Learn. Share. Evolve…

0:00:03 People are still trying to figure out how everything works.
0:00:09 What I love about the vibe writing concept actually is it’s a place in which full autonomy can be fulfilled today.
0:00:15 You’re prompting, although it’s English-like, it turns out you’re just programming.
0:00:17 And you’re just programming in prompt.
0:00:22 I’ve had so many conversations with product managers over the last two years about the death of product management.
0:00:25 It’s the end of the field, why we need PMs.
0:00:29 It was extreme in 1990, and it’s extreme today.
0:00:33 Where are we really in the AI computing shift?
0:00:37 Is this the Windows 3.1 moment or more like the 64K IBM PC?
0:00:46 In this episode, part of our This Week in Consumer series, I’m joined by A16Z general partner Anisha Charya and board member Stephen Sanofsky,
0:00:50 former Microsoft president and one of the most influential product thinkers in tech,
0:00:54 to unpack where we are in the AI platform cycle and what’s coming next.
0:00:59 We dig into the framework shaping this moment, partial autonomy, jagged intelligence,
0:01:02 Vibe coding versus vibe writing, what builders are wrong about agents,
0:01:05 what Google’s I.O. signals about platform strategy,
0:01:09 and why the future might be less about killer apps and more about control sliders.
0:01:14 We begin by discussing this week’s talk from Andre Karpathy on why software is changing again.
0:01:16 Let’s get into it.
0:01:21 As a reminder, the content here is for informational purposes only.
0:01:24 Should not be taken as legal, business, tax, or investment advice,
0:01:27 or be used to evaluate any investment or security,
0:01:31 and is not directed at any investors or potential investors in any A16Z fund.
0:01:37 Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast.
0:01:40 For more details, including a link to our investments,
0:01:44 please see A16Z.com forward slash disclosures.
0:01:54 Anish, Steven, we were having such a good conversation offline that I wanted to get this on the podcast.
0:01:55 There are a few topics we wanted to discuss.
0:02:00 First, we were all fascinated by Andre Karpathy’s talk at Startup School.
0:02:02 Steven, what did you find so interesting?
0:02:04 What were your takeaways or reactions from it?
0:02:07 Well, I totally loved the talk.
0:02:12 He did an unbelievable, like a philosopher king version of where we are.
0:02:16 And I just found his metaphors really compelling.
0:02:21 In fact, what I might do is even take it further back and just say,
0:02:24 since he used an analogy of like where we are in computing,
0:02:27 I’m talking about the Windows 3 era and stuff like that.
0:02:35 And having lived through all of them, I tend to think we’re at the 64K IBM PC era of the microcomputer.
0:02:39 And the reason I think that is actually a technical one,
0:02:44 which is that we’re at the point where people are still trying to figure out how everything works.
0:02:54 And all the coding and all of the energy is working around like these very basic working problems.
0:02:57 Like with the PC, it was like, okay, we have 64K of memory.
0:02:58 And our programs are all too big.
0:03:00 And we have no display and all these problems.
0:03:03 And with AI, people are like, it’s going to replace search.
0:03:05 It’s going to replace Excel.
0:03:06 And it’s going to replace all these things.
0:03:08 But it doesn’t add very well.
0:03:10 It gives you a lot of errors.
0:03:14 Like the thing that you say it’s going to do, it just doesn’t even do yet.
0:03:19 So I feel like we’re at a point that is just so, so early.
0:03:24 And he did a fantastic job of sort of making that arc.
0:03:26 You know, the thing that struck me the most was
0:03:29 he talked a lot about our relationship with this new tool.
0:03:32 You know, and in a sense, we want to use it in the same way that we’ve used
0:03:35 all the other computing tools and technologies we’ve used in the past.
0:03:40 But he really talked about this kind of inversion of the relationship of LLMs as people,
0:03:43 spirits, the fact that they have jagged intelligence.
0:03:46 So to me, that sort of meta point he made was one of the most interesting.
0:03:51 We have to relearn how to use this type of tool before we know how to be productive with it.
0:03:55 I think tools is a super interesting point because the talk is anchored in tools,
0:03:57 but the world itself is anchored in tools.
0:04:01 And the early stages of a platform are always about tools.
0:04:04 And so you kind of get a little confused.
0:04:07 Like right now, of course, he was talking about vibe coding, clearly,
0:04:10 because he pioneered the term, invented the concept, and is living it.
