This Week in Startups: Demoing Google Suite's AI plus legal docs, detection, and more! | E1835

Jason Calacanis Jason Calacanis 10/27/23 - Episode Page - 48m - PDF Transcript

and I did that in Google Docs,

we'll get to that in a second,

but I just pasted that in here.

Okay, this is a performance review for Greg.

Right, written by Google's AI,

I'm gonna show you that.

And so basically I pasted it in here,

and basically you can see it did a AI detection scores,

as 100% of it was written by AI.

Oh, wow.

Yeah. How does it know that?

This week in startups is brought to you by lemon.io.

Get access to Lemon Hire,

a platform with more than 80,000 pre-vetted engineers

that you can interview within 48 hours.

Get $2,000 off your first hire at lemon.io slash hire today.

Codecademy, build the future you wanna see with Codecademy.

Codecademy Pro helps you learn everything you'll need

to shape what comes next in the tech space.

Try it free for 14 days.

Visit codecademy.com slash quiz.

And NetSuite.

Once your business gets to a certain size,

the cracks start to emerge.

Things you used to do in a day, take a week.

You deserve a customized solution.

And that's NetSuite.

Learn more when you download NetSuite's popular KPI checklist,

absolutely free at netsuite.com slash quiz.

All right, everybody, welcome back to this week

in startups.

Sorry we didn't have a Monday module for you.

I am traveling all over the globe

and we had to tape our AI segment and our demos

for you later in the week, but Sunny Madra is back.

Welcome back to the show.

Sunny and Deep Madra.

Yes, awesome.

Sorry about the delay here.

How's the travel been, Jacob?

Give us a...

Very fantastic, you know, I'm learning a lot.

And it's been really a great eye-opening trip.

Had a lot of conversations, you know,

staying up till 1 a.m. talking about technology

and venture capital and politics and geopolitics

and the pod and everything, it's just crazy.

And what's been the mix between founders, investors

and fans, like how does that, what's the mix?

Pretty interesting.

I have to say, you know, I get,

when I come to an industry event here,

because, you know, Americans are not in Saudi

or UAE, Dubai, Abu Dhabi all the time,

I get a, you know, a hero's welcome and a lot of selfies

and then just tons of requests to meet.

So I've been meeting with a lot of interesting companies.

I met with my new favorite startup.

It's called Bread Fast.

Okay.

Bread Fast.

Get a like breakfast.

But it's Bread Fast.

It's Uber for bread.

It's the greatest brand name ever.

Once you hear Bread Fast,

you can never get this name out of your brain.

Because it's breakfast, but it's Bread Fast.

And the bread, Sonny, comes fast, like an Uber.

So basically they partner up like with local bakeries

and get you your favorite.

I think they are the bakery.

And they're the Uber layer.

So they get you your bread fast.

And so I met the founder.

They're doing like 100 million in bread in the region.

And yeah, and then he asked me,

oh, where you're staying?

And I told him where I'm staying.

And then at like eight in the morning,

I could knock on my hotel room door.

They gave me three giant shopping bags

full of every possible amount of bread.

So I have a little bread and I had a little chocolate croissant.

And then I gave my driver,

I got a driver here when I'm in the region.

And I gave him great bags of bread.

He was really happy.

Quite charming.

So shout out to all the founders I've been meeting with.

And yeah, there's tons of LPs out here, family offices.

The family offices are absurdly sophisticated.

These are like family offices that have been running

50 years since the inception of the region.

So they've been building real grinded out businesses,

like operationally hard businesses to like build a nation

like concrete and gas and oil and Pepsi and money transfer.

You know, it's just been really eye-opening.

And the amount of change going on here,

social change, economic change, political change,

everything is just wild to watch.

And there's graffiti behind you.

Yeah, well, I mean, a really cool co-working space

in Riyadh called The Garage that houses 150 startups.

I kid you not, I'm walking around here.

This is a, yeah, and this is not like a fake thing.

Like they're here working, you know,

and taking selfies with me.

You know, there's 35 million people in Saudi

and they're really well-heeled customers

with a really high ARPU, the average revenue per user.

And UAE, same thing, you know, it's smaller,

but a lot of revenue.

And then you're gonna place this like Egypt

with 100 million customers,

which is waiting there to engage with products.

And the American companies don't come here that quick.

So, you know, there'll be a local buy now pay later.

That's doing incredibly well.

There'll be firms that come here, you know,

I don't wanna say copycats,

but they'll build localized versions

of other ideas from around the world.

Kareem being the delivery service here.

But then they put a spin on it.

So like Kareem, the local service here is like a super app.

You can press a button

and have your apartment or house cleaned.

So like that whole Uber for cleaning your house,

like that exists here, right?

So they kind of add to the mix

and yeah, tremendous, the startups here

and the founders are kind of like,

I would say New York in the 90s, if I had to pick,

you know, like an emerging market like that,

where then, you know, the second decade,

you see like, you know, really big companies come out of it

and the third decade, giant companies come out of it.

It's kind of like Australia is now, you know,

or maybe before Atlassian, that kind of situation.

So it's gonna be a big story.

The startup's coming out of here,

the investment coming out of here.

It'll be the number two region for venture capital investment

and backing venture capital funds within 10 years,

I predict.

Do you say like North America, Middle East, Europe?

That's what it'll become, then Asia?

I think that'll be the order, yeah.

And I think Europe is like just not gonna,

with the exception of like Sweden, you know,

and I just think they're just not as hungry

and the growth in population is occurring

in Asia, Southeast Asia, Africa, North Africa.

So if that's where all the growth of customer is

and those, many of those markets are what they call

emerging frontier markets.

And then above that is emerging markets.

