Nov. 13, 2023

My GPTs, and avoiding the OpenAI tidal wave as a founder

My GPTs, and avoiding the OpenAI tidal wave as a founder
The player is loading ...
Hallway Chat

In this conversation, discussions revolve around the recent OpenAI Dev Day, the Reflect note taking app, and how founders should internalize the "threat" and partnership opportunities with OpenAI. Fraser & Nabeel discuss that it's still "fart app and flashlight" early days for developers to truly utilize the capabilities of technologies like ChatGPT. The conversation then shifts to Reflect, a note-taking app that utilises AI to optimise features like speech-to-text and backlinking. The speakers debate whether one company can excel in both consumer-facing products and developer platforms, highlighting how OpenAI is in a unique position to potentially 'run the table' across various booming markets.

  • (00:00) - Are My GPTs an app store or chrome store?
  • (01:00) - Open AI Dev Day
  • (01:33) - Exploring MyGPTs
  • (08:44) - Difficulty and Promise of App Stores
  • (11:28) - Surprises of the first year since Chat GPT
  • (14:14) - Investor influence on founders
  • (19:36) - Coming to terms with the AI fog of war as a founder
  • (21:54) - Exploring Reflect: A New Note-Taking App
  • (26:45) - Will OpenAI be able to be consumer & enterprise?
Chapters

00:00 - Are My GPTs an app store or chrome store?

01:00 - Open AI Dev Day

01:33 - Exploring MyGPTs

08:44 - Difficulty and Promise of App Stores

11:28 - Surprises of the first year since Chat GPT

14:14 - Investor influence on founders

19:36 - Coming to terms with the AI fog of war as a founder

21:54 - Exploring Reflect: A New Note-Taking App

26:45 - Will OpenAI be able to be consumer & enterprise?

Transcript

Are My GPTs an app store or chrome store?

Nabeel Hyatt: I'm stitching together an AI future.

Fraser Kelton: Nabeel, you're stitching together an AI future, is that mid

twit? What you just walked through, and it's

Just listen to the podcast, is on either end of the curve?

And it's I've stitched together the future of AI to

get these.

Nabeel Hyatt: Hello, everybody. First big news is we have a podcast name.

Fraser Kelton: What was it? Next token prediction.

Nabeel Hyatt: NextTokenPrediction. No, it's HallwayChat!

Fraser Kelton: Hallway chat.

Nabeel Hyatt: have spoken to our illustrious friend and ambassador. I sent, I sent a text message to an ambassador to get a password to a podcast

so that we could take over the name HallwayChat.

He put the password as past the torch, which was very teary eyed.

But happy to be here with you, Fraser.

Fraser Kelton: Oh, happy to go see you, as the ambassador of Sweden in a decade.

Nabeel Hyatt: I don't think so, man. I'm just not capable of being that political or well behaved.

It's pretty obvious what we're going to talk about this week.

Open AI Dev Day

Nabeel Hyatt: Thanks, everybody. If we're talking about anything, which is we're going to talk about OpenAI Dev day.

Fraser Kelton: That was the thing that happened this week. Yeah, I heard about that.

Nabeel Hyatt: No one cares about Apple keynotes, by the way, anymore. But it's good to know, that there's still a keynote that gets everybody a little riled up.

Fraser Kelton: Oh I, I will tell you there was a buzz the weekend before, wasn't there? And then I was so you know, that's a good call out. I was giddy like I used to be for Apple Keynotes for this. And then I was just so proud of my buds over there for shipping all this stuff.

Exploring MyGPTs

Nabeel Hyatt: So my first question for you... Given your background and where you come from, how much of what, without saying anything confidential, how much of what you saw

there was stuff that you knew about a long time ago and at OpenAI has been worked on for a year. and how much of that stuff was, hmm.

Fraser Kelton: Sorry, I can't help it. The idea that they work on stuff for a year

is amazing. If you, if you

squint the vision for

my GPTs. was there as we were building early versions of, of what became ChatGPT, right? I think it's not, an elastic imagination to say, Hey, this LLM is really great at certain tasks.