0:04:14 And it’s very interesting because I actually think coding is one domain
0:04:17 that always works best early in a platform because, well,
0:04:20 all the customers of the platform are developers,
0:04:23 and they’re going to make their tooling kind of work and come along.
0:04:26 But I really think that the most interesting thing for me,
0:04:30 what’s being underestimated in the near term is sort of vibe writing.
0:04:33 I mean, it seems weird to say anything with AI is underestimated
0:04:36 because Lord knows that’s not where we are.
0:04:40 But the thing is, is that vibe writing is so here.
0:04:44 Like if you’re in college, you’re already vibe writing.
0:04:48 And businesses are still working through the, well, can we use this?
0:04:49 Doesn’t seem appropriate.
0:04:52 And that’s a thing I’ve definitely lived through with word processors.
0:04:55 You know, I had to get permission from the dean in college
0:04:56 to use a computer to write papers.
0:05:00 But this vibe writing is absolutely a thing.
0:05:04 And it is really, really no different than when calculators showed up
0:05:08 and all of a sudden just doing math homework involved using a calculator.
0:05:11 And people are like, well, you’re not going to know how to do math in the future.
0:05:13 And it’s like, I won’t have to know how to do math.
0:05:15 That’s like the whole point of a tool.
0:05:20 Like I have a power drill, so I do not know how to use like one of those Amish drill things,
0:05:22 you know, and the world moves up the stack.
0:05:24 And so that’s where we are.
0:05:25 And it’s just super exciting.
0:05:28 What I love about the vibe writing concept actually is
0:05:31 it’s a place in which full autonomy can be fulfilled today.
0:05:35 So you can ask the model to vibe write something, you know,
0:05:38 really detailed and compelling, and it’ll do a great job.
0:05:41 Whereas with vibe coding, I think there’s a ton of constraints
0:05:44 as to what the model can actually do versus what it can conceptually do.
0:05:47 And understanding those boundaries and constraints
0:05:50 is going to define a lot of the text-to-code stuff for the next two years.
0:05:52 Well, I’d push back a little bit on that
0:05:54 because, of course, I agree on the coding side.
0:05:58 And I think one of the things developers do early in a platform is they love to tell you
0:06:03 that they’re doing something every day and it’s working, but it actually just isn’t.
0:06:05 And that’s just what happens early in a platform.
0:06:09 They tell you all these things that they say are easy and they’re actually not.
0:06:12 And they spent 18 hours struggling with something that didn’t work.
0:06:17 But on the vibe writing side, it also hits a point that I just think is so, so important,
0:06:21 which is, yeah, you can prompt it to spew out a bunch of stuff.
0:06:26 But if you have a job and your salary depends on you submitting that,
0:06:28 or you’re a student and your grade depends on you submitting that,
0:06:30 it actually better be right.
0:06:34 And you can’t just say, look, vibe wrote this and here you go.
0:06:37 And I think people don’t get confused when it comes to like math.
0:06:40 Like everybody knows you have to go check to the math if you ask it to do a table
0:06:42 and then add a column that does math.
0:06:49 But we’re going to just see endless, endless human wasn’t in the loop vibrating things.
0:06:53 And it’s just that with programs, you can’t really see that right away
0:06:56 because in order to actually distribute it or get someone to use it,
0:06:58 you have to at least fix the initial bugs.
0:07:02 We’ll only see them later when there are security bugs, authentication bugs,
0:07:05 passwords stored in plain text, or a zillion other problems
0:07:07 that are going to happen from vibe coding.
0:07:08 In a sense, we’ve seen this already, right?
0:07:12 We saw a bunch of lawsuits that were citing case precedent from cases that don’t exist.
0:07:16 So maybe this is actually the operative point, which is there’s full autonomy,
0:07:17 there’s partial autonomy.
0:07:21 I mean, maybe partial autonomy in writing is moving us from writer to editor,
0:07:22 but you still have to be the editor.
0:07:24 Yeah, we should also give him credit.
0:07:28 Many people have talked about this, but he did a fantastic job using the Iron Man analogy
0:07:34 of how we’re going to have autonomy, partial autonomy, and a slider to control what you want.
0:07:38 I actually think that’s a fantastic analogy and a way of thinking
0:07:40 that gives you a very clear picture from the movies.