Those are the markets that really like, you know,

if you invest in and you put products

and services in those markets,

you're gonna grow with those markets

as they adopt technology and-

That's the JQL sweet spot too.

That's the JQL sweet spot.

There you go.

So it's, I'm not investing in any of these regions yet.

It's not part of our mandate yet.

But, you know, the amount of interest, you know,

this week in startups is incredibly popular over there.

People watch this episode.

A couple of people specifically mentioned the AI demos

and they love them.

So let's get to it.

You mean-

Let's do it.

Awesome.

All right.

Okay.

So we've got some great demos lined up today.

We're gonna start with something that's boring

but important and it's called Legal Now.

Especially, you know, we're talking startups,

we're talking building companies.

And one of the things that Silicon Valley law firms

have done for a while is definitely partner with startups

to kind of give you the package of goods.

And the thing I've seen as current iteration is that,

you still get that done, but, you know, firms are,

you know, still pricey and it still costs you

significant amount of money to get set up

and sometimes you use some template things.

And so this company's called, you know,

legalnow.xyz powered by AI.

And, you know, you buy yourself the credits

and they've got some pre-templated documents

that can be created for you.

Employment agreements for interns,

like an equity incentive plan,

terms of use, privacy policies, NDAs.

You've done all these things before, right?

Sales agreements, size agreements, yeah.

And so I did one earlier here.

I had this like internship agreement, right?

The summer is gonna roll around pretty quick.

We want to get interns.

And so basically it gives you the template here.

You fill in the name of the company, the effective dates.

And basically it spits out the document for you

with all the pieces that you need

and you're off to the races.

So there have been template sites like LegalZoom

as an example.

So how do you feel this is different in using AI to do this?

Yeah, I mean, fundamentally,

and like, you know, I thought the same thing

when I was looking at it, it's a cost thing.

And so, you know, LegalZoom is great.

Still way cheaper than using, you know,

like a Silicon Valley law firm.

But this brings it down

and this lets you iterate it a little bit more,

which I thought was, you know, pretty interesting.

And you can basically ask for more edits

through their AI assistant here as well.

And so you can kind of tune it a bit more like,

it's not a straight template.

That's what you'd sort of get out of LegalZoom.

It allows you to sort of modify it for your use cases,

which I think is helpful.

Yeah, awesome.

Yeah, I mean, a shout out to the LegalNow team.

This is going to get more and more interesting.

The thing I would love to see

is upload all of your legal documents to a service like this

and tell me, explain them to me.

Teach me what this means, you know?

I think you could do that with Claude, you know,

if you uploaded a legal document.

Hey, you're a lawyer, explain in bullet points

to your client what each of these segments mean.

And, you know, I think,

do you think you need a vertical AI to do this

or a general language model to do it well?

Because LegalStuff has big ramifications.

Yeah, I mean, you know, the interesting thing is

there was some data that came out of

one of the partners of OpenAI.

The OpenAI funded this company called Harvey

and one of their partners, like a large accounting

slash legal firm, they basically posted

the amount of savings they're getting by leveraging AI.

And so what that says to me is,

and that's built on OpenAI.

So I think what you really want to do

for these use cases right now is start with a, you know,

a large frontier model like OpenAI

and then give it some fine tuning information that you can

and then go from there.

You don't have to start kind of from scratch

and build something in this era.

If you've ever had to hire developers,

you know it is a giant headache.

Are they built from the startup grind?

How do you know?

If you hire the wrong developer, man,

maybe they're gonna drag your whole team down.

Maybe they take you down some rabbit hole

and you don't know what's going on.

And then not only are you back to square one,

you might be behind the eight ball, right?

But if you hire the right developer,

that could be a game changer.

Let me introduce you to LemonHire.

It's a platform that helps you find qualified,

senior developers who want to join your team.

LemonHire pre-selects vets

and matches tech candidates with high growth startups

so founders can focus on building their vision.

Here's why you can trust LemonHire.

They got 80,000 pre-vetting engineers

and you can interview any of them within 48 hours notice.

You pay nothing until you've started onboarding.

LemonHire is gonna replace any of your hires

for free in the first 30 days.

So just go to lemon.io slash hire, H-I-R-E,

to get a special offer just for quiz listeners,

$2,000 off your first hire.

That's right, two grand off at lemon.io slash hire,

L-E-M-O-N dot I-O slash hire,

hire with confidence at lemon.io slash hire.

I've got another demo queued up for next week

because I don't have it ready here,

which will do what you're saying,

but for your company documents.

So all your financing documents and everything else.

And so yeah, we're very excited to share that one.

We can look for errors,

look for things that contradict each other, right?

Explain things as well.

Like, tell me what share classes exist

and who owns how much and all that kind of stuff.

So it's very cool, we'll have that up for next week.

But very excited, so let's go to the next one here.

All right, so the next one is, I really liked a lot.

And it's interesting because we're gonna kind of flip back

and forth between startups

and some big company stuff today.

So this one is called rose.com.

Think of it as like an online spreadsheet

like Google Sheets is.

Yeah, it's like Google Sheets, yeah.

Yeah, exactly, but it has AI built into it.

So you have this like little tab here.

And so, you know, I just got like a CSV of some sales data.

Okay.

The standard one that you use for testing

and things like that.

I just inserted it into here.

And off the top, like you hit this like little button

right here at the AI analyst button.

Yeah, everybody uses this little like diamonds

or stars in the sky as the AI button now.

I don't know how that became the standard,

but it's like a twinkling star button.

Yeah, I don't know.

It's like kind of magic, right?

Is it maybe it looks like magic, right?

Yeah, sure.

And so what this does here on its own,

it does some quick insights, right?

So without me doing anything,

total number of units sold, average price,

maximum unit cost, minimum revenue.