That have some utility, but then low utility tasks individually aren't that interesting. If you can find a way to make it easy to create those, discover those, share those you can probably get a bucket of small utility in aggregate that's quite useful, right?

And I might have, I might have five MyGPTs that are useful for me. You might have... Two of the same, but you might have then six other ones.

Nabeel Hyatt: You're saying it in that way you mean predating plugins?

Fraser Kelton: Yes.

I have heard that there was a blog post that went out about a confidential talk that Sam gave in London where he said that plugins didn't quite have product market fit. And I may share that point of view , it was nice to see them get back to, to a place that I think is going to have tremendous value for end users.

And then by extension, real value for them, everything,

Nabeel Hyatt: So as the outsider, I'm going to take a read of that is, my colleague fraser didn't think plugins should ever be done. He liked my GPTs when he was there. They got sidetracked. They're back on homestretch.

You don't, you don't have to respond to that fraser.

Fraser Kelton: think the world was, had a lot of euphoria at the start of this year because it's the, the world's response. And I wonder how much of that euphoria. Led to a plugins decision, but MyGPT is, I think is, is elegant. It just makes sense. We'll see how people adopt it, of course. Everything

Nabeel Hyatt: think.

Fraser Kelton: Everything else was new.

Nabeel Hyatt: Everything else being new is interesting, since some of those things feel like they are complicated undertakings or long engineering efforts, and I think it does tell you, given that you weren't at OpenAI that long ago, does tell you a little bit about it.

How quickly this industry is moving, what would you say if a founder was talking to us today and saying, what to take away from OpenAI Dev Day? Did we learn anything new

Fraser Kelton: I don't think much of what we would have said in the past would have changed because of this. You need to have a differentiated view in the world and have opinionated product that you're delivering. And it's absolutely okay to be building with hosted models. That stuff's all the same. My observation, if I look back over all of our demos and releases and this one's probably going to be no different is the demo is going to overstate in the short term, what we all think these technologies can do.

And then we're going to wake up in 6 months or 12 months, and a lot of that stuff is actually going to happen, right? The Codex demo from Greg, I don't know if you watched it, I remember being on the Twitch stream when it played out, and it was so lovely to see because people were blowing their minds on the Twitch comments, much like we had been internally when we saw it.

And then, and then there wasn't much, right? There wasn't much with Codex. But we're starting to see the really creative use of these types of technologies in decoding applications that will rewire the

Nabeel Hyatt: It feels like even with Copilot around, there's been an explosion of, coding startups.

Fraser Kelton: Yeah, for sure, right? Because the models are excellent in that area and it just happens to be a problem that these developers have for themselves, so they have good intuition on what can be done.

But like the Codex demo was some number of years ago. It has taken a while for us to get there. And my sense is that this is going to be very similar.

Nabeel Hyatt: It's funny that we first talked about GPTs, MyGPTs, which is of course not a developer focused thing, really. That's actually a consumer focused thing. On the developer side, we heard things like they updated the knowledge base cutoff date. There's some more input and output controls.

The API now added Vision and DALL-E and Text to Speech. What else was there? Fine tuning controls were also added. The token length was also extended. A whole host of things that would seem to be helpful. And, and what devs want. But the prevailing thing that it feels like I walked away from even though it's a dev day.

Is just, given the weight and power of chatGPT on this company, we didn't hear any announcements about SLAs. And in fact, the day after, we had some serious SLA problems, I use an app called Reflect. It wouldn't work well because OpenAI was down, and it was using Descript, and I was trying to edit podcast notes from a previous podcast, which should go up tonight, and similarly it was down, and it reminded

me that, I don't know, that as much as this was a dev day, I'm just not sure this is going to end up being a developer focused company. I don't think it can.

Fraser Kelton: I think they're, rather ambitious. And I think that they're they're playing a game across a number of different fronts. The, the sheer size of the success of ChatGPT is, is outrageous. But they're building a good business.

