The universal adoption of this as a consumer technology and then bleeding into prosumer is it exceeds anything I've ever experienced.
And I think it is, it will just fundamentally change people's sort of daily pattern.
This is all early adopters and early adopters are very forgiving of mistakes on purpose.
When something is brand new, a culture around it develops.
The early internet people didn't complain that the internet was slow.
Right.
The more senior small teams that use AI are superhuman.
Yeah, yeah, yeah.
It's like they woke up and they were all f***ing Tony Stark.
It's unbelievable.
And like their productivity is insane.
Should the U.S. put a price on H1B visas or would that shut out new talent?
Are AI coding agents truly boosting productivity or just heightened?
And in this AI platform shift, who wins, incumbents or new AI-native startups?
Today I sit down with Box CEO Aaron Levy, alongside A16Z's Steven Zinovsky and Martin Casado, to debate H1B reform, why Box now ships a third of its code from AI, The move from writing to reviewing code, and why bottom-up AI tools beat top-down pilots.
Let's get into it.
First I just want to comment.
You posted in the group chat that the news around autism updates your PDoom.
It only works if you show the image though, so you'll have to do the overlay to make that make sense.
But there's so many memes you can do with that Fox News headline.
Exactly.
First, I want to get into the immigration news.
Oh, you really want to kick off just like really with the fun stuff.
Exactly.
Martin, you had some interesting reactions.
Please, what were your reactions to what you think of the policy?
Well, it's interesting because it seems like any time the administration touches immigration, there's a huge outcry, a knee-jerk outcry.
And we saw a lot of that from VCs even.
But it's also very interesting that Reed Hastings, who is a classic lefty and has long been, was like I've been doing policy for immigration for 30 years, and this is the right approach.
And this is very much my thought, which is...
This system has been game for a very long time.
It's very hard for startups to hire because of the lottery system.
It's locked up by the large companies, the consultants, Amazon and Google.
And that has to change.
And I think a very reasonable way to do it is to set price, because you've got a market and you need to allocate supply.
Price is a great way to do it.
So I'm very, very positive on it.
I comment about that and a lot of people seem to disagree.
So I think it's an active discussion.
Well, there's a couple of elements to this.
So one is first of all, Reid was ultimately responding to a thing that was no longer the actual policy.
So he said 100K a year was a great policy, and obviously the internet had moved on.
It's not, to me, obvious.
I wouldn't conclude the same outcome that you just concluded, in that I think that you'd have a situation where the Amazons and Googles would probably actually capture the vast portion of the talent.
In this situation.
So it's not clear to me that like startups sort of come out ahead.
Well, we'll step back.
From this particular implementation.
Maybe Amazon and Google who are probably more easy to regulate.
But there are a number of organizations that are consultancies.
Yeah, 100%.
That actually are price sensitive.
Yes.
That would be squeezed by this.
I would think that given that they're in the top 15, they make up like four or five of them.
That would be a significant freeing up for a higher level.
I think the my thing would be like if you could just get all the people in the room that have an opinion on this topic and you, but you actually have the practitioners in tech in the room as well.
And the, let's say, the most kind of you know.
You can't even say like right wing, because actually I don't think this is even classic Republican.
So it was just like the polls if you got everybody in a room and he says he's going to say what are we optimizing for?
Are we optimizing for, we don't want to have wages go down.
That's an interesting thing.
Are we optimizing for a particular kind of job, not going to, let's say, certain populations of Americans?
Are we optimizing for just ensuring that we only have the highest merit people on the planet coming here?
Like those are all totally different kind of goals to optimize for.
And I think that the framework you end up with and the system that you end up with should probably hopefully have like a cohesive sort of strategy behind it.
My strategy would be we want the absolute best in the world here.
There's not exactly clear that there's a fixed number on that.
Some years there might be 5,000, some years there might be 50,000, some years there might be 80,000.
We probably want them to be net positive to wages.
So let's agree that you know, in any given industry or locale, wages should go up with this talent pool, as opposed to down.
So I think that's actually totally reasonable.
And so you should have the market kind of, sort of some market dynamic to that.
And you shouldn't be able to kind of game and exploit the talent pools for saying, now in Detroit we can go wipe out IT jobs because we can go and offshore those.
Like I think you could build a system that basically meets all of those goals while still ensuring that You can get somebody that goes to their master's program and name your state school.
They come out of it.
They're an AI engineer.
They're not yet at the sort of meta is going to pay them 100 million, but they are going to be totally valuable contributors to our economy.
It's all sort of positive sum.
It's not taking a job from anybody else.
It makes us more competitive.
And I think there's a way to do that without sort of overly, let's say, putting constraints in the system that make it maybe so a startup wouldn't be able to kind of economically viably participate in this.
And I think 100K per year would be at a point where the startups would be directly impacted.
Working with a lot of startups.
I'm not sure that's still Respectfully.
The kind of startups that Andreessen Horowitz sees are not all of the base of startups in the world.
So you could quibble about the number.
Is it 20K, which Keith Rebois said was very sensible, or is it 100K?
I don't know.
But like the idea that you get... Wait, Keith threw out 20?
Heath threw out 20.
Well, let's just go with Keith's number.
If Keith threw out 20, I think we can be good with a Keith number on this one.
I think the number, it's easy to fixate on the number.
Yeah.
But you have to always look at what is the number replacing.
And I don't think the average person having this debate, other than the people that really work at this, have any idea the amount of productivity that is lost working this system.
I mean, the incredible amount of resources.
And, of course, the bigger companies, the ones that you mentioned, have enormous teams that spend all of their energy like literally essentially as lobbyists working this system.
And then the back end of that is are all the justifications and all of the management and all of the, And now they've just deployed all their resources to manage in-house call centers to deal with getting their employees back to the United States to just deal with it.
You hit on one point that I think is really important to this debate that is sort of getting lost, which is there's no doubt that within the big tech world that they, for the past 25 years or so, they really went on this sort of bifurcated curve which is hiring for the people in the office, focusing on say, 25 or 30 university departments.
And then basically everybody else was like well, it's so much easier if we just hire huge numbers of people from these eight international locations and schools.
And I think a lot of this is missing that a big part of this is well, the tech industry, like if you look at Intel and where they all went to college and if you look at the history of Silicon Valley.
It's all these people from all the schools in the middle of the country, none of which are like the target recruiting schools by the main tech companies these days.