And then it gives you some other deep dives you can do.

So you can do deep dives by region.

And this is all stuff is doing on its own.

And then it has like a text box here

where you can ask questions.

You can say, was the earliest sale, right?

And so it'll kind of analyze the table

and they'll come back with that results.

And so I got that one wrong,

but maybe the earliest order date, you know,

what is the earliest order date?

I love that it's anticipating with the quick insights,

other questions you might ask, the maximum unit price,

the minimum usage size, the average user price.

This is where, you know, look,

the idea is good, right?

And, you know, and then we can see.

We might have that data in there too, yeah.

Yeah, yeah, maybe there's, I don't think so,

because this is like a well-published thing.

So we'll have to ask these guys to comment on it, right?

Or okay, what is the maximum number of units sold?

And let's see if it kicks that one out for us.

But yeah, there we go.

So that one kind of came out.

The maximum number of units sold was 9,000.

And if you said, which month had the highest sales?

You know, people like to know that question all the time

from sales data.

Yeah, yeah, yeah.

Which month had the highest sales.

Yeah.

It would be interesting.

I think it's that one.

Show me the table showing the month of the highest sales.

So this kicked out down here, which is fascinating.

So we can say, yeah, insert table.

And so, oh, no, yeah, I got that wrong.

But the idea here is, you know, a couple of fold,

like one, you know, this is near and dear to us.

Like we do this for very, very large data sets,

so not spreadsheet kind of style.

But like...

You're talking about definitive intelligence.

You're talking about business.

Yeah, exactly.

Yeah, exactly.

So we don't do it like on a spreadsheet level.

We do it like on a data warehouse level.

But, you know, this is a really interesting problem.

It's becoming more and more relevant for folks that have data.

And I think if you're dealing with spreadsheets,

like something like this is the way to go for it.

And so...

All right, I give this a C.

It's not working too well yet.

I give legal now like a C plus.

It's interesting.

Maybe a B minus.

What did you give it?

You know, funny enough, I used it.

So because I used it, I'm going to give it a B plus.

Yeah, there was like a piece of documentation

I needed for definitive.

And I said, oh, I just kicked it out of there.

And it worked very related to the demo I showed you.

So for an intern, we're hiring.

I said, well, you know, why should we go and...

Yeah, and so I'm going to give it that...

Sounds like a low risk.

This is a very hard problem.

They're UI and everything is excellent.

But like, yeah, like you saw here,

there are some things that aren't working.

But...

R-O-W-S.com.

Not Rose as in R-O-S-E, the flower,

Rose as in columns and rows.

Got it.

Great domain name, by the way.

Shout out to the rose.com team for getting a four letter domain name

in the dictionary.

That's about a quarter million dollar domain name.

Yeah, maybe at least.

Yeah, wow, that's great.

I like the UI.

I like how they anticipated things.

So, but, you know, they didn't get things to work.

So I think with that, we have to give them a C.

But, you know, they can come back and they can improve it.

So...

Yeah, keep working, everybody.

Sharpen your problems, keep working.

I'm going to go to our next one,

which I find this one really important

because when I get things in my inbox now,

I'm starting to get suspicious if it was written by AI or not.

Yes, I have this problem right now.

A lot of my replies on X feel like they were written by an AI bot.

So this is called originality.ai.

In fact, one of the next demos, I created a performance review

and I did that in Google Docs.

We'll get to that in a second.

But I just pasted that in here.

Okay, so this is a performance review for Greg.

Right? Written by Google's AI.

I'm going to show you that.

Got it, Greg.

And so I basically, I pasted it in here

and basically you can see it did an AI detection score.

It says 100% of it was written by AI.

Oh, wow.

Yeah.

How does it know that?

So you put in something from ChatGPT4 and it figured it out.

Yeah, well, this was from Google.

It wasn't from ChatGPT, but yeah.

But it was from an AI writer and it put it in here and it detected it.

And it costs four credits.

And so there's been a lot of published research

because this has been a big problem since the release of ChatGPT.

And so a number of folks have written papers and methods

that can be used to determine if AI is being used.

And my guess is, because we've talked about this a few times,

the underlying data sources that came to provide

the core of the training material are open source.

If you use a model that has the majority of that,

I'm sure you can kind of figure out that this has been written by an AI

because you can kind of see the same patterns.

So my guess is they're using AI to detect that this was written by AI.

They're using AI to detect it.

So we don't actually know how it's doing it.

It said another way.

I mean, just for them, maybe we can have them come in and talk about it.

So I don't want to speak about their technology in too much detail.

They don't have a lot of details.

Anything else on this one?

No, it's just like an AI detector.

Yeah, I just, yeah.

I mean, I'm giving it a B plus, I think.

I mean, I would like to know how it works, but it looks like it works.

So yeah, it does what it says it does.

So what more do I have to give it?

It's like a single purpose.

I just want to give a shout out to our friends over at OpenAI.

Look, I asked it to give me a performance review for Cousin Greg session.

Did you read the session?

Did you watch the session?

No, I haven't watched it yet.

Oh, okay. Cousin Greg is like a bumbling idiot.

You know, trust on kid and adaptability fair demonstrates learning curve and

adapting to the corporate culture of Wastor Royco.

Communication needs improvement often struggles with clear and professional

permission prone to nervous rambling, which is true on the show.

Judgment exhibits growth in fair exhibits growth in judgment but still displays

naivety and certain strategic situations.

I mean, pretty professionalism lacks traditional corporate professional

however shows authenticity and interactions.

That's pretty good.

This is pretty great.

I don't know if producer Nick is here as a big fan of Cousin Greg.

What do you think, Nick?

I think a big theme of the show is how much better chat GPT is than every other product

that's out there right now.