They're building a good business with the API. So there's like a lot of complexity. And I think the world hasn't necessarily seen a situation like this to this degree where you're, you, it's not even that you have a platform where you're then chipping off on features. It's that you're. Your consumer product consistently is getting better on an arc where it's just absorbing the functionality of what's getting built by many of the people on the platform.

And that, that is going to be interesting. I, I so listen, they may be a, a ChatGPT first party. But you have to have a a big imagination to think that UIs are going to fade away and that you're going to have no third party products for bespoke, purposes.

Nabeel Hyatt: I think that's true. Look, I checked in with a handful of founders yesterday that operate in the space just to get their hot takes on what they think about this. And what I heard a lot was of folks that are horizontally oriented. DevTools

oriented and these kinds of things they felt pretty threatened.

And folks

that felt vertically oriented were going

after a specific use case.

We're building a certain kind of behavior. We have a differentiated user interface. We're building, and then we're building on top

of this model in that

direction.

It was, they all seemed

fine, you

know?

Fraser Kelton: Yeah. Listen, they launch a text to speech API and I think it's like 10x cheaper than, than the market. I might be

wrong on that, but it's

like dramatically less expensive. And, it's a happy by product of investments into other areas that they can just bring to market for, for developers to, to utilize.

It's amazing.

Difficulty and Promise of App Stores

Nabeel Hyatt: Yeah, I think my last takeaway from those check ins was like, nobody seems particularly enthusiastic about... The GPT store. As much as I think I will use GPTs, like I actually already have a handful of prompts. I am one of those humans with a Google Sheet filled with prompts. And I actually built a little silly chatbot, um, side project last summer that would go fetch the prompt for the job and stuff like that.

So I will certainly use GPTs. But what I didn't get a sense of I assume OpenAI wants this to feel like the App Store, right? That every entrepreneur... Should go now

30 percent of the YC class should be folks trying to make GPTs, which is why they launched with monetization coming soon, monetization,

Fraser Kelton: I

Nabeel Hyatt: they want it to feel like the

mobile revolution

did, and I, I don't suspect that that,

I don't know, what do you, do

you think that's going to be a thing?

Fraser Kelton: have high. Confidence that what you said you're going to use it for is what a lot of us

are going to use it for. We have these small tasks that, that we can stand off these rough

edges in our day to day lives, specifically at work.

Nabeel Hyatt: Yeah,

Fraser Kelton: am I going to use it as a consumer ?

I think that's to be determined. I think that's to be determined.

Nabeel Hyatt: think we forget how many app stores post the app store the iPhone app store just haven't worked, There's a ChRoam app store. I don't know if you use that that much, like you can, and it never worked. It's still basically extensions,

which is what it's always been. It's never really turned into a gold

Fraser Kelton: Even the group who has shipped the app store has a watch app store a iMessage app store, a macOS

Nabeel Hyatt: Yes,

Fraser Kelton: and they're all like just,

Nabeel Hyatt: Apple tV app store. They actually spent a bunch of money to get people to move games onto the tV app store at Apple.

It's just never been a thing.

Fraser Kelton: yep, yep, yep, yep. So these things are, are hard

because I, I think companies

want them.

It's not necessarily clear

that. Consumers and developers want them.

I do have, I have high conviction that I have

10 tasks that my

GPTs can help add a little bit of structure so that I

can more easily get value out of this model.

And probably even more importantly, I can then share these three tasks with my colleague who doesn't quite know how to do prompt engineering or whatever within ChatGPT to get the value out of the model. And so that, that, that will happen. Am I going to use it for consumer y like things that people will build big businesses off of?

I don't know. To be determined, I'd love to see that. I'd love to see it, right? I would love to see it. It would be so awesome if somebody creative walked in.

Surprises of the first year since Chat GPT

Nabeel Hyatt: Fraser, last night we had a AI dinner with a handful of founders, including Andrew Mason over at Descript. And I thought he asked a really good question. If most of the things that we're seeing right now are playing out the way that we think they will I think his question was something along the lines

of as investors, what surprised us versus a

year ago. And I know you haven't

been an investor for a year, but what surprised you over the last year that

you would not have prognosticated?

How has it

played out differently, you thought?