And that's been a place where I think that the big companies have been somewhat lazy.
And as a person who spent decades flying to all of these schools and recruiting, there's work that the universities have not done to be better programs, and that there's work that the big companies have not done to be clear what it is.
That why they've stopped recruiting or haven't seen the numbers.
And that change would be better for everybody to really make.
So my expectation of what gets impacted is kind of an even different job set than that which is.
If you go to Florida today and you try and get like an IT job for 100K, you just can't.
And so that I think is actually the area that's the most directly impacted by the large consultants.
Meaning there aren't jobs that pay 100K or that you just can't find a job?
They're all taken.
They just don't exist.
Like any sort of IT administrator, like the services, like the basic consulting gigs, like all of that has been saturated.
It's very, very tough to get a job between like 80 and 120K in much of the United States because of this right.
And so this isn't about a new grad being a software engineer, because the reality is the expected value of a software engineer over their lifetime is high enough that I think that, like the market kind of navigates that.
But...
It's almost these kind of lower level more like IT admin jobs that have been squeezed out.
And listen.
If we want to bring them back, then I do think and we don't want to do this kind of arbitrage that a lot of these companies are doing that we're going to have to change the pricing.
But wouldn't a minimum salary band effectively solve that problem for you?
Sure.
Yeah, for sure.
Yeah, I think there's a lot of mechanisms to do it.
I actually agree with you.
We should actually talk about the problem that we're trying to solve.
So I think we would all agree.
If you come to the United States, you get a college degree, you should have a visa.
That's for sure.
Do we all agree on that?
Okay, yes, I agree.
I don't think that the people proposing this strategy actually agree on that.
Well, so Trump famously said this.
He famously said a lot of things, for sure.
And then he unsaid it.
This wouldn't be an interesting debate.
If we all agree and we have a policy that says if you go from IIT and then you come to Kansas State, Then you get a job.
But that became part of the gamification of the system, because you would do the IIT thing, then you would get a company sponsorship for a master's degree at some school.
So it didn't really accomplish the goal of investing in coming to the US the same way that going to a four-year school would have done.
Wait, why is that?
What was the problem?
Because you ended up getting sponsored by a company.
Yeah.
And it changed the whole dynamic of were you seeking out the US or did a company pull you to the US?
It's a little bit different.
Okay.
But I also think so often this discussion goes the direction this one is going on is we focus on, like new grad software engineers?
And I actually don't think that is getting impacted right.
I really don't.
I really think it is like what do the consultant shops do?
They do yeah, admin work, it work, and it's just a different salary band.
Yeah, i think these are, these are body shops and like an approach like this directly targets them.
I think in a way, that's actually quite positive.
Yeah, I just think there's.
I mean, I would just then favor Keith's approach, because I think there is a number in which it becomes You're then making other trade-offs in your business just to be able to Yeah yeah, yeah.
The whole system can do without this immense cost and uncertainty.
And any solution should really be.
If you address that, then the whole, the rest of the dynamics will follow.
But as long as it's a huge complicated, expensive system, then the big companies are going to continue to benefit from it disproportionately.
And I will tell you right now like it is much harder for a startup to deal with the lottery system than it would be for them to pay 100K.
Like, at least for the startups.
But I don't think it changed the lottery system.
Right?
Right, it didn't.
It's $100K to participate in the lottery system.
Right.
Oh, I understand.
Hopefully, it'll change the calculus of a lot of people that are in the lottery system.
But I would prefer to remove the lottery system.
Yeah, I think you can absolutely pull off a system that for probably in a shared definition, for us and anybody else would say this is clearly a high merit job.
It's going to increase the wages in this particular sector on average.
And we want to make sure that we've got the best talent in the world that comes in to do that.
It's not going to drive down wages.
We probably want as many of those individuals here as possible.
And so I think you know, some elements of this are intriguing in that they push the conversation forward on the dimension of, you know, like the 100K is like a hammer to do that.
And maybe there's a more nuanced approach that I would certainly prefer.
But again, it's important, like the 100K will scale with the skill level.
So the higher the skill, the more like that is amortized.
And so you could argue that this is just kind of a dial to get higher skills.
Well, once you put a dollar amount on it, you have to keep in mind that people are going to pay it.
That's right.
They're going to make up their own mind.
It's an adaptive market.
And the skill might not be relevant to some people.
And so that's the tricky part.
Yep, yep.
I want to segue from labor markets to What's the next really interesting political topic that we can engage in.
Yeah, exactly.
From labor markets to labor productivity with AI.
Offline Aaron, we were talking about the meter paper, and the paper suggested that their developers were actually less productive with AI.
But that doesn't square with your experience talking to a lot of different startups, seeing a lot of different startups and how they're so much more productive.
So why don't you talk about where you're seeing startups say they're more productive, and why is that happening?
Yeah, so I'll first just represent our own case study, and then there's the really extreme version.
So our own case study is we've adopted a few different kind of AI coding tools.
You know, Cursor being a super, super popular one internally.
And you know, as I talk to people, let's say, in the hallway, who have you know yeah, maybe they're trying to get me excited by AI but, like I think they know, I'm bought in.
So so the the kind of qualitative answers I get from from people, then I'll give you our internal metric.
You know, some, some people say I'm getting, you know, a 20, 30% productivity gain.
Other people will say 75%.
Interestingly, I have not been able to pinpoint the demographic difference on the answers.
Oh, but this is self-reporting.
This is self-reporting.
How happy are you?
So, no, but we have internal metrics as well.
So about 30% of our code right now is coming from AI.
So we've got the... 70%?
30%.
So we have some of the kind of pure internal metrics that show this.
But what's interesting is that, like I have a two by two of, I have senior people that are saying that they're, you know, getting 75 productivity.
I have junior people that are saying they're getting 75% and then vice versa on the 25%, let's say.
And it's.
I haven't been able to quite figure out a pattern, maybe except for and we've talked about this a little bit, but you kind of see this online except for maybe the biggest criteria is just the people that actually push the AI to do more, which is sort of this other new kind of psychographic, which is just like who is willing to just be like you know what.
I'm going to YOLO this task and just see what the AI comes up with.
And your sort of willingness to just do that, I think probably somewhat then shows up in the ultimate productivity gain.
So that's us as a relatively larger company on the startup side.
Oh, it's crazy.
And this is the thing that just blows my mind.