At least that's what it sort of feels like to me.

Like it feels like chat GPT can do a lot of things better than verticalized

AI tools by themselves.

And I wonder if there's like a, I don't know.

I guess you could call it like a Kanye West problem going on where and this is

pretty crazy Kanye like everybody was always dubious about taking beats from Kanye, although

because he was the world's best producer, but you were always getting his like B material

when you were using his beat because he would always use his best beats for himself.

And Sonny, I'm curious, like, do you think that open AI is using their best version of GPT,

whatever that is for their own models and whatever they're letting people build on

is sort of a not nerfed but like slightly worse version of the model that they're using

themselves? Like will they always have an advantage?

Well, you know, it's interesting.

I was at a dinner last night, we were talking about this kind of from a different perspective.

When you're using models, right, there's two things that are happening

and I'll get around to your answer long-winded, but I think it's important for you to understand.

When you go to non GPT models that you're using directly like Lama or something like that,

you're basically interacting with the model itself.

When you use open AI, you're actually interacting with an API and that API could be doing it is

definitely doing a lot of things beyond what the model can do, right?

And you know, one of the things is that for GPT-4, I think it's well agreed that it's a mixture

of experts models. It's not a single model. There's a few models underneath the API and that

gets kind of the request gets forwarded to the appropriate one. And so that's one of the things

that you may be seeing just in the general use of GPT-4 versus other models that are out there.

Now, specifically to your question around like when you use the API versus you use the chat,

I do think that you're getting different versions because the APIs get locked to specific model

numbers and versions and you see that when you're developing with it. So I would guess that the

latest and greatest are always out in their consumer offerings because the terms of service

and the level of service are different than what they have to offer in the APIs.

So that's a long way of saying, yes, you're getting some different version of it. And who

knows? Maybe they could be saying bagging. Yeah. Yeah. Yeah. What did you give? Did you give

a grade there? Out of all, I've tried a few of these different ones. I'm going to give this one

like a B plus. I mean, I don't know what... It's cumbersome. You got a copy and paste, but I hope

the folks that are out there that need to deal with this can use it. I don't know what else it

could do to be better. Maybe it should just be like a plug-in or something. No. Well, yeah,

if it was a plug-in and when you were surfing the web, if it told you and showed you, but that would

require a lot of processing power, you'd be processing everything. Everything. Yeah. Maybe

you just pushed it like in your Gmail or something. If you were editor at the New York Times, it would

be kind of cool if when you were on your computer editing, it was always showing you in the CMS

if it was edited or making it better. Yeah. You know, kind of like Grammarly is always

looking at any editing I'm doing and working on it. So that's going to be kind of an interesting

thing that if everything you're looking at is being processed by AI in real time. Yeah.

But it could embed in a model locally that runs on your machine. Like, you don't have to go to

GPT-4, right? They could take a smaller model, embed it locally on your machine, and it's running

on your machine locally. So you don't have to worry about that. I never thought about that. What if AI

was running, you know, like the co-pilot when I was surfing the web, what could you do? Like,

what would you want to do? If it was telling you this was plagiarized, that'd be kind of

interesting. Or if it's telling you like another person who disagrees with this point or other

people talking about this topic or if it thinks there's a fact wrong. So that'd be really interesting.

Like, if I'm surfing a New York Times story and it's like, I think this is wrong,

or I think this is biased language, kind of cool, right?

All right, everybody, we all know technology is moving at a blistering pace. We talk about it

here every week, especially AI. And Codecademy's goal is to help you stay ahead. And so if you've

got some experience or you're brand new to coding, Codecademy is going to offer you everything you

need to build your tech career. You want to build that tech stack, you want to build your skill stack,

and you specifically want to gain skills in AI, full-stack engineering, data science,

cybersecurity. Twist listeners can try Codecademy pro-free for 14 days. We all know the world

large percentage of people, they don't think that college is the most important

factor when hiring people. They want to know about specific vertical skills. So listen,

when you got to lose two weeks free to check out everything Codecademy has to offer,

this includes 400 courses covering everything from in-demand languages like Python and JavaScript

to new courses in generative AI, large language models, and more. You're going to get step-to-step

guidance on your career path, learn how to do real-world projects to build your portfolio.

They're even going to help you with interview prep so that you nail that really important

interview. And listen, once you get the job, you can work towards your next promotion,

which is really what you should be doing. When you get that job, you should be looking at,

hey, who's my boss? How do I get their job? Who's their boss? How do I get their job? Well,

it's going to be experience and it's going to be skills. Codecademy is going to help you with

both of those. Start your free trial today. Visit codecademy.com slash twist to try Codecademy

Pro free for 14 days. That's codecademy.com slash twist.

There's a rumor that the next MacBook Pro is coming, I think we're like 32 or 64 GPU cores.

Did you see that, Nick? Probably tracking it more than I am, right? I saw it yesterday.

I saw that there's a big last-minute Apple announcement that's coming on October 30th,

and they're doing it at night and nobody knows what's going on.

Yeah, so the 5 p.m. Why are they doing it at 5 p.m.?

It's 5 p.m. Pacific too. Prime time Eastern.

8 p.m. I wonder if it's something to do with a live sports network or something.

You know how they've been talking about maybe doing some deals with sports networks,

so I wonder if it's something entertainment related. Is there anything going on that night in

sports that they would be a lead into? Because if there's, I don't know, if it was TNT NBA,

maybe they're making an announcement about the NBA.

That started yesterday, though. That started yesterday for us.

The first game of the season was yesterday. Yeah, the Knicks were playing

tonight. So tonight, yeah, can't wait to shout out to the Knicks.

Yeah, Apple already has an MLS deal and an MLB deal for baseball and American soccer.