Fraser Kelton: hmm. I don't think we've seen the level of creativity in how

to bring these technologies into brand new products, like breathing life into beautiful new products. We've seen a bunch of First order thinking of, hey, this can, complete text, so let's do copywriting, or let's write an essay those types of things .

It, maybe the parallel is we, we have seen... The equivalent of Instagram and all of the photo sharing applications. And probably somebody will build a nice business there that's, that's very apparent. What we haven't seen is the second order thinking like, Oh, you have this device on you that has location.

Now we can summon a car and let's do Uber. I went, I went to OpenAI because I thought that we are going to make this technology accessible and when it becomes

democratized, the real, beautiful, creative people out there are going to show us all of the wonderful things that can be built with it. It is, it is remarkably accessible. Somebody else last night, this is a funny aside, they said that they have no problem fine tuning the world's most sophisticated model, but they have no idea how to do very small ML traditional models, which I thought was a wonderful place statement of where we

Nabeel Hyatt: Yeah.

Fraser Kelton: today.

But yeah we haven't seen, have you seen anything creative that just knocks your socks off?

Nabeel Hyatt: We have seen some things. Founders that we're talking to now or some of the investments that we've made, but you're right. I think I would have assumed it would be a Cambrian explosion of the craziest, weirdest ideas ever as people are playing with this new tool that is really this alien amongst us. I, I with these use phrases like alien amongst us, the thing you expect the next week is, is going to blow your mind.

But, I also have to remind myself it's early. To use your mobile analogy, uber is two and a half years in. We're not even a year, we're barely a year out of ChatGPT.

And we're still in fart app land,

Fraser Kelton: Yeah.

Nabeel Hyatt: Like, we're, we're still in flashlight land and calculator land. And like there is, we're still in folks trying to learn to live with a thing for a little while before they really get to second order thinking. And so I think that's just the way it is.

Investor influence on founders

Nabeel Hyatt: I think we also want, I think the investor class quite frankly has contributed to this problem. To be honest, we have a whole bunch of investors that became much more industrialized over the last four or five years and they're factory farming B2B SaaS startups.

They have absolutely no taste. And so a new crazy thing happens in the world and their instinct is it's a gold rush. We should sell picks and shovels and that then I don't have to make a decision about anything. I just go do picks and shovels. And I'm not saying we'd never do a picks and shovels business, but it's better being highly differentiated.

Because you're talking about a situation where the model layer is already changing every single three weeks, every single three months. You have some new open source model comes out, then, OpenAI decides that they're going to release a multimodal model .

And so in a world where the bottom layer of the infrastructure stack is changing crazily and frankly, like people at GPT 4 at the team and people at Anthropic working on Claude don't fully understand every single capability of that model. That's the nature of the probabilistic model like this.

And the consumer behavior is changing like crazy, right? We have a situation where a year ago, ChatGPT happens. Every public CEO now has to give some talk. on, TV when they're doing their earnings report about their AI strategy, which means they tap some stupid person in their executive staff to become the AI person.

And that person's happily trying to grab more territory. So they're not the AI guy, they got a budget.

And then that flows into some wreck that flows into some startup. And you fast forward 90 days, that person still doesn't have any idea about what they want really inside of their large org.

And so 90 days later, they're that much more sophisticated . And so if you take a Internal middleware company, say Unity, in games, Unity benefits from the fact that at least one of those two parts, above or below them, demand or supply, is static. Unity was relatively flat as a company for a while, until what comes along is oh, we have this problem, iOS and Android, and I gotta ship for two.

Two, it's not 55, it's two, and we know which two they are, and then you got a long tail of the game developers that are trying to build for those two platforms. Perfect DevTool situation and so I actually, I blame a little bit of, kind of investor class for running around and chasing everything that had a little tiny bit of momentum and applying their B2B SaaS playbook to what is a very, Very unique situation where there's a really, really high fog of war, sometimes lots of founders have a vision that is. Five or ten years long, and it's their calling and this is what they're building. But there's lots of other very, very successful companies that started out by reading a market, and talking to a customer,

and talking to investors, and triangulating, and building, and and a lot of

those companies, not all, but a lot of

those companies, unfortunately, will get swept

by as the

waves just move too quickly.