I will regularly talk to three five, 10 startup founders that self-report.
They might be getting somewhere on the order of like three to five to 10x productivity improvements.
And...
And the big difference is that a year ago, if we were to have this conversation, the conversation would be about AI doing type ahead.
And it can add maybe a few lines of code to your productivity per incremental unit of work that you give it.
And then now obviously the big phenomenon is background agents, where i give it a very, you know, detailed prompt.
I send it off.
It comes back.
You know people talk about it as like a slot machine of like some percent of the time it's not going to come back with the right thing.
You have to decide which what you actually, you know, kind of pull in from it.
But the kind of startups that are getting like real multiples of productivity gain are are just, they're fundamentally engineering in a different way.
They're sending off a task, task goes off, comes back in 20 minutes and then they're really in the business of doing code review, not code writing.
And it's going to obviously change.
You know quite a bit of what computer science looks like in the future.
And then the only question is like, you know, what are all the things that that's good for?
Where does that break down?
What kind of teams can actually evolve to that state?
But that one has been blowing my mind the most recently.
And I think that kind of fundamentally changes what the future of you know kind of engineering looks like.
I think what you said is super interesting.
Let me ask you I think that there's an overlay that goes beyond junior senior, which is we're all talking about characteristics that have two, a solution that has two really important characteristics right now.
One is that it's engineers doing stuff for engineers and they understand the domain super, super well.
And I think that that's a really, really important part and a really big thing that people aren't talking enough about, which is maybe what's going on, is that you have AI accelerating for people that work in the domain and are very smart.
And then the other is we shouldn't forget and this is to your self-reporting a little bit, but also which is that this is all early adopters, and early adopters are very forgiving of mistakes on purpose.
And it's a super interesting dynamic where, when something is brand new, a culture around it develops which just lets anything happen.
I mean, you know, like...
The early internet people didn't complain that the internet was slow.
The only people complaining the internet was slow were the late adopters who were like wow, this is so much slower.
Or take online video, which was like, I'm watching this tiny postage stamp video.
And the early adopters were like, this is the coolest thing I've ever seen.
Everybody else was like, why would I want to watch anything like that?
And the same with downloading music.
And so I really feel like...
What's going on is just this incredible.
Do you remember that, you guys?
Do you remember this watch that you guys made?
The Spotwatch?
Yeah, Spotwatch.
So in like 2004, 2003 or something?
Something.
So I bought one.
And it used like AM or FM... Unused FM radio white noise.
So you could get like stock quotes, 45 minutes delayed if you were outdoors in a field.
Yes, and so I had one.
And I was in this camp, like, this is the coolest thing in the entire world.
And obviously, this is going to be the most mass market product of all time.
And, you know, I was like 20 years too early with the Apple Watch.
No, turn-by-turn GPS.
Like the first time you could put turn-by-turn GPS in your car.
It was super cool, except for the fact that most cars move faster than the ability for the computer in the car to calculate when you were going to turn.
Right.
And so you just drove around.
It was like a U-turn machine.
But I think that it's just so interesting because I think that people use the AI tools today and they assume, like well, I'm not a doctor.
I don't know anything about being a doctor.
Let me ask you to cure me.
Right.
And diagnose me.
And it's like, whoa, that is the worst.
And just like all the people who see failure, they weren't great programmers to begin with.
And they didn't know how to ask.
They didn't want to review it.
And that's just how you do it.
I think there's an aspect of this that makes it very difficult to measure.
One of them is and I don't think it's just an early adopter thing like these models are so magic that you get dazzled.
So even if it's not what you want, you're like, it was great, you know?
And I think it's very easy to complete that with being productive.
Like, it's not what I wanted. but it was amazing.
So therefore, it must be great.
So maybe over time, we just abdicate having an opinion, and the model does everything.
But right now and I see this a lot people are so enthusiastic about using AI, but it really hasn't impacted their output.
They're just enthusiastic.
The second one is, I feel like...
There's almost shadow productivity.
Oh, sorry.
How would you verify that with a five or 10 person company who kind of empirically is operating at like a 50 to 100 person company?
Like you.
Just you can see the sheer scale of their code and you're like okay, you could not have done that 10 years ago.
I actually agree with everything that Steve was saying.
So listen, this is anecdotally.
Anecdotally, I work with a lot of companies.
Anecdotally, the more senior small teams that use AI are superhuman.
Yeah, yeah, yeah.
It's like they woke up, and they were all fucking Tony Stark.
It was unbelievable.
And, like, their productivity is insane.
But they're all super senior.
And look they were.
I don't want to take anything at all away, but those companies were also incredibly productive relative to a 10-person team at a big company.
Because there's no code, and they're starting from a clean slate.
No, they really wake up.
Yeah, yeah, for sure.
And, like, you know, they're very senior.
But they're also almost to a person who are AI skeptics to begin with and are incredibly sober about the value.
And so they just use it in these very pragmatic ways.
There's one other category that I'm seeing because I just want to be intellectually honest on the full spectrum.
I'm seeing these 19 and 20-year-olds that are like, first of all, I don't know what is in the water at, like Stanford, MIT et cetera right now, but like everybody's dropping out.
So like just like literally like people are going there for like a week just to drop out.
But there is a tendency of that cohort.
It's the 996 people.
And there is this tendency which is, you know, they would have been maybe 10x engineers in a prior world, but now they're like 100x engineers.
And so you know senior in terms of in their own kind of relative cohort, but like the way that they are building their startups are just like completely different than it's probably the, if I look at you know so we dropped out of college God that's scary 19 years ago.
If I look at how these companies run versus today, it's the biggest change in how you start and run a company that I've ever seen.
Yeah.
And like, and I think if you looked at like in 1995, if you were to drop out of college, versus 2005 when we dropped out of college, I don't think you like, I don't think like the company building process was all that different.
You get into a garage.
The internet fundamentally changed.
No, no, post-internet, post-internet.
Okay, yeah.
So you're building.
You have to rewind to the internet because that was a.
Yeah, 100%.
We're not in 85.
So at 95, you're dropping out to do an internet startup.
Okay, you drop out to do an internet startup.
By 2005, other than the fact that our resources were in the cloud versus you know you'd have to go to a data center.
In our case, we actually still went to the data center.
Like, not that much about the company building process was different.
Today, in 2025, everything about how you're starting your company is completely different, because of AI.
