great like people would say sometimes i'd see somebody send an email i'd be like what the You work for me and you're misspelling this word or you're like, it's so sloppy or it doesn't make sense what you're saying.Now you use Grammarly.It's like really hard to be, to write something bad in Grammarly and get it past Grammarly is hard to do.Like to have a mistake in Grammarly, I think it's impossible. I didn't know that they had a, I'm going to try.I have it.They don't have these navigation tools.They don't have that.
But I do know in Siri, you know, when you hit the microphone keyboard, the microphone icon on your keyboard, on your eye, I don't know what they call that dictate.The Apple dictate tool has gotten better.And if you do say new paragraph or strike that last word, like some of those work, but not a lot of, not like this, not like, Hey, start over.Not like, Hey, make it a bullet point list. So I agree, they should keep working on this and see where it takes them.
SPEAKER_05: It's a world-class notes app.I think everybody would use it.It'd be awesome.
SPEAKER_04: Yeah, I mean, every couple of years, somebody finds a new vector.Evernote was the one that was cloud-based and mobile.Grammarly is based on AI, machine learning, and making you a better writer.This one is based on formatting.So yeah, there's always an opportunity with note-taking apps.I agree. but yeah, I give them a big plus.Keep going.
SPEAKER_05: I'm going to give them an A. I'm going to give them an A. I like it.Great.Keep going.Okay.So the next one on our list is, um, so we saw a couple of weeks ago, Devin, which was the automated coding that had like, uh, sort of multiple workspaces to work on a problem.And this team has basically made an open source version of it.And so if you go to SWE-agent, this team has come together.It's an open source project. And they compared themselves against some of the bigger models out there.
They have just a demo in place here.I didn't want to do this from scratch.And so you can see here they have like a 30-step demo.So it's like they have an issue that they pull from their repository, basically.They have it create like a... a bug, they reproduce the bug, and then they move the code around.So I'm just going to go through this quickly.And so it reproduces it.And look, it's open source.And what's awesome about this is they've made this available to everyone very quickly, shows the pace of development we're at.
And I think folks are going to take this and really kind of build on top of it very, very quickly. And, you know, it's kind of just walks through the same workflow that we saw.
SPEAKER_04: Tell me the name of this project again.This is called?
SPEAKER_05: SWE-Agent.
SPEAKER_04: Ah, so it's a software engineer agent.Agent, yeah.And it's now open source, hosted, and you can go play with us at SWE-Agent.com.Agent.com, yeah.
SPEAKER_05: And you can get access to the code and you can see, you know, what everything is behind it.They're going to release their paper.They show it how they're scoring in the software engineering benchmarks.
SPEAKER_04: So do you see this as being like pair programming where you and an agent are working together eventually?What do you think this is going to look like when it's finished?When I say finished, it's like reasonably trustworthy enough for you to use every day.And somebody would want to take the time to use it.And it's not slowing them down.It's actually speeding them up.
SPEAKER_05: I think that's a really good analogy.I think it starts out as like a pair that's kind of giving someone superpowers.And so I read this really interesting analogy, and I can't find who came up with it.And I'll try to find it in Twitter, but the search is really difficult to use.
SPEAKER_04: The search has always been terrible.
SPEAKER_05: Yeah, it's always been difficult.You may know someone that can fix the checkout.
SPEAKER_04: I mean, you know, it's, you know, it's the thing about search.It's like such a rarely used feature compared to all other features that it's always with the exception of Google, which is certain.I know, but the problem is most people don't.And it really has always been, they did like a very advanced Boolean search where you do like from at Jason greater than this many likes or whatever.That's not what people want.People want to play language, interpret my search, you know, tell me about the Knicks or people talking about the Knicks playoffs. And that's where I think language models will help something like Twitter X leapfrog, right?Where the language model will be like, oh, I understand what your intent here.There's a website called Stat Muse.Do you know about Stat Muse?
Yeah, of course you do.Everybody in the NBA uses this thing constantly, and I guess it's got some machine learning.So if I said, you know, which point guards have the most 40-point games this season, it should have Jalen Brunson or whoever come up with, yeah, Jalen Brunson, number two, Luka Doncic, number one. And so that's just a natural language search, right?And it just nails that.I think that's what people want.And I don't know if this is a language model doing it or it's just machine learning. Like, look at this.When I type in, yeah, for stat muse, which point guards, you know, have four games a season.And then if I said with seven assists.