0:07:46 But at the same time, people are very, very aggressive on their timeline of agents.
0:07:50 And there’s a very, very long history in trying to automate things
0:07:54 that turn out to be very, very difficult to automate.
0:07:56 And he did a fantastic job.
0:07:58 He said, people are talking about like the year of agents.
0:08:01 Yeah, that’s a good consultant phrase.
0:08:03 Just like he said, we’re in the decade of agents,
0:08:10 and it’s going to take a decade for things to be anywhere near living up to agentification as a meme.
0:08:12 It’s an interesting point.
0:08:14 I think a lot about agents as applied to financial services.
0:08:19 And I think there’s a set of problems in financial services that are high friction, low judgment.
0:08:23 So for example, when I want to go refinance my personal loan,
0:08:26 I don’t really feel attached to any specific brand of a personal loan provider.
0:08:28 I just want the cheapest rate.
0:08:30 So it’s actually a very low judgment decision.
0:08:34 But going and researching and applying for a personal loan is a high friction process.
0:08:36 That’s something I would love to delegate to an agent.
0:08:38 I think it can do a nice job.
0:08:42 Whereas doing my taxes, wow, like, Stephen, how much risk do you want to take on your taxes?
0:08:44 How many things do you want to report or not report?
0:08:46 That requires an enormous amount of judgment.
0:08:48 And of course, it also is high friction.
0:08:51 So when I think of the two by two of where is automation going to come first,
0:08:53 I think a lot about high friction, low judgment.
0:08:57 I want to build on that because I actually think it’s super important to also consider
0:09:01 that for anyone to offer the alternatives to the market,
0:09:05 there has to be an ability to differentiate, to explain.
0:09:08 And so you end up with this kind of thing where I just want the cheapest flight.
0:09:13 And of course, for 20 years, all of the flight searches and stuff has worked on the cheapest.
0:09:16 But it turns out that’s not actually what you want.
0:09:16 That’s right.
0:09:21 Plus, a lot of people want to intervene in presenting your choices to you.
0:09:21 Yes.
0:09:27 And so this idea that all choice in life is going to be reduced to some headless API.
0:09:27 Right.
0:09:28 I don’t understand.
0:09:31 People have to go build that and make a living building those things.
0:09:36 So to your example of refinancing a home, like the only reason that it can exist as a search
0:09:42 problem today is because the different people who want to refinance you can target you with an ad
0:09:46 and attract you as a customer and differentiate themselves on that offering.
0:09:51 And if you can’t do that, then your ability to actually automate that task isn’t going to exist
0:09:54 because there’s no economic incentive to just be,
0:09:59 hi, I’m the headless, faceless, nameless, low-priced mortgage leader.
0:09:59 That’s right.
0:10:00 It’s not really a business.
0:10:02 There’s nothing there.
0:10:05 Just headless, faceless, nameless food isn’t a thing.
0:10:07 It doesn’t show up in a white can labeled food.
0:10:10 And then you consume it and you’re, okay, all good.
0:10:11 I have food now.
0:10:13 Well, maybe Soylent, but yes.
0:10:16 In the future, in the dystopian future of Repo Man, that’s where we end up.
0:10:17 But that’s not going to happen.
0:10:20 I want the cheapest flight as long as it’s not on Spirit Airlines.
0:10:20 Right.
0:10:23 I want the cheapest flight, but I’m traveling with a family of three.
0:10:24 I don’t want to leave at 5 a.m.
0:10:25 No red eye.
0:10:25 Yeah.
0:10:27 Like, I want miles on this airline.
0:10:31 A lot of things don’t add up to that.
0:10:32 This is a real thing in business.
0:10:35 It’s a thing on the producer and the consumer side.
0:10:39 Consumers really, really want much more choice than they often think they do.
0:10:44 And anyone who’s bought anything on Amazon knows they complain about the choice, but they
0:10:48 really don’t want just, like, phone case to show up as the thing because it was $6.
0:10:54 I think this is a real through line through the talk, which is partial autonomy, jagged
0:10:55 intelligence.
0:11:00 Karpathy is just talking a ton about the constraints of the technology, which I think is the right
0:11:02 thing for us to be thinking through trade-offs around as builders.
0:11:06 And he does a great job, very much as this philosopher that I love.
0:11:08 His delivery, his tone.
0:11:09 Don’t just go read the summaries.
0:11:11 Don’t read a post.