Fraser Kelton: I don't have anything to add to that. That's, yes, plus one.

Nabeel Hyatt: that's my

rant, but you might have heard that rant

before, Fraser.

Fraser Kelton: No, that's all brand new to me.

The,

I think there's a bunch of different things going on. The technology is so new. We are just inventing new, new things that we can, we're discovering new things that we can do with it. I think this comes back to the discussion about the OpenAI Dev Day where, the challenge is that the models themselves have these behaviors that you're not entirely sure what they're going to be capable of until you get your hands around them and you start playing with them.

And so this is not. Maybe this, maybe this is an answer to like, where's Google, right? Because if you have these processes where you need to have a fully formed PRD that people sign off on, that gets circulated around stakeholders and they all have buying and alignment, et cetera, et cetera, et cetera, you need a very long planning process.

You have a model that trains, you get a checkpoint at some point, you start exploring it, and you're trying to figure out what it can do and what it can't do. You, you, you want to be shipping as close to the model's completion of training as possible, for obvious reasons. You want it to be and not be able to be easily misused, etc.,

etc. But then the, the counterbalance there is that you need the model to be done training in some extent to be able to know what you have. And

Nabeel Hyatt: right.

Fraser Kelton: Arrakis or whatever the rumor is that they weren't able to get something done and they had to abandon it. And then what they shipped is, is,

Nabeel Hyatt: yeah, you can't, you can't exactly plan your announcement

and your pipeline

nine

months

ahead of time to run your show when you don't even know if the thing you're making is going to work and you're going to have to kill some of those things

along

the

Fraser Kelton: That's right. And so if. If the people building these models don't even know what they have until they put it out in the world, and that's

actually such a beautiful thing because then we all collectively can help discover creatively how to mold them and bend them into interesting products, maybe it's not surprising that we haven't necessarily seen anything yet.

It just takes time.

Coming to terms with the AI fog of war as a founder

Fraser Kelton: It takes time. I'm very

Nabeel Hyatt: A part of that that's incredible and wonderful and is an advantage to startups because startups run with agility. Yeah. And large companies win with straight running speed in one direction. So agility favors them. If the world is changing constantly, then that should favor a startup over a large company.

But it's also terrifying, right? It's the world,

you, you try and make a pitch to the team about why you're joining. You try and talk to an investor about why they should put money in.

You've got an internal worldview that makes you convinced you should go quit your job and go do This startup. And then, and then everything changes every couple of months.

That's. That's pretty terrifying.

Fraser Kelton: that's the startup journey. It's always terrifying. This is just a new level of terror that you

have to keep your eye on, right? The game the game is changing. This is new. But the tradeoff is that you have

this beautiful new technology, right? That you didn't have before and that nobody has like fully plumbed and discovered what's possible with.

I would take that, I would take that trade off every single day, by the way. I would take the fear of it changing underneath your feet every couple of months in exchange for every couple of months like some beautiful new piece of technology that we thought was impossible two years ago just appears in the world.

And it not only appears, it's available via a simple API that costs like a thousandth of a penny.

Nabeel Hyatt: yeah, As long as I

think you've internalized the culture that way. I think one

Fraser Kelton: Yeah, yeah, yeah, yeah,

yeah.

Nabeel Hyatt: belies is having a culture that is with constant realignment,

Fraser Kelton: Yep,

Nabeel Hyatt: okay with new plans on a very regular basis and is very execution oriented. Otherwise,

Fraser Kelton: yep. And that comes back to your comment earlier about you need to have... Verticalization is, is a benefit in this world because what does that mean? That means that you can be so close to your end user that you understand what their needs are, their problems are. And then as these new

technologies come along, you can say, Oh, wait we discovered six months ago that they actually wanted to do X, but that was impossible half a year ago.

And now look at this for a thousandth of a penny, we can solve it for them.

Nabeel Hyatt: Yeah, that's right.

What else we got for today?