But the NBA season is starting, so is this happening this week or next week?

This is next week. It's next Monday. It's this upcoming Monday.

Next Monday. Hmm, Monday Night Football?

Could be.

Could be Monday Night Football related. I wonder if they're going to announce some sort of NFL deal

or something. Well, I think their glasses are coming out in the new year, right? You're going

to be able to buy their glasses in the first quarter. Is that right?

The goggles. Yeah, the vision program. The vision goggles. Yeah.

That'll be first quarter for three grand. By the way, I got the iPhone 15.

Yeah, me too.

Yep, I'm kind of liking it. The USB-C.

The USB-C, I never, this is the level of like pathetic, but it really is a game changer

to carry one cable and not be like going around looking for two cables. That's where we're at

with technology today. But boy, does the new operating system and the new processor feel snappy.

I don't know if I, you know, or maybe my old phone didn't feel as snappy in the battery life

feels a little better. I like this new, there's a little thing up top where the camera is,

the front-facing camera. Did you notice it puts like logos in there? Like if you're watching CNBC,

it puts the logo up there. Oh, the Dynamic Island. Yeah, but what did you upgrade from?

So you've been missing because I had the 14th, so the Dynamic Island is very cool.

What do they call that thing? The Dynamic Island.

Well, you know what, try it with Uber. So the apps that do it the best are Uber, Uber Eats.

So shout out, you know, to your thing. Shout out Uber Eats, yeah. Shout out Uber and Uber.

Exactly, exactly. United, United Airlines. When you're on United, it will tell you the gate up

there and then a countdown to your... It is the most amazing thing of all time. It's my favorite

thing ever. It's unbelievable. It's an unbelievable thing. I literally thought that is...

They give you a countdown clock to when your flight starts boarding.

Yeah, and it shows your little gate. It'll be the gate F2 and then countdown to your boarding.

They're just literally like, what are you opening up and looking at all the time?

The gate anxiety, they just... They got you on the gate anxiety. We got you.

Then when you're flying, it has a reverse countdown. You can kind of see like how long

to your destination. So it's like counts down. I don't know if you saw this trending on Twitter,

but the woman who worked at Twitter, Jessica, I'm forgetting her last name right now,

she was on that flight. I was going into Seattle, Alaska Airlines, where a co-pilot in the jump seat.

Oh yeah, I tried to turn the engines on. Turn off big engines. Did you see this net?

Yeah, it was big news locally here. It was huge on the news.

Yeah, big news on the West Coast. This crazy guy, this pilot just decided he would... Yeah,

Jessica Verily. I know her sister because she does the all-in meetups in the Bay Area,

and she was on this flight with her two kids and an off-duty Alaska airline pilot was booked

in jail on 83 attempted murder charges after he allegedly tried to shut off the engines

of the San Francisco bound plane. Oh my God. The pilots apparently like tackled them and then they

put those, you know, wrists, restraints, zip-tied them. Yeah, exactly. Yeah.

Yeah, I'm trying to let anybody know, man, I'm a child of 9-11. Anybody tries anything on my plane?

You're done. Yeah. I'm taking you down. I'm literally... I'm the first guy getting up.

Can you give the Taekwondo a picture of J. Cal? No, I'm just... I'm team Let's Roll.

Anybody tries any shenanigans on my flight with my kids?

Let's roll. And I'm counting, by the way, if you get saucy with the flight attendant and you decide

that you're going to get into it with the flight attendant, I'm the first dad getting up.

You know the dad who gets up on the video, on the TikTok video, who tells the other person to

sit down and shut up? That's me. I'm the first dad. Yeah, let's go. Gotta be that way. No nonsense

on the flight. No, not on flight, not anymore. I don't understand what's going on. I understand

during COVID people lost their minds, but listen, hey, everybody, it's three years after COVID.

No moss. You get on a flight, you behave yourself. That's it. I don't want to hear anybody saucy with

the flight attendant. I don't want to hear anybody. I was at an Emirates out here next to me

in first class. He's pounded scotch like it's iced tea. He's on his third double. I never saw this

in my life. The flight attendant said, listen, and I certainly never saw this in business class.

But I said, listen, I don't want to give you another... Yeah, you're like shaking his glass

kind of situation. You know, the guy shaking the glass. Yeah, terrible. Yeah. And he's like,

you need another one. The guy's like, listen, I don't want to give you another one right now.

You've had three. And I hear the number three. I'm like, am I going to really have to tell

this guy? Am I going to have to go into dad mode and tell the guy, listen... Yeah.

On a 17 hour flight, too. That was exactly what I was thinking. We were in hour two

of a 17 hour flight. Look, there it goes again. I held up my peace sign and I got a bunch of

balloons in the AI. Maybe that's the sign for balloons. Try it. Now it's not. I've seen it come

up a couple of times. Yeah, I know. All right. Let's keep going. Just one thing on the Apple

thing. This was the announcement I was talking about. So looks like M3, Max and M3 Pro are

going to have 16 and 40 GPU cores. What can you do with that? Should I put tires on my MacBook Pro

and have it drive itself around town? I mean... I think you could start running the AI locally.

That's an absurd amount of processing power. Maybe if we want to do a little speculation.

A little speculation could be a little local AI. Apple's so far behind that maybe they think,

you know, the problem with Apple, and you see this with goggles, is in the race of like,

don't worry, be crappy and get the product out there. Apple doesn't play, don't worry, be crappy.

That's not their game. They want to do, take your time, tight is right. Not, don't worry,

be crappy. They can't play in this AI game and Siri sucks. Yeah. But I will say my ChatGPT

virtual assistant. I've been playing with that where I talked to it. Yeah. Quite compelling.