I think that the key, the through line in all that is velocity.
Yes, exactly.
And I think that that cutoff, the internet increased velocity.
Yes.
And AI increases velocity the same way.
And I think that that's just super, super important to what's going on.
So basically, if you go back, even in the early internet, like when you started a company, There was still a lot of old school like what's your business plan?
What's your plan?
We're going to be stealth for two years.
There was all of this stuff.
And it was really Mark and Ben at Netscape that changed the velocity of how companies work.
And the cloud was an accelerant to that acceleration of velocity.
And AI is a refactoring of how velocity works.
Yeah, but what was interesting is even in the cloud, like that was this great virtualizer of the physical stuff you would have to deal with.
But that was a two-year build-out.
Yeah, exactly.
You know, like, you had no... You could have customers.
Yeah, 100%.
As soon as you had code, you could have customers.
You're right.
I mean, PLG and a lot of these kind of high-velocity startups actually started pre-AI.
Yeah, but they're pretty, like, phenomenal.
Like, think about, like, GitHub.
Think about, like, Slack.
Think about, like, Figma.
Like you know, you have some pretty remarkable companies that came pre-AI, that were drafting on it, and a lot of that was like basically cloud and then the SaaSification and then unlocking kind of different routes to go to market.
Zoom was a great early example of this.
The thing that AI is is an accelerant in building the product, whereas what SaaS and cloud did was accelerate getting paid, which was a thing that used to take two or three years.
I mean, arguably, the cloud also made it much quicker to build a product because you would have, like, these big... Well, it made it much quicker for customers to all have the same one.
I don't think it made it like five times faster.
Fair enough.
So like I think it was like it was like yeah, like you can see your website a lot quicker, but like I could see a website in 98 pretty quickly.
Yeah, but well, maybe this is an infrastructure thing, building a big distributed service.
No, no, sure, sure, sure.
I was not doing distributed infrastructure in 98, so.
Yeah.
So I think AI productivity is hard to measure for two reasons.
The first one I just mentioned is it's just really dazzling.
So I think people kind of feel like oh, it's amazing.
The second one is I think a lot of the productivity is actually hidden and people measure the wrong thing right.
Shocking that people measure the wrong thing in productivity.
That's right.
Literally the history of productivity measurement.
Yeah.
But also what happens here is, like, you have the board, and the board is like, we need more AI.
And then so what happens?
And they go to, like, some CTO or some innovations lab, and then the innovations lab do AI.
And so, like, whatever.
They, like, bring in, like, build some internal tool, and it'll fail.
Of course that'll fail, right?
But the reality is, is, like, this AI wave is so personal.
Like... probably most people in the company are using ChatGPT.
Probably there is, you know, some personal assistant.
Probably, you know, they're using Cursor or some coding thing.
And that's much, much harder to measure just because, you know, it's not advertised.
And so if you actually look at the reports on, like, enterprise things fail...
You guys look at what they're measuring and it's like yeah, clearly some internal project pushed down by the board, where they hired some consultant to do it, is going to fail.
Those always fail.
But that's actually not what's going on.
The movement that's happening is a very secular movement.
This is the next time that bottom-up adoption is really changing the productivity equation.
And that's a thing that it defies.
Big companies do not know how to deal with that because they need to control it.
They worry about safety and security and privacy and all of their corporate rules.
And then also the other thing I think to overlay on, that is, AI is a very unique if that's not a bad way to say things innovation in that it's like non-deterministic.
And so, all of a sudden, you have this very personal and non-deterministic thing which the real problem in a large organization to all these pilots and AI projects is that you can't not just measure, but you don't even want to put out there a non-deterministic solution.
Because your whole thing is like, well, we operate at scale and we have 60 countries.
We can't put like a customer support solution out there if five agents in five languages all have different answers based on the way that the customer or the CS agent asked the question.
And I really feel like that is going to be the hugest challenge in large organizations figuring out how to adopt things.
They're going to just get all bunched up over nondeterminism.
Yes.
Yeah, I think that we will have to have some new form of measurement, probably in general in this, because so much of the improvement in productivity will sort of be this subtle change of like.
I used to go to Google for that.
No, my EA is writing emails using chat TPT for me.
And so where is that showing up?
And other than just like again when we talked about this in the last podcast.
We're going to just start to do higher levels of work.
And that will just end up looking different.
But you won't be able to be like okay, how do I measure what the productivity was when it's just like I'm working just totally differently?
But, I mean, Coda, I think, is such a great example.
So, like, what do people, in my experience, the more senior folks, actually use AI for code?
It's like documentation, writing, testing.
I mean, it's a lot of the other stuff that may not actually like increase, like the shipping schedule, but like you get a lot more robust code, a lot more maintainable code, a much better architecture, a much better future-forward architecture.
So we could be building tremendously better software but still be shipping features at the same velocity.
We might need to just measure like quality of life as another metric of like.
Literally, I'm just happier.
I don't have to do stupid shit.
Developers don't want to write documents.
Exactly.
But this is, you know, it's so easy to get accused of hyperbole and overstating it and things.
But what I think is just so key to what's going on is this is what you were talking about, which is like the whole notion of what you're going to do with the job is going to be really different.
You know, this is my obligatory super old person thing, but the pre-spreadsheet, post-spreadsheet is a perfect example of this.
Classic example.
I didn't say classic.
I said classic.
But before spreadsheets.
You would be a banker and you know you would be like I'm supposed to help this company get acquired.
And you would come up with a financial model and you'd have 50 you know recent MBAs all churning away with their HP calculators figuring out the financial model.
And then you go like, okay, let's do this again, but if the interest rate changes.
Or if their source of funds change.
And you're like, okay, well, that's like a week.
And everybody has to do it all over again.
So you actually, your quality of decision was really bad.
And so what happened was that just in 1985, like that job just completely changed.
And by 1990, instead of asking the recent grads to do it, you were doing it yourself.
I actually remember this absolutely crystal clear.
My two cousins were University of Chicago MBA grads in 1985.
They did not use a computer to when they got their MBAs.
And when I was talking about going to Microsoft two years later, they were like well, you know, we have these kids who use the computer for us.
And they ended up using computers and stuff but, like their whole notion of banking was defined by this multi-week turnaround.
And then all of a sudden, it's just more interns doing their Lotus 1-2-3 thing.
And this is it gets to the.
You know what's going on with code and startups is it's just.