Fraser Kelton: Try any apps the past week, any products using AI that are worthwhile exploring, discussing?

Exploring Reflect: A New Note-Taking App

Nabeel Hyatt: I've been using, not just this last week, but I have a new note taking app that I have been using for the last, probably, month. I think I've mentioned it briefly in the hallways, but I never really gave you a full download on it. I'd love to talk about it. First of all, the world doesn't need another note taking app, but I try all of them.

And it's actually a very good example in a constant conversation that we have about products, about whether AI is sustaining innovation or whether it's a disruptive innovation. And so the company's called Reflect. It looks a lot like Obsidian or Roam or some other note taking apps when you first open it up.

Clean UI, very well structured. The major differences that are pretty awesome. are, one is the iOS app in particular just opens up and it has a record button right in the middle. And it is a it is a speech to text, I'm walking out of a meeting, I open it up, I just start talking, I talk into it, it automatically turns that into a note.

There are some bespoke apps that have existed, like AudioPen

and some others. That are the let me speak into my phone

and then turn it into a note, company's on their own. But this is well

integrated inside of a full suite. And so it's it's that,

benefit of integration,

Fraser Kelton: is an amazing,

Nabeel Hyatt: like.

Fraser Kelton: you

know, we we just talked about.

The lack of creativity that we are seeing. But here's this new technology that

is commoditized text speech to text, and, it's, it's changing the way that we interact with, with devices, with computers, with machines. So you

Nabeel Hyatt: That's right.

Fraser Kelton: note app

that, there a place, is it, is the prominent feature still to be able to input through the keyboard, the notes, or is it?

Nabeel Hyatt: Yeah, you open it up and it does what, what Roam does, which is it pins you

to the current date. So it's not trying to tell you to open up a new note. It's trying to make low friction. It just opens up to the current date, and you're supposed to just start typing, or you can hit record and just start talking.

But importantly unlike iOS, it's it's not record under the keyboard. It's not tucked in as a third third tier feature. It is first class citizen in the product that makes you feel differently about the product, even though a lot of the rest of the ChRoam looks very similar.

And then, then the second really wonderful thing is they have some of the normal things that lots of note taking apps has, which is allowing you to then use AI actions against

a little co pilot y stuff against that text. Summarize this text, write a title, the kind of normal, what is becoming table stakes in

Notion or anything else that has AI writing.

Fraser Kelton: So hold on, wait one second,

because what you just said is

Nabeel Hyatt: what do you mean?

Fraser Kelton: a MyGPT's. You're going to be

able to go in and create an action that has a record button that then does LLM type tasklets against it, summarize this, or have a button that lets me synthesize

this. Is there going to be a a version of this app? Will this developer be inspired to go build that into my GPTs as a platform?

Nabeel Hyatt: Is there a world where going to ChatGPT is an input into a note taking app? Of course. That's possible. Do I still need a specific workflow to look through all my notes, combine them retrieval, augmentation of that, write a blog post based on the last 15 things I talked about?

More importantly, the thing I love about a note taking app in the modern era, which was spurned by Roam, is the idea of back linking similar concepts. So I end up with a lot of disparate concepts that now get gathered into, hey, what is every single time I've talked about note taking apps? And I can look back five years, and I've got a rolling bit, or...

How many times have I, have I talked about like the nature of trading card design or educational pedagogy? And then can I get a page that kind of gathers all the notes I've taken on educational pedagogy over the last five years?

Whatever thread you happen to be pulling. And I love the idea of backlinking. Unfortunately with Roam and other services, backlinking, it takes time. You're constantly tagging things. What should I tag? What should I not tag? So actually the kind of like second AI feature that I love is that yes, it does summarization and yes, it does normal headline AI, but you can also grab a big section of text, like for instance, me blabbering on in an audio note that just got recorded in their reflect, and then just ask for it to backlink the whole thing.

And it will analyze the whole piece of

text and then add in little backlinks on all the relevant.

Fraser Kelton: Yeah, nice.

Nice.

Will OpenAI be able to be consumer & enterprise?