SPEAKER_05: Yeah, this is like, this is RAG, right?This is retrieval augmented generation.So this is using an LLM to drive a database. Because this is structured data exists in a database, but the SQL for this is not trivial.And the LLM is able to take this English and turn it into the SQL and go and get the results for it. And interestingly, Google just launched this for BigQuery and Snowflake just made this available for Snowflake.Powered by Gemini for BigQuery and powered by, I think, Anthropic for Snowflake.But let me just check the Snowflake.
SPEAKER_04: It's just amazing when you think about something simple like being an NBA analyst, right?And you see on my screen here, I said, which point guards have the most points per minute this season?That's a statistic that, I don't know, you would probably do the day before. You had Dallas Mavericks playing the New York Knicks, right?And then somebody would give that stat to the people who were doing it.Now you can just do it on the fly.Which point guards have the most points per minute this season?And as you can see here, again, Luka and Jalen and Steph are the three best point guards.And it's 0.9 points per minute for Luka.And then Jalen and Steph have 0.81.
And then it drops off from there.But how wonderful to be able to pull up this level of detail on the fly.This kind of knowledge was... You know, pretty amazing.
SPEAKER_05: Yeah, I mean, it's Mistral.Sorry, they did it with Text as Equal with Mistral on Snowflake.Yeah, this is going to be the world going forward.So let me share something with you.Just a quick aside, J. Cal.And this is not an original thought by me.I read it.I'd like to find the person to give credit to. which is when the world went from pre-industrial revolution to industrial revolution, let's use an example of making cars.We went from like making cars like in a bespoke kind of way, which is like you'd make one a day till you had the Henry Ford production line and factory and all of a sudden you make a thousand a day.
Similarly in farming, right?We went from people working in farms to having machinery and all of a sudden you could farm thousands of acres with machinery and feed lots of people. That was the industrial revolution impact on our lives.What someone said, what AI is doing, and we're looking at it in different ways, but that person's framework was, we are now having the industrial revolution for digital.Their example was, look, you used to make someone make a mock-up or a picture, and then they'd make one.A great illustrator, maybe take a couple of days.Now, using Midjourney, I can make 100 in a minute if I really want to.
SPEAKER_01: Yeah.
SPEAKER_05: Or I could make 1,000 songs in a minute.
SPEAKER_04: Just like a factory can make 1,000 buttons in a minute or 1,000 t-shirts in a minute or 1,000 books.But does it mean the book or the button or the t-shirt is going to be high quality or not?But it can spew a lot out.
SPEAKER_05: It can.And so now what we're seeing is for digital... And, you know, we've had these eras where we had digital stuff pre the Internet.That was like sort of when you were working in Florida.Right.And then you had the Internet show up, which, you know, kind of changed distribution.And we had mobile, which was just like, again, going to like a different form factor.We had cloud, which made it easier.But this Gen AI movement is the actual thing. industrialization movement of making more images, making more software, making more mock-ups, making it easier for us to access data quickly.
And so I think with that, when you apply that back to software engineer agent, Right.It's we can now, you know, the same way we went from making one card eight to 10, we can 100x ourselves.And so me as a software engineer, I can work on 10 different problems at the same time or 30 different problems with the help of these agents.What do you think about that?
SPEAKER_04: I feel like this existed already where I had writers, no, no, in a, in a minor way, I had writers who would come to me and they'd have like, I would say, Hey, come to me with like great story ideas.And then, you know, I'd have bloggers or writers come to me in the magazine day and somebody would have seven incredible ideas. And then other people would have one.And I'm like, how did this person come up with seven really great ideas?Because people came up with one.And that person had different resources they used to come up with story ideas.They had different techniques, right?
SPEAKER_03: Yeah.
SPEAKER_04: And so they, net net, were seven times more valuable to the magazine or to the blog, right?Yeah.And then you had people who had source.It turned out they had sources in some cases.They might have three or four sources where they would say, you know, to that person, hey, tell me what's going on in gadgets this week.And, you know, or they would just be shooting the ish or chewing the fat with somebody and they would get these ideas for stories.And it was like they had some better technique, which just then made them 10x better than the next writer or 100x.And that's just but one example in journalism, like your sources matter, right?So some source might
get you the content you need, some leak that makes you a Pulitzer winner.And then somebody else might be a better technical writer than you. And so I think what's going to happen is there are going to be some people who figure out how to use this pair programmer, how to use Sora, how to use, you know, whatever the tool is, they're going to find some like little edge and be able to use it in a way to perform at a much higher level.And yes, we will have the same way, some knowledge workers, some developers, some writers with sources, whatever it was, or 10 times better than their contemporaries.This could be like a hundred X. And so if you know how to use these tools, you're going to be just so valuable to different companies.And then companies are going to start to learn.If you can't use these tools, then think about how little value you provide to your company.It would be like trying to go work at a company in 2024, and you didn't know how to use Microsoft Office or email.You didn't know how to use an Office suite.