0:11:13 Go just watch the video immersively.
0:11:18 Well, I want to get to automation and employment, particularly on the entry-level side.
0:11:23 But first, I just want to ask the broader question of, there was this idea of AI plus human,
0:11:26 I think it was chess, could beat AI for some period of time.
0:11:28 And that was kind of the co-pilot view of the world.
0:11:30 Human plus AI will have a better product.
0:11:33 And then it turned out, I think it was chess, maybe it was Goic, or maybe it was both, that
0:11:36 actually that was a temporary thing and AI is just better.
0:11:42 And then there’s the question as to how much of the world is like chess, where a human plus
0:11:47 AI is only better for a certain period of time and then the models get better, or how much
0:11:48 the world is like something else.
0:11:51 We’re always going to want that human plus AI is just better, or we’re just always going
0:11:52 to want humans to do it.
0:11:56 Look, my view is that in a domain in which you have a formal definition of correctness,
0:12:00 the path will be no autonomy, partial autonomy, full autonomy.
0:12:05 In domains where you don’t have a formal definition of correctness or where a ton of human judgment
0:12:10 is necessary and human choice and sort of a human direction, we’re just the right product
0:12:12 design is not to go all the way to full autonomy.
0:12:15 I would argue that chess and Go do have a formal definition for correctness.
0:12:18 So it makes sense that those were fully automated over time.
0:12:23 We’re back to the early stages of where things are, which means that a bunch of programmers
0:12:26 are sort of defining what success looks like.
0:12:31 And programmers are very good at either works or it doesn’t work, or I just want to automate
0:12:36 this, or I’m going to reduce your job to a tiny shell script kind of mentality.
0:12:40 And I just look at the world as everything is gray.
0:12:45 And everything is much harder than it looks when you don’t actually have to do it.
0:12:51 Ages and ages ago, I visited a really giant hospital in Minnesota to help them figure out
0:12:54 how to use Excel within the medical profession.
0:12:58 And the doctor just looked at me and he’s like, I don’t think you understand.
0:13:02 He was like, my job is all uncertain.
0:13:05 Every aspect of what I do is uncertain.
0:13:12 So adding something that pretends to be certain, like a spreadsheet, to my uncertainty doesn’t
0:13:13 actually help me.
0:13:18 And so fast forward, first, I’ve spent 25 years with a doctor, but that’s a different, but
0:13:21 if there was a story this week about radiologists.
0:13:26 And so very early, actually, if you go to ImageNet, everybody was immediately radiology is doomed.
0:13:29 Oh, like you never need to get a skin cancer biopsy.
0:13:32 You’ll just take pictures of your mole and it will just tell you.
0:13:34 And then you find out, wow, there’s judgment there.
0:13:38 And there’s even judgment in doing the biop, in how to do the biopsy, and then what to biopsy,
0:13:39 and all this.
0:13:43 But it turns out the radiologists have, like, fully embraced AI.
0:13:50 But they embraced it no different than they embraced the latest MRI technology or the latest
0:13:53 software update from GE for a CAT scan.
0:13:59 Like, I just think there are so many things like that, and so many jobs are either very,
0:14:05 very uncertain, or most of the job is basically exception handling.
0:14:05 Right.
0:14:08 And, like, people are like, oh, we’re going to automate our taxes.
0:14:13 Okay, taxes are literally a giant cascading if and switch statements of exceptions.
0:14:14 Yes.
0:14:20 And so the idea that you will just automate that, well, you have to know the answer to
0:14:20 all the exceptions.
0:14:25 And if you’re going to prompt it with the answer to all the exceptions, then you’re doing your
0:14:26 taxes manually.
0:14:30 It’s sort of like once you reach a certain income, you have to get help from an accountant
0:14:30 to do your taxes.
0:14:34 And the first thing the accountant does is ask you for your tax planner.
0:14:38 And as a software person, I look at it, I’m like, the tax planner really, really looks like
0:14:42 the input fields of the software you’re using.
0:14:46 So maybe I could just buy that software and then type it in.
0:14:50 And I said that, and he’s like, well, you’re welcome to, but you will go to jail.
0:14:55 And he explains, because every time I give him a number is a whole decision about where
0:14:56 to apply it.
0:14:57 Does it work?
0:15:00 And I’m like, well, you’re not really a farmer, so don’t fill anything in on that form and stuff
0:15:01 like that.