There's just like a whole different mindset over how much you can do and how soon, and iterate.
Figma and Dylan, they're always talking about this.
Like.
Figma is going to now change the trajectory of a design idea to go from like oh, let's iterate over here.
To like, let's just do.
Yes.
I mean, this is where it's like.
Again, a weird thing to think about the percentage of my productivity as an example.
My day is not representative, obviously, because I'm bouncing between too many different things.
But there'll be so many times where it's 10 pm.
In a prior world I would have sent off a task to somebody you know, some analyst or chief of staff, you know type, role go research this thing.
It comes back three days later, and then you find the answer.
And then now it's obviously just like kick off a deep research, a cursor, to generate a prototype, do some kind of analysis, and then you have it back in 10 minutes or 20 minutes and I've just compressed.
Whatever that was kind of then going to be serially connected to that task is now just fully compressed.
And so then by the morning, you're now kicking off whatever that project was.
Again, like, nearly impossible for me to peg, like, what is that as a number?
It's just like a fundamentally different thing on just what work looks like.
And because you just compress so many different steps of a workflow into a single action.
And so it's just like a completely different way of thinking about work.
Where does that fit in for you in what we were talking about?
What I was asking about earlier, which is, how does the expertise your expertise really contribute to that?
Yeah.
And, in particular, I think it'd be interesting for people to understand like, when you talk to customers, how do you help them to avoid trying to get people to make AI, make them do jobs they couldn't do in the first place?
Yeah.
Because that's an easy point of failure.
Yeah, I mean, it actually is.
This is this really counterintuitive thing where and you've talked about specialization on the last one is like the biggest gains of AI go to people who have some degree of expertise in an area to know what is actually true, what is not going to work, what should I integrate from the output of this AI?
Yeah, you know, like what are the 2 of things that maybe are hallucinations or, you know, took the data in the wrong direction?
If you don't have a deep understanding of your particular space or field or domain, you aren't able to then have the right judgment to make all of those decisions.
So I think the experts just get more powerful in this world.
And so I would.
That's why I'm not even like convinced that you can tell a college student to learn anything different than any other.
You know period in history.
Like, be really good at a particular field.
And then AI is merely a turbocharger of your capability in that particular field.
But like, if I didn't know, if I didn't just generally know the things I know about SaaS which is obviously a really weird expertise, but I'm okay at understanding SaaS then the things I give to a deep research agent that I then go and incorporate back into work wouldn't make sense to me.
I wouldn't have all the context for that one thing that it mentioned.
How do I form that into the overall strategy?
But because I have some understanding of this particular industry.
That just makes me way more productive.
So I don't think expertise goes away at all.
And I think any of the experts in their particular area just become more powerful.
So we actually have a fair bit of anecdotal market data on this.
So it's very, very interesting.
So if you take like a lot of these, let's just take kind of a non-text example, like image or video.
And if you look at the customer base for any of the popular platforms, very interesting.
So if you draw a dollar at random that's monetized, it's from a professional.
And for obvious reasons, because they can produce.
If you draw a user at random, it's casual and it's in the tail.
And so it's fairly clear that this is a prosumer movement from monetization.
And I'm associated with a number of companies that work with say, professional designers or professional creatives.
They spend just as much time on the AI tools as they would on traditional tools.
It just turns out the output is far more rich, and it tends to be.
They have a taste, that is Yeah, but it's still human taste.
There's still very specific requirements.
And so I think, if there ever is an ads model that ever shows up for AI, I think that there's going to be a long tail of people that want to use AI but they actually don't have the financial incentive to do it or it isn't tied to their actual job.
And we're already starting to see that bifurcate out and there's going to be another subsection that do.
And my sense is, let's say I'm writing, whatever, I'm a casual developer, and I'm writing a 3D game.
And I want to have a 3D asset.
I've got one of two choices.
I can have AI create it for me, or I can contract a professional to do it.
Like me, as a developer, I'm not going to create.
Even if I use AI, it's not going to be a great 3D asset, right.
And so I think that you know you're going to have that same line that you have today is either you can hack something up yourself or you can go with a professional, and the professional will be using AI.
And I think what's super exciting is that AI has created a third category, which is I am...
I'm not trying to do this as a job.
I'm never going to monetize this.
But there's some product utility gain to me that is worth $20 a month.
So my ability to now generate prototypes when I'm by myself just brainstorming.
Again, I'm not going to do anything in the ultimate delivery of that into any functional code.
But it's worth $20 for me to be able to go and realize the thing that I'm thinking about.
Yeah.
And so there's, just like all new ways of capturing TAM, because there's all this utility that gets unlocked.
And there's one more thing that's worth noting, which is people that are interested in an area get there through AI right now.
It's like this is the number one opportunity for somebody you know that wants to enter an area to do it as an AI native right.
Because A like the tool actually can teach you, you know, and then there's just such a paucity of talent out there that you can fill.
This is also I mean, this is the history of productivity in general, which is the more tools that you have available, first the experts use them and then more people are able to become experts.
But that's part of managing your own career.
I love that point you made about, like, you still have to be really good at something.
And I think that...
You know, if you want to be good at finance or at sales, you should become really good at it, and assume you're going to use AI for that.
And you're going to be better than the people pretending.
And I think that that's very much like, if you just take, you know we were going back and forth about you know oh, it makes a PowerPoint slide deck.
Well, it turns out to make a good PowerPoint deck is still a skill.
And people still pay McKinsey huge amounts of money for better PowerPoint decks with better pictures.
So listen, I contract a lot of work for videos and art and I have for a very long time as part of different companies.
Here at A16Z I have seen the shift for contracting something that uses traditional techniques to AI.
Dollar amounts are the same.
And so again, this is Jevon's paradox all over again.
They spend just as much time.
It's just the output tends to be more dazzling or whatever it is.
And you should see more versions of it.
It should run more simulations and other ideas.
That's right.
You've got more iteration with them.
You've got more control.
Listen, I can have a video where I'm like, I want a dragon to fly out of the sky.
So you have a lot more control as the customer.
But for sure, this is not somehow dropping the cost of output.
Were you about to jump in?
I want to circle back to a point you made earlier about that.
There are 20-year-olds who are building companies in new ways.
Because I remember a few years ago, I think, Patrick Halston and a few others were asking hey, where are all the Gen Z super successful founders?
Remember that?