You didn't know how to use spreadsheets, word processors.You didn't know how to use Zoom.You didn't know how to use Gmail.But you're like, I have a phone. And I have a calendar that I write shit in and have like a written calendar.Maybe like, okay, short, you can give it a shot, but you would just be like this.You would be like this Neanderthal.Right.And I think that's, what's going to happen here is the people who know how to use these tools are going to be slowly become bionic and outpace their contemporaries.
SPEAKER_05: You know, like I said, I, um, And that framework was really interesting for me.You know, I sold my last company to Ford and I spent a lot of time understanding the history of the company and really just, you know, what they were able to do in terms of industrialization.And when I read that, it just clicked for me.And I said, all the stuff that we spend time thinking about is wrong. Because we are about to undergo this moment that those folks did like 100 years ago, when it was over 100 now, right?Early 1900s, where they were like, people were making like one car a day, and it couldn't change the world.And then all of a sudden, when you took it to 1000 a day, and made it affordable, everything changed, the world fundamentally changed. And I think software is still in that bespoke era.Even though we've had all these advancements in making software, better programming languages, we're not doing punch cards anymore, right?
We have great, amazing IDEs.But I think we're about to have the thousand X jump in software development.
SPEAKER_04: I mean, what does that even do?I mean, can we even comprehend?Like, we can comprehend what a 10X developer does.You ask them to build something.You thought it would take two weeks. It comes back tomorrow.Okay, yeah.Instead of 10 days, it took one day.Great.You love that developer.
You give them more stuff to do.A thousand X person, like, okay, we have a roadmap to build something over the next three years.And it came back tomorrow.It's like, wait a second.Okay, so should we just start another company or should we start another product line?Like, that's really wild when you think about
SPEAKER_05: yeah or like or think about like you know using like some kind of service and instead of like trying to integrate and you know maybe even pay someone to use it you know this thousand x world you just get something built bespoke for you and so it really has a huge impact on saz companies it has an impact on you know yeah i mean it would be like
SPEAKER_04: i'm trying to think of the analogy here but you know i i'm envisioning i want a car and i want my car to be a sports car but it also has a ski rack uh or like some way for me to put my skis on it and uh you know has snow tires and can you know be in four feet of snow but i also want it to be a sports car because i'm like all of a sudden this franken car is on your doorstep right you just yeah you're going to be able to make very custom things that didn't exist before, which in manufacturing, we actually have now, if you wanted to get a, the idea that you could get a t-shirt made or a pair of sneakers made to your specification with specific designs, we kind of take that for granted right now and nobody really does it.But if you want a sweatsuit, Like I see some people do this, like they'll have a birthday party or something for somebody.And then everybody shows up with like a sweatsuit.That's the person's picture.And you're like, okay, that's corny, but charming or whatever.
SPEAKER_05: We should do that for our friend.We should do it for our special friend.
SPEAKER_04: Oh my God.If we all wore a Phil Helmuth sweatsuit, which is pictures of Phil blowing up. Oh my God, that'd be hilarious.We should totally do that for the Phil Hellmuth birthday weekend.Oh my God, if we all showed up in a sweat, a tracksuit, a Phil Hellmuth tracksuit with pictures of Phil Hellmuth on it, that'd be hilarious.
SPEAKER_05: Or just the tracksuit with logos at least.
SPEAKER_02: Just the tracksuit with the Phil Hellmuth logos on it.Oh my God, so great. Any other debuts are we ready to wrap?This is a great Saturday show.
SPEAKER_05: This is a good one.
SPEAKER_04: We're ready to wrap.It's Saturday.All right, listen, everybody.Go to check out grokconsole.groq.com.Yeah, the cloud.Go there.
SPEAKER_05: We've got 85,000 developers, 25,000 apps.It is really, really awesome.We're going to cross 100K soon.
SPEAKER_04: I love it.I love it.Continued success.And if you have any great demos you want us to do here and give you a letter, great.And give you some candid feedback, just at Jason and at Sundeep, S-U-N-D-E-E-P on x.com slash formerly known as Twitter.And yeah, just reply to us and we'll do it.And we'll see you all next time.Bye-bye.