But the demo that we did when you put it in audio mode. Yeah. It's kind of working now and they

have a bunch of different voices and it's a little bit faster. Yeah. I'm kind of liking it. I hate

to say it. But I was just in the car and I was, this is the, after we did our thing and I was,

I was like, tell me the history of Saudi Arabia. Tell me the modern history of this. Tell me that.

And I was asking you a follow up question. Yeah, your AirPods in, or you don't use AirPods,

you use the other ones. I was driving in the car. Yeah. And so I just had it up there,

like it'll be her. And I'm just, you know, I was going to pick up my daughter and I had time and

I was like, I can listen to a podcast or I could talk to a chapter first. I started wrapping the

chapter and getting a little history lesson going. Oh, right. Okay. Pretty cool. Yeah. Awesome. So

for history, like for your drunk history uncle, shout out, David Sacks, history uncle, history dad

can get a lot out of this. All right. Let's keep rolling here. So, Jacob, you sent this next one

over. Oh, yes. Yeah. And so this is called some cool because summarized.tech, very simple and

clean UI. So I think they're just going for functionality right now, which is great. And to

be honest, everyone's been talking about this Jared Kushner, Lex Friedman podcast, but it's

like three hours long. And so I popped that in here. And this does a couple of interesting things,

which I want to give them some credit on. If someone else has already done it, you can kind

of go here and see recently summarized videos. And so, which is really good because you don't

have to wait for it to do the work, which is smart, right? Because, you know, it makes no sense to

kick kind of keep doing that as well. And so I thought that was really well done. And then from

the summaries, in this case, you know, this was a three hour 45 minute podcast. And each hour,

but then you can do like a little C more. And then if you want to kind of get into each section,

you can kind of scan through it as well. So I thought that was really well done.

You know, I thought it was, you know, simple like simple UX. Again, like this should probably

be a part of YouTube. That's the problem. Like it doesn't shouldn't be like a, you know, like

an external feature. It's a feature of YouTube rather than an external service. But I really

like the simplicity of it. I like how they've taken advantage of caching and, you know, taking

advantage of sort of other people that are doing it. You get this one basically happened

instantaneously for me. So someone else must have done it. Pretty neat. I like it. So I give

this one like a B. Yeah, I think I'm the same as you on this one. Like again, the only risk for

them is this this naturally for the folks at Google should be a YouTube feature. I mean,

it could be doing other interesting things here. Like if it summarized this video,

you know, it could like this or some like some recommendations could be like here's a summary

of this section. But then here are the next questions to ask. Tell me about the history of

Hamas. Tell me about the history of the Israeli Palestinian conflict. So remember, we were in

that other Google sheet and it was giving you suggestions in the wizard, anticipating your

next question. Each of these could have some questions, you know, you could prompt it. I'm a

podcast listener looking to understand these conversations deeper. Tell me what question

to ask next for each hour of the pod. And it just comes up with 10 more questions to ask.

So that could be kind of interesting. Scaling a business is hard. It's really hard. Let's be

honest. And when you're growing things that used to take a few hours, well, now all of a sudden

they take a week, they take two weeks, maybe they never get done. So you wind up getting bogged down.

That's not what you need. You need to be efficient. You need to be light. You need to stick and move.

And one way to stay efficient is to understand your KPIs. What does KPIs stand for? No need

to be embarrassed. Key performance indicator. So basically, these are the important metrics to

track for performance against your objectives. Right now, you can download NetSuite's popular

KPI checklists for free at netsuite.com slash twist so you can catch up on a KPI checklist.

If you're in SaaS, you've probably heard some of these numbers and some of these acronyms,

Mao, MRR, ARPU, average revenue per user, and NetSuite's KPI checklist will teach you to identify

and understand your strategic objectives and then how to collect and analyze the data and so much

more. Identifying KPIs are going to make you more efficient. Your team is going to be happier.

You're going to be happier as the boss. And right now, increasing margins and getting profitable

is the name of the game. KPIs is the road to that. So here's your call to action. NetSuite is

everything you need all in one place. Download NetSuite's KPI checklist for free at netsuite.com

slash twist. That's netsuite.com slash twist to get your KPI checklist right now.

Okay. Next one. So we're going to pop back. We're going to do a couple with our friends at Google

again. They keep innovating. They keep innovating. And so it's good to see it like it's an era of

innovation. Okay, Sundar. All right. So it stocks down a bit. But they got to keep pushing. Yes.

Okay. So here, what you see now is in your Google Workspaces, if you have AI for Workspaces enabled.

When you start a new Google doc, in this case, like a blank one here.

Untitled doc. Classic. Untitled doc.

Classic. Untitled. You can help me write. And so...

Oh, look at that. It took prompt right in the white space there.

Right in the doc and say, okay. Yeah.

Okay. Help me write a performance review for Nick. Okay. I want to highlight his strengths

Okay. Let's just say as being on time, doing good research. Give me one more, J.K.o.

And his jovial disposition. Okay. Okay. Okay.

Areas he needs to work on are his hairstyle. That is doing his personal hygiene.

Personal. And inappropriate banter.

Sand. What's going here?

You're getting a little inappropriate. Here we go. I sound like Chris Farley.

Basically. Yeah. Absolutely. All right. So I'm going to do this. And so...

And I showed the Greg one before while it generates that. So I have this one here that I

did for Greg. Sure. Fictitious Greg. Oh, wait. We're still learning. I can't help with that.

Try another request. Wow. Really? What? It's crazy. It just did it. This is always weird.

I know. I hate when it's not to do it. And then you tell it, it's okay. You can do it.

And it does it. Yeah. Just do it. Right. Okay. Here we go. And I did it. Second tries.

Yeah. Yeah. Yeah. And so you can kind of go here. And I really like this. You can say

elaborate tones. You can change the tones. You can say, make the tone more formal.