And of course, there was Dylan Fields and Alexander Wang, but their companies took a few years to really work.
But now you know we're seeing the Cursor founders, the Mercor founders sort of you know, get to massive scale in a very short period of time.
And maybe it was that the foundation model companies required, you know, a certain level of, you know, experienced founder because of the fundraising amounts and maybe the applications are, you know, more conducive to younger founders.
But what's your reflection on this?
Well, I don't remember exactly the date at which he mentioned that, but I do think there was a period in the sort of mid-2010s to early 2020s where we were actually in kind of a bit of a lull as an industry.
And the reason for that was like, like we kind of did like check off a lot of boxes of like the core things that people needed in the world.
And so we like checked off like a lot of the like once you have slack, you don't need five other chat tools.
Uh, once you have zoom, you don't need five other video conferencing tools.
And so it gets kind of derivative.
You know past these kind of core platforms and so once you have like sas, you know kind of check off all the major like things you do at work.
And then in the consumer world we had ways of delivering food and listening to music and watching videos.
So there's not an infinite set of things that we do as consumers.
Then what is the 20-year-old founder supposed to work on?
It's like you have pretty finite opportunities as compared to in the mid-2000s, let's say.
The whole world was open.
You could start anything.
And because every single category had to be reinvented kind of post-mobile, post-cloud maturity.
So we now have that era in AI, and that is why I'm so unbelievably pumped up.
And it's because you have a complete reset of the landscape where there is incumbent advantage in distribution.
But that is it.
There's no other real advantage.
Well, there's a bunch of disadvantages.
Yes, and then there's a bunch of disadvantages.
But I just, yeah, go ahead, sorry.
No, no, no, but like, I mean, you know where I'm going.
So like.
So you have this.
You have the exact makings of a landscape where new startups can come in and do things that incumbents either can't or there's no obvious incumbent to even do that thing.
Because again you're taking maybe like services and turning them into AI labor, and there was no software incumbent previously to even attempt to do that.
And then you have incumbents that have a whole lot of complexity in terms of their ability to go and execute in some of these spaces.
And they're not going to retool their entire internal engineering workflows to move at 10x the pace.
And so a brand new startup can go and do that and then instantly get the scale of a larger company.
So it's the first time in history where you have none of the disadvantages of a big company.
And the traditional advantage you have as a big company is you have scaling of distribution.
Scale because you can look at a feature and you say, we're going to go build that next month.
And obviously it's harder because there's just lots of complexity to that.
But at least you have the human power to go do that.
Now, as a startup, you instantly have scale because you know background agents etc.
And so then it's a distribution game and a lot of these you know, pieces of software can go viral now in a way that wasn't possible 10 or 15 years ago.
So we've kind of neutralized a lot of the incumbent advantages and so thus it's a ripe opportunity for brand new startups.
Often will be people just like coming right out of college saying hey, it's my first time building a company like, like they're crazy enough to not know how hard it is, so they'll jump right into markets that otherwise we would assume are like it's art, like The market's already solved for it.
There's no way that you're going to build a company and you'll just have new startups that actually go and do it and actually produce real companies in these spaces.
Yeah, because I mean this is just so critical, because what's really happening is this is why you know it's an actual platform ship.
So Silicon Valley has seen this movie many times before, and that's why often there's a lot of this.
You know, is this crying wolf or not?
Because everybody knows that when there's a platform shift, that's the moment in time that startups are at an advantage.
And so each time there's a platform shift, everybody's like oh, this is it, this is going to reinvent everything.
And then it doesn't, and people get really like, oh, it's always incumbents.
But historically, the advantages to incumbents are wildly overestimated.
And really I mean, this is this one where you know, Like you know, did the internet undo Microsoft or not.
Undo Microsoft is a super interesting thing.
Because, of course, there's a $3 trillion company now.
But not on the internet in a way that you think about the internet.
Like none of the consumers, none of the platforms, none of the assets that we had in the 90s became internet assets.
I mean, even if you look at Azure today, it's an amazing accomplishment.
It's not running Windows anywhere.
Right, yeah.
And I think that that's why you know it's not crazy to go wow, is this going to be good or bad for Google?
Because there's a bunch of stuff that becomes really, really difficult if you don't make the transition.
And then it turns out historically, even if you do make the transition, you really didn't and you just have to wait for time to pass.
Well, and this is like Intel with the GPU.
Like, they missed the GPU in 2005.
Right.
And they missed the opportunity to buy the company, to do the work, or whatever.
And they kind of missed the data center, too.
It just took a longer time to figure out that they missed that as well.
Well, and we have a pretty narrow definition of like disruption in the sense of like we expect that the incumbent has to lose for this new startup.
And that never happens.
And never happens.
Yeah, yeah.
It's the whole radio, TV, you know, theater, you know, movie analogy.
It's like no, like it turns out that Microsoft can be a 4 trillion company and you can have all these new categories emerge that maybe Microsoft should have owned if everything was like perfectly analogous to the desktop days.
But they just don't.
And it all works together as one sort of ecosystem, because it just turns out like software did eat the world and these markets are actually just like 100 times larger than what we realized and so incumbents can grow and then you have new disruptors, that's of emerge along the way.
Yeah, anytime you bring in a new technology that brings the marginal cost down, then the market's going to expand and the incumbents can do it.
I will say incumbents are very bad when new user behaviors and buying behavior show up in particular.
They don't know how to cater to it.
AI is definitely a new user behavior and a new buying behavior.
And so this is very much an advantage of startups.
Just because to change a large company around a new user behavior cuts to the entire company, everything from basically marketing all the way to support in the back end.
That's just too much of a lift.
I mean, if you look at the best practices of even how you would create an agent, in the last 18 months I think we've gone through two to three architecture pattern changes.
And so I can barely keep up as a mid-sized company.
I can't imagine if you just had so many more people you had to organize around that.
Disruptive.
New technologies require people to understand how to use them in consumption in different ways.
That evolves over time as you get best practices.
That sort of flexibility can only come.
It's actually a very interesting question of how Microsoft actually did Copilot to begin with.
Because it was actually one of the first ones of these products was very successful.
I learned recently it was created by OpenAI, so that kind of explains it.
You had a startup person.
And conveniently, he was a startup guy.
Even then, by the way, it's remarkable that it came out of Microsoft.
What's always remarkable is when something new and defining comes from these big companies.