0:15:03 Automation is extremely difficult.
0:15:08 And it’s exception bound, it’s judgment bound, and it’s all uncertain.
0:15:12 You know, a field in which this question comes up a ton is product management.
0:15:16 I’ve had so many conversations with product managers over the last two years about the
0:15:17 death of product management.
0:15:20 It’s the end of the field, why we need PMs.
0:15:25 And I think our sort of developer generation has developed a real resentment towards product
0:15:26 managers, which is a different conversation.
0:15:31 With that said, I think that the product management job is the job of addressing ambiguity.
0:15:34 And it’s ambiguity that prevents progress from being made.
0:15:37 Sometimes it’s execution, decision making, product design.
0:15:39 That will not change.
0:15:44 The nature of business and human interaction and companies is these complex adaptive systems
0:15:46 where there will always be ambiguity.
0:15:49 I think you’ll always need judgment, and you’ll always need somebody who looks like a product
0:15:50 manager.
0:15:54 Yeah, I think that really gets to the vibe coding challenge we’re dealing with, which
0:15:57 is like, how fast can we go text to app?
0:16:03 And I think here, what’s so interesting in the long arc of platform transitions is that we’re
0:16:08 also having this platform transition happen not just out on the open.
0:16:09 We’ve had that before.
0:16:13 Like back when, in the earliest days of computing, these platform transitions happened in user group
0:16:17 meetings, like at the Cumberly Community Center down the street, or in magazines or newsletters,
0:16:20 and then with news groups, then the internet and so on.
0:16:24 The whole internet was all ICQ, and it was all in the open.
0:16:27 But now it’s like happening on CNN, on the nightly news.
0:16:33 Everyone knows about the platform transition that’s happening, in particular on social,
0:16:34 in Discord.
0:16:38 And so what’s happening is you’re getting a lot of like vibe coding for clout.
0:16:44 And so you’re getting a lot of this, I had an idea, I prompted it, and it worked.
0:16:45 And here I am.
0:16:47 At some point, I just go, I’m calling BS on that.
0:16:48 That’s like not a thing.
0:16:50 And then I sound like an old person.
0:16:54 And because some people think I am, I don’t, but some people think I am.
0:16:54 I don’t either.
0:16:57 It looks like, hey, you’re just being old.
0:16:57 Yes.
0:17:02 But then you dig in and you find out like, wow, you’re prompting.
0:17:06 Although it’s English-like, it turns out you’re just programming.
0:17:06 Yes.
0:17:08 And you’re just programming in prompts.
0:17:08 Yes.
0:17:12 And people are like, oh, this is what we’re going to do is we’re just going to get
0:17:15 the model to require a little bit more structure.
0:17:17 And I’m like, you’re writing a new programming language.
0:17:18 Yes.
0:17:22 And this path of text to app and vibe coding is just developing a new language,
0:17:24 which is super cool.
0:17:27 Lord knows the world is built on programming languages.
0:17:32 In the 80s, if you drove slowly past the computer science department trying to get a PhD,
0:17:35 they would just invent a new programming language right then and there
0:17:37 if you stood outside the building for too short a time.
0:17:42 But we can’t lose sight of the fact that the arc of programming has been one of basically
0:17:44 over-promise and under-deliver.
0:17:49 When I was in college, like, the theory was the market was going to need so many programmers
0:17:53 that the whole employment force, the whole workforce would just be software people.
0:17:55 And that never happened.
0:17:57 And now here we are, we’re not going to need any.
0:17:58 They’re all just going to go away.
0:18:02 And I think it was extreme in 1990 and it’s extreme today.
0:18:07 And I think that the big thing is this over-promising at each transition.
0:18:09 Even just most recently, low-code.
0:18:11 Who even says that word anymore?
0:18:13 Like, we’re not allowed to even mention it.
0:18:17 And it’s because it’s always the same thing, which is, yes, if all you’re doing
0:18:21 is a very straightforward app that looks like all the other straightforward apps,
0:18:24 but with a domain spin or a branding logo or something.
0:18:28 We see this with Wix and with website templates, like it’s possible.
0:18:31 But you’re not going to run a company on any of those.
0:18:33 I totally agree.
0:18:37 Where I disagree, actually, is I think that the language, the language model in this case,
0:18:42 but the language in your metaphor, is improving at a dramatic rate underneath these things.