Nabeel Hyatt: And so now, it's

not just some add on writer,

Fraser Kelton: Yep.

Nabeel Hyatt: it is in most apps,

it's actually helping

you use the product more

Fraser Kelton: Yep.

Nabeel Hyatt: more deeply integrate into the surface area that the product is

already

laid out beforehand. And I love

that.

Fraser Kelton: Beautiful. That, that type of like depth of functionality and interconnectedness is why I

don't think, I

don't think there will be know, a platform that has all of these first party products sucked into

it. I think that you're

Nabeel Hyatt: Now, oh,

Fraser Kelton: there's clearly a need, but then if there's clearly a need, who is going to supply the API for all of the AI that's woven into that product?

you don't think there's a developer business for OpenAI long term around that?

Nabeel Hyatt: no, I'm not saying that there's not a developer business for

OpenAI. I, I just think that it's very unlikely that it will be OpenAI. I, I would take the bet again. Because I do think... It's so important. That it needs to be some company on the planet's job one.

Fraser Kelton: Mm hmm. Mm

Nabeel Hyatt: And, And, and, it's very unlikely that if it's somebody on the planet's job two or job three, that they won't get lapped, eventually.

And, and in the case of OpenAI, I just think ChatGPT creates so much gravity around the or look, if you could just, if you, if you could go work at OpenAI tomorrow, which I, you definitely don't want to do, but you, cause you're here. But if you, if you were going to OpenAI in this chapter of its life not when you went.

Way back when, when OpenAI was still a research org trying to figure out how to be a product org. But in today's post GPT, post large volume model, like this world, if you were to join now would you rather work on ChatGPT?

Would you rather

work on AGI? Would you rather

work on core

models? Or do you want to

like,

work on the rate limiter for the API?

Fraser Kelton: I'm not answering the

question. There are some...

There are some...

Nabeel Hyatt: Or, I don't know, the like, corporate fine tuning area

of it's just it's an org that is oriented around certain priorities. And so I think it will

be somebody else

that will win that game. I don't know who,

Fraser Kelton: Yeah. This is actually what I think is the most

amazing piece of this moment. Is... There is an organization that is trying to run the table across a number of

different monumental businesses across

different markets that are at odds with one another. I don't know, I don't know if there's an example of an exceptionally great consumer product company that has also been a developer platform that's first party, like first rate, first rate.

And there's like a bull story, I can tell you the bull story,

Nabeel Hyatt: way,

we're both

our explosive, brand new, multi billion

Fraser Kelton: right, right,

Nabeel Hyatt: of war. No

Fraser Kelton: right. And then all of the tensions that exist within a relatively small

organization that is, still relatively resource

constrained to do

those two things is, there's a bull story for why that can

happen. I think the traditional story is a bare story,

that you can't do both of those at the same time, and it's a future that you foresee happens.

I think, I think that they're as well positioned as anybody in history to be able to do it, to be able to run the table, and I don't think we know yet.

Nabeel Hyatt: I think doing too many things will make all of them

mediocre. And I hope OpenAI does one well versus seven increasingly worse over time. But We'll see.

See. Oh, the last little bit, by the way, of Reflect, I have to get this in, which, because it loops back to the previous week's podcast, which is the other very interesting thing about Reflect is that I have been using Snipd as a podcast service.

Lately, and now I can triple click and create a note. The note that gets transcribed,

speaks to text by AI, then gets dropped into ReadWise,

and then ReadWise has a sync directly to Reflect, which then drops into Reflect, and then gets backlinked to all the other concepts. And so I've got... I got...

I'm stitching together an aI future.

Nabeel Hyatt: I, see. I see. You're saying if you check back into me in a year, I'll be talking about

the apple Notes app

again. It's

Fraser Kelton: I,

Nabeel Hyatt: just so clean and simple. Fraser, it's just, I

just open it and it works.

Fraser Kelton: yellow. It looks like a notepad.

Nabeel Hyatt: Exactly . Let's wrap it for this week, man.

Fraser Kelton: cool. Thank you. See

Nabeel Hyatt: to you next week. When we'll see who announces what. Take care.