100 of the time you look into it and you're like it was basically Skunk Works.
Basically, they had nothing to lose.
And it didn't interfere.
Like, the iPod is this classic one.
Or the iPhone, even.
People always talk about like the brave.
It's like their computer business was dead dead, dead.
Like, it was like 3% share and going nowhere.
So, like, it was like a Hail Mary.
The iPod was a Hail Mary.
And then the phone, they weren't in the phone business.
It didn't matter.
And the fact that they made a phone a little computer is what Nokia was like, whoa, what is that?
And I think people really need to wrap their heads around just the fact that, to your point, that the big companies stay around a very, very long time.
And this is something I've seen a bunch of people in the past couple of weeks, you know.
Oh, it's basically Clay Christensen, Innovators Lemon.
Of course, nobody's ever read the book.
And Clay is a great guy.
You know, he was down the hall when I was teaching there.
And the thing is is that this is a book like about two and a half inch disk drives and a bunch of really crazy industries.
I really don't think it applies to modern systems.
But the thing that he missed aside from he missed the cost and low high-end and things, but was also that the companies don't evaporate.
And your point about that is really important.
And so what it does is there's this shadow that everybody is worried about.
And so you just see it constantly.
Well, like, I assume, like, when a company is like, we're not worried about if Google does this.
That's what you want to hear.
Right.
Because, like, it's just, I mean, and you've lived this.
Because you were like the classic, you know, Steve Jobs said you're a feature.
Mm-hmm.
And not just that.
There's like two whole companies that do this stuff now.
Fortunately, he only said that to Drew.
Right, right.
I got the Gates version of that one.
Yeah, yeah, yeah.
But the one thing that is very timeless about Clay would be the thing that does transcend floppies or whatever thing it is.
The incumbent doesn't want to do something that's against their business model.
Of course, of course.
And that part is fully timeless.
And to your point, it's actually very rare whenever we say these companies disrupted themselves.
It's almost never the case that they disrupted themselves.
It's the case that they went after a market that they had no actual market share in, and it just worked.
Which was the development tools for Microsoft.
Because there was no business in Microsoft development tools anymore, because nobody was writing Windows programs.
So really, it was like, what could we do for the cloud?
Honestly, it's even a little bit of a silly discussion to do this with AI.
AI is very disruptive, so we're like okay, well then, it allows startups to work against incumbents.
But even in non-disruptive technologies, startups often have a play against incumbents.
So, for example, every year I'm an infrastructure investor.
Every year for the last 10 years, after AWS reInvent, I move into therapist mode and all that.
All of my founders call me.
And they're launching my open source.
They're competing with it.
It happens every single time.
And I'm always like, you know what?
You'll be fine.
I can't think of one company AWS has ever put out a business by launching a service.
And you know what?
They've all been fine.
And so, in some ways, even in the normal world, like state of business, without massive disruption, startups still have.
Yeah, it was, don't build a SQL server and compete with Oracle directly.
Don't build a word processor.
Those are two things that were pretty evergreen for a long time.
Never do a spreadsheet.
There are some categories where people should call us first.
Because they won't change.
There's a far side.
We should definitely show people that.
It's a really, can I have 10 cents?
Why do you want 10 cents for your startup?
So I could buy a loaf of bread and beat you over the head with it for such a dumb idea.
It's a good one.
We should have a call-in on this thing, right?
Oh, yeah, yeah.
Here's my startup.
The other thing that I just don't think we've had at least I don't know of a modern kind of case study for is again this opening up of non-software, TAM for software.
So there's not even incumbents in the classic sense.
The incumbents are really just professional services categories of work.
And so it's really for the first time ever, you're packaging up intelligence for a particular domain and workflow.
And so there's no software company you're competing against for those dollars company right like yes, you know like you have to become an ag company, but then but, the vertical company will probably also be your customer company.
Yeah, they're also your customer.
So it's actually this amazing thing where where the people you're probably disrupting on paper are actually the primary users of your technology, actually take advantage.
Yes, And so then it's like there's really no inherent competition until eventually more companies flood that space to do that idea.
So this plays out in practice.
If you have an AI company that goes after say, agriculture or construction, they end up realizing the competitive asset are agriculture and construction.
The buyer knows how to price things like agriculture and construction.
They end up becoming basically agriculture and construction companies and then they end up doing exactly what you're saying is selling to the agriculture.
Because they're not good at that, right?
There's a whole world.
In fact, the earliest PC software was extremely vertical.
Like, if you actually look at the TRS-80 catalog from the early 1980s or late 1970s, it would be like this is crop rotation software.
Literally. like, okay, this is what you should do.
Like and you and the salesperson for Tandy would show up in Nebraska and sell crop rotation software.
And then there was like, I run a dentist's office and this is scheduling for a dentist's office.
And what happened was and this is what I think is going to happen is these professional services organizations are.
There are going to be some that are like computer savvy.
Yeah.
And today they're really good at using existing.
They're just going to go, you know what?
We should just build like a company.
Yes.
Yep.
Huge numbers of these verticals are just going to be existing pro-serve that turn into software providers.
I think it's an incredible time where let's just pretend that you just are committed to not building a software company.
Like you don't want to do that or you don't maybe have the team to do that.
So you're going to build like a real world company.
It's an incredible time if you started from scratch with now AI as your foundation.
If you wanted to be a new systems integrator and your whole point is that we are a systems integrator but we use cloud code or we use cognition or we use you know cursor to get the output, you will have such an advantage over any incumbent because the incumbent is not going to be able to rebuild that.
So I've seen examples of people building new ad agencies.
Oh, yeah.
Because obviously like, if you can do literally a million dollar ad, you know video, you know campaign for, you know 5000.
Somewhere between those two numbers, you can charge the customer.
So there's like this incredible time where you can just be building all new kinds of companies from the ground up, leveraging the breakthroughs that we've now seen in AI.
That was an early internet thing that did really happen, particularly in the advertising space.
And I think it's going to happen in everything that's text, which was there?
Were these digital native ad agencies that just knew how to use Flash.
And they would get bought for a billion dollars.
The same thing happened with social, by the way.
Yeah, exactly.
What did you think of this survey that was how many people use AI every week?
I thought this was pretty interesting.
Yes, what was your Pew response?
Well, it turns out like the number of people it's like up to 75 of adults are using it many times per week.
And of course, you just see it when you use Google search.
It's all self-reported, so you don't really know.