And then you can have it elaborate. And you know, J.K.o. I'm bringing this up because like, look,

like last week, we did the images inside. So it says here, Nick is a valuable member of the team.

He is punctual and completes his assignments on time. He's also a good researcher.

But then, however, Nick needs to prove on his personal hygiene. He has a habit of coming to

work with unkept hair and body odor. Come on, Nick. He also needs to be careful about his jokes.

He sometimes makes inappropriate jokes that makes his colleagues uncomfortable. Nick, please.

Yeah. And I love that you elaborate. We should elaborate on the body odor.

Yeah. I elaborate on the body odor.

And like, but you know, J.K.o. like, look, you know, you've been building companies and managing

people for a long time and performance review is important. And sometimes it's hard like to,

you know, build that old. Get started. Yeah, get started. And so you do this.

Cold start problem is real. And like, there you are. You're done.

I mean, the crazy thing about this is if you actually know your points to put them in full

sentences, this does a wonderful job. So for people who have writer's block,

this is going to do it. I was writing a deal memo, and we have a section in our deal memos

for the syndicate where it says the bet. And okay, you know, people have a hard time writing the

bet. So I've been having my associates managing directors, you know, et cetera. I give a shot

at writing the bet. And then I basically approve it or write it. And every time it's not as good

as I want it to be. Yeah. And the last time I said, you know, I just highlighted what you wrote

in my notion AI. And I said, you know, make it better. They have a button that says make it

better. Make it better. Yeah. I was like, it's so much better, the notion one. Yeah. So like,

your original ideas are just presented much better. Yeah. Can you do me the honor?

Yeah. Save us all. But now like, you know, you're using notion, but like, you know, if Google

Docs is your platform, like think about how, you know, what this is going to do. I think it's

really all executed. That's four hours of work that, you know, somebody puts on their to-do list.

That's going to take them two, four hours. And if they got to do 10 of those, and it's

20 hours of work, it's three days of work, like, they'd be like, you know, the boss would be like,

Hey, what are you doing this week? This is performance review week. Week. Yeah. Like,

week. No, seriously. And now it'll be like performance review day. Yeah. And you know what,

now highlight that review. Here's what I want you to do. I want you to take that review. I want

you to say, your Nick, write a rebuttal to this review that asks for a 20% raise. Okay. But that,

they don't, they don't allow that to happen. Oh, yeah, you do that. But you could copy this into.

Oh, yeah, I could copy it. You could copy it, put it into chat, and say, this is my performance

review. Write me a rebuttal that doesn't make me look defensive and ask for a 20% raise.

That's a good one. And so what should happen in the future? Your AI should know automatically to

do this and try to get you to do it. Right. It should know when you get a performance review,

you should write a not a, you know, a review that's not defensive. Let's see what happens here.

Dear manager's name, J. Cal, I want to express my gratitude for the

emergency performance review. I genuinely value the opportunity. I understand. Oh my god,

this is such a whole lot of horseshit. Firstly, I'm proud to acknowledge the positive feedback,

particularly my research goals. Regarding feedback about my personal hygiene, I sincerely

apologize for the discomfort I may have caused to my colleagues. Nicely done, Nick. I recognize

the importance of maintaining a professional appearance and the impact it has on the work

environment. Please understand that personal circumstances have temporarily affected my

routine, but I'm committed to addressing this immediately and sure it doesn't continue. Wow,

what a great excuse. Well done, Nick. It feels like character assassination.

It's just a joke. Nick does not. Nick's hygiene is, seconds are none. I mean, it's incredible.

The guy smells like roses. As for the jokes, I understand that humor can be subjective and

sometimes, despite my best intentions, a joke might not land as intended. This sounds like me,

actually. I appreciate the insight. I was going to say, I'm usually the get back guy for Jason.

Exactly. Yeah, this is me talking to my wife. I appreciate the insights of this and will be more

vigilant about ensuring my humor aligns with workplace appropriateness. Not bad, Nick. You

did a good job. Yeah, and then it comes to the race. Considering my commitment to the company,

the constant positive feedback in several areas and the value I bring to the team,

I kindly request a salary adjustment of 20%. Well, that's a big number.

I believe this reflects both the current market value of my role. I mean, my lord, this would

have taken somebody like two weeks of like hand wringing and edits and stress and, you know,

like asking a friend, like you go to your friends, be like, oh, I got a reply to my manager,

what should I say? So now I want you to copy that. And then I want you to write

I want you to put that into Claude AI and then tell Claude AI that market conditions don't

allow us to give a 20% raise, but we're happy to do a 4% adjustment for the cost of living.

Don't do it. You can see where this is going, everybody. So I'm going to just say that Google

AI writes for that AI writes for Google as presented. That's an A for me.

Yeah. Okay. Another A. Okay. Yeah, I think it's an A. Like, you know, in fact, I found it because

I was trying to work on something and then basically I've been using it like every single day now.

I just want to say as a writer. Yeah.

I would say this is a bummer for me because like I used to be able to do all this stuff,

but I'll be totally honest. I hate doing this stuff. And if this is like people who work for me

can just do this because what happens is ultimately myself or producer Jackie who used to work on the

show and is now a managing director, all this writing stuff winds up coming to our desk. Someone's

like, I need help writing this. And it's like, oh, producer Jack, he's the writer at the company,

Jake house, the writer at the company, go to them, they'll write the copy.