But it was pretty interesting because I pulled a 1999 Pew study on internet usage.
As you would.
As I would.
Like, basically, in 1999, like, half the country owned computers.
Yeah.
And they were all online.
Yeah.
Even four years post-Netscape, it was, like, still half the country.
Yeah.
And so you look at that as being slow or being fast.
Well, it's fast back then.
I mean, that felt fast, actually.
Well, you had to spend $3,000.
Activation Energy was buying a computer.
But still, you know, if you discount the search AI, there's still, you know, activation energy to figure out what this new thing is.
Nobody knows how to ask questions to a blank edit control, like make me smart about something.
I think the I mean the universal adoption of this as a consumer technology and then bleeding into prosumer is it exceeds anything I've ever experienced.
And I think it is, it will just fundamentally change people's sort of daily patterns.
Like my sister, not in tech at all, a teacher, Like she was in town and she was like yeah, I was asking chat this question.
And, like, I had to, like, do a double take.
I was like, chat.
And it was like, oh, ChatGPT, that's what normal people call this.
And it's just like, it's completely pervasive as just a standard technology.
And so that, to me, is just like okay, we now have the conditions laid for the next phase, which is and we've now seen this for a couple decades which is consumer adoption.
Now goes first.
And then it gets basically pulled into the enterprise, because you go to work and you're like why can't I ask questions of my enterprise systems the way that I can everything else in the world?
And why am I not getting that same level of productivity gain?
And then you're going to have the kids coming out of college that, like they only know how to do homework with ChachiBT.
They come into the workforce and they're like why would I spend two weeks writing this report when I just came out writing essays in an hour?
Like, obviously something has to kind of give on this.
So this just lays the foundation for why we're going to see just a massive upgrade cycle in the enterprise.
Also to build on your point earlier about distribution, and the reason that distribution is not the advantage that it used to be is because it already exists on 7 billion phones.
And so at every other platform shift there was an upside to getting new distribution that didn't exist before.
But you had to overcome that.
Like, you had to get, like, the internet to people who didn't have the internet before.
You had to get SaaS to people who didn't.
Now everybody has all of the ingredients right now.
Any time your business strategy relies on Comcast showing up in a neighborhood for me to get distribution like that, you're going to have some problems.
So this is a very different trend.
Honestly.
Last quick point on this is like for the first time in a very long time we're seeing brand effects with an early technology.
And what I mean by that is, if you look at like whatever the major model providers, you know how much better are they from each other.
Like, you know, maybe, you know, a little bit, maybe not.
Like, you know, it changes all the time.
But you actually see clear leaders if they break out early, just because people learn like the house online, people learn mid-journey, people know open AI et cetera.
People will just adopt you.
But don't discount, you know, like, the early leaders of search were, like, Excite and Yahoo.
And so there's, like, this is just to not discourage people from thinking.
Right, like, it's so early that names that you never heard of existed before Google.
And that's going to be really important.
It's never the first people.
When we look at mobile, there were big companies built, you know, like Uber and WhatsApp and Instagram and TikTok, but the biggest beneficiaries were Facebook and Google.
In AI.
Do we think it will be different that the biggest companies in the world in 10-20 years from now will be created after ChatGPT, or will it be similar that?
What's your time frame? you know, post-2019, I don't know.
No, no, no.
How many, 10 to 20 years you said?
Oh, yeah, sure.
Okay, so we can't know if we're wrong and really do this podcast in 10 years.
I think this is so boring, but I think it's going to look like what we saw in something like SaaS or Cloud, which is the incumbents get bigger.
But then there's all of these new categories that we would not have been able to predict.
And then there's lots of 10 and 20 and 50 and $100 billion companies that also emerge.
And then over time, those will just continue to scale.
So similar to mobile?
And some don't make the transition.
Yeah.
Some will go down on a relative basis because their market wasn't as ripe for agentic kind of workflows.
But I think that you can kind of say you know, maybe this is then for another conversation is just like if you have a current system of record that has a set of workflows on it where agents make sense to make that workflow much more powerful, that's a good position to be in.
But I bet you that if we look back in 10 or 20 years from now, the vast majority of things agents do don't relate to just those things that we are currently looking at, because there's just so many more fields that are now open.
And so all of those use cases, I would favor the disruptor or the insurgent.
And then in today's spaces I would kind of favor the incumbent on the margin, but the markets are so large that you're going to see kind of growth in all of them.
There's a key attribute across all of those, which is sort of like thought leadership or like who is really setting the agenda for what people are talking about.
And I think that's the thing that really changes.
The incumbents become bigger, but nobody wakes up in the morning wondering what they're up to.
Right.
Nobody starts to wonder, well, if they're going to do it, we need to understand it.
And that's the shift.
And you can think of it in the enterprise space or the business space.
Like, what do the CIOs, who do they wake up thinking about?
And that was a huge shift that sort of goes under the radar.
And in the consumer space, it just, like, it becomes the, I understand, I use ChatGPT at school.
I need ChatGPT.
And there's nothing you could do about it as a company.
Yeah.
I think the more provocative question is, are there any laggards that will use this to get ahead?
And we've seen this in the past, right?
Like, will Cisco do something interesting?
Oracle is making some kind of crazy moves.
Like.
Are we going to see those that like missed like social?
No, I think everybody missed Oracle as an example, right?
From three years to today.
Like, you would not have been like, oh, definitely the next billion dollar company.
Microsoft had its moment of like you know, you didn't know the future, and then it used cloud and Azure to kind of come back.
And so you know this actually is an opportunity for laggards that are behind the curve to come back.
Yeah.
To the Cisco point, like, data centers are sexy.
Like it turns out that we're just going to be building out lots of AI factories everywhere, so you're going to get more scale from parts of the stack that we stopped paying.
Like, Broadcom.
Like, again, people were not... Hawk?
Yeah.
Everybody looked at Jensen.
It may be somebody else, so... We'll table the rest for the next conversation.
Thank you so much for coming on.
Thanks for listening to the A16Z podcast.
If you enjoyed the episode, let us know by leaving a review at ratethispodcast.com slash A16Z.
We've got more great conversations coming your way.
See you next time.
As a reminder, the content here is for informational purposes only.
Should not be taken as legal business tax or investment advice or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any A16Z fund.
Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast.
For more details, including a link to our investments, please see a16zcom forward slash disclosures.