And I just rather everybody be able to write good copy from the get. And then I can just do my

pixie dust, you know, a little bit of flair on it. Yeah. You know, and then eventually it'll

probably understand how I do flair. And it's just like, you know, we did the originality,

maybe instead of being 100% AI, you add it and it becomes like 90% AI, but the 10% flair is the

human and that's okay. Perfect. I like it. Great job. I'm giving it an A. So send this to Sundar,

you got an A. You finally got your A. Not a plus, but a. Okay, last one. Okay, I'll say I have to

Google Sheets one, but I'll save it because I have two more to go. Actually, let's do Google

Sheets and we'll save the bard for next week. So very similar theme here inside of Google Sheets.

You know, when you start a new sheet, you'll get this little, you know, tab that comes in from

the right hand side. This helped me organize and so it gives you ideas like, you know, wedding

event planner, wedding event plan in a zoo with different animal kitchen remodel task list, you

know, and so you guys run conferences, Jake house. So what I was going to say is let's,

let's pop in one, like if you were wanted to create some kind of checklist for a conference,

what would you, let's make a, let's make a run of show for a corporate conference that's three days

long with 16 speakers a day. We'll see if it understands run of show. Okay, that's it. That's

it. Let's see what it does. I don't know if it's like the same kind of issue or let me grab this.

I don't know if it's like a refresh issue because it's like it's a weird error that it gives there.

Yeah, it's a weird error. Okay, so let's give it a shot here. Let's do it again. And

that's terrible. Oh, that's not good. Okay, so, okay, Sundar, this is, this is the don't worry,

be crappy mode that we're in with AI. You got to build stuff and you got to throw it out in the

world and it's saying we're unable to generate any results based on your prompt. Please try again.

Let's do another prompt. Please create a budget for a three day corporate conference for 150

executives, include food, lodging and entertainment. You know, this is something somebody might do,

right? Somebody on my team might do something like this. Somebody on your team might do an

offsite meeting like this. Let's see if it makes an actual budget for a three day corporate

conference for 150 executives, including food, lodging and entertainment. All right, lodging,

okay. Okay, let's hit insert and see what happens. What happens next?

Yeah, just kind of dropped it in here. It's kind of weak sauce. Let's try again. It didn't do a

good job on that. It just put like $200 a person for dinner and it didn't even put themselves. So

what will we do here that could be very interesting? Please create a PNL for a startup in the software

industry. That's a pretty good one, right? That's like a template they must have somewhere. I must

understand. Yeah, you would think so. Yeah. See if it does it here. This could be a chance to, okay.

No, we did a quarterly report. Yeah. Okay, so sorry, Sundar, you got an A for dots. The Google

Sheets team has failed horribly and they're getting an F. There's an F. You failed.

Well, I will say I picked one of their examples, which was like a new product launch.

Sure. Yeah, and this one, sorry. Yeah, I'm sorry. Nailed the example. Yeah, okay, great.

I give it an F. What do you give this pathetic performance? I'm in that range, you know, like

these things I have to do. Yeah. Yeah, plus from everybody. This has been another episode

of This Week in Stardust for Sundar Madra. I am Jay Kao, the world's greatest moderator. We'll see

you next time. Bye-bye.

Machine-generated transcript that may contain inaccuracies.

This Week in Startups is brought to you by…
Lemon.io. Get access to Lemon Hire, a platform with more than 80,000 pre-vetted engineers that you can interview within 48 hours. Get $2000 off your first hire at http://lemon.io/hire today!
Codecademy. Build the future you want to see with Codecademy. Codecademy Pro helps you learn everything you’ll need to shape what comes next in the tech space. Try it free for 14 days. Visit Codecademy.com/TWiST
NetSuite. Once your business gets to a certain size the cracks start to emerge.  Things you used to do in a day take a week. You deserve a customized solution - and that's NetSuite. Learn more when you download NetSuite’s popular KPI Checklist - absolutely free, at NetSuite.com/twist

*

Today’s show:
Sunny Madra joins Jason to demo an AI-powered legal docs tool (6:46), detection tool (15:38), Google’s workspace AI (35:15), and much more!

*

Time stamps:

(0:00) Sunny Madra joins Jason
(6:46) Sunny demos Legal Now - AI-powered legal docs tool
(10:01) Lemon.io - Get $2000 off your first hire at http://lemon.io/hire
(11:10) Sunny demos Rows - Web-based Excel
(15:38) Sunny demos Originality AI - AI detection tool
(22:36) Codecademy - Try Codecademy Pro FREE for 14 days at http://codecademy.com/TWiST
(24:02) Apple testing M3 MacBook Pros with 16 CPU cores and 40 GPU cores
(31:32) Sunny demos Summarize.tech - Summarizes YouTube videos
(33:57) NetSuite - Download your free KPI Checklist at http://netsuite.com/twist
(35:15) Sunny demos Google Docs AI write
(43:49) Sunny demos Google Sheets AI
*

Check out Legal Now: https://ai.legalnow.xyz/
Check out Rows: Rows.com
Check out Originality AI: https://app.originality.ai/content-scan/6910415
Check out Summarize.tech: https://www.summarize.tech/www.youtube.com/watch?v=co_MeKSnyAo

*

Follow Sunny: https://twitter.com/sundeep
Check out Definitive Intelligence: https://www.definitive.io/

Read LAUNCH Fund 4 Deal Memo: https://www.launch.co/fourApply for Funding: https://www.launch.co/applyBuy ANGEL: https://www.angelthebook.com

Great recent interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland, PrayingForExits, Jenny Lefcourt

Check out Jason’s suite of newsletters: https://substack.com/@calacanis

*

Follow Jason:

Twitter: https://twitter.com/jason

Instagram: https://www.instagram.com/jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

*

Follow TWiST:

Substack: https://twistartups.substack.com

Twitter: https://twitter.com/TWiStartups

YouTube: https://www.youtube.com/thisweekin

*

Subscribe to the Founder University Podcast: https://www.founder.university/podcast