0:18:46 So while I think almost all these products today, they’re good at prototyping,
0:18:50 they’re trying to push into refinement, they’re not really usable as things that you can actually
0:18:52 deploy to production at all.
0:18:55 In fact, most of the cool demos you see on Twitter don’t work three days later.
0:18:59 So they’re very much in the prototyping phase, but the programming language in the metaphor
0:19:01 is improving dramatically.
0:19:06 I think we’ll get there, at least make more progress than we think, versus a traditional
0:19:10 programming language like object-oriented, it didn’t feel like it 100x the number of programmers
0:19:13 or 1 100th the amount of time to ship something to production.
0:19:17 We just got new tools and sort of new problems to solve.
0:19:19 Well, of course, you’re benefiting from hindsight.
0:19:20 Yes.
0:19:21 And that’s a key thing.
0:19:22 First, I agree.
0:19:26 We’re in an exponential improvement cycle with the models.
0:19:26 Yes.
0:19:29 So any predictive power goes out the window.
0:19:29 Correct.
0:19:33 And anyone who says like something negative, you’re going to be the next person who says
0:19:36 the internet is going to be a faxing fad like a fax machine.
0:19:39 And that’s a bad, you just don’t want to be there.
0:19:43 And it turns out also, that’s a case where having lived through them, you get very shy
0:19:48 about making predictions because you see how foolish people look for a long time.
0:19:50 But take something like object-oriented.
0:19:53 I mean, this thing was hyped to the moon.
0:19:55 This is a wave of programming languages.
0:19:58 Just to give you an idea, again, how the speed things move.
0:20:00 They started in 1980.
0:20:05 And by 1990, they finally reached like peak hype.
0:20:05 Right.
0:20:07 So it was like 10 years of incremental improvement.
0:20:10 And by then, they were also over.
0:20:15 Like any programmer would have kind of said, eh, it’s sort of just changing the old programming
0:20:18 paradigms of abstraction and polymorphism and stuff like that.
0:20:23 But meanwhile, the magazines, which was the key measure of success at the time, there was
0:20:27 one magazine that had a picture of a baby of diapers, not a picture, a drawing, on the
0:20:28 cover of programming.
0:20:30 How programming will get made easy.
0:20:35 I remember seeing it at the newsstand, and I was working on the C++ compiler at the time.
0:20:40 C++ was a brand new language in 1990, and it didn’t work yet.
0:20:46 And here was a baby who was going to make programming possible for other babies at baby care or something.
0:20:52 And whether it was that or all the database programming languages like Delphi or PowerBuilder,
0:20:56 in algorithmic sense, they were all constant improvement.
0:21:01 Like just, they added a constant factor, like plus seven, onto programming.
0:21:04 None of them changed the mathematical order of magnitude.
0:21:09 And what I believe is, with writing right now, it’s changing order of magnitude.
0:21:11 And so it’s here, it’s happening.
0:21:13 Accuracy isn’t there.
0:21:18 But one of the things about writing is, like, actually, when you read it, most of it in business
0:21:19 is not really accurate already.
0:21:21 It’s very much like autocorrect.
0:21:27 Like autocorrect fixed all the common typos, like T-E-H to T-H-E in English, and just replaced
0:21:32 them with these wild new autocorrects that just replaced what you typed with a word that
0:21:36 has no meaning in the context of the sentence, which is what we face on phones all the time.
0:21:41 So what we’re going to see is a whole different set of errors in business writing or academic
0:21:46 writing in schools that just replace other errors that have always creeped in.
0:21:46 Totally.
0:21:49 I remember Smalltalk was the hot language, right?
0:21:51 Well, Smalltalk was the start of it, and it was called Smalltalk 80.
0:21:52 Yes.
0:21:56 And then it really didn’t ever achieve any momentum outside of Palo Alto.
0:21:57 Yes.
0:21:59 But then C++ came along.
0:21:59 Yes.
0:22:02 And there were 50 languages in the middle that people don’t talk about.
0:22:02 Yes.
0:22:06 Like Objective-C being one of them, that was the iPhone language, which was really one Steve
0:22:06 Jobs.
0:22:11 There was Object Pascal and Pascal, Pascal with a relational database attached.
0:22:13 But this was my master’s degree.
0:22:14 Then I quit grad school.
0:22:17 I could go on about this one for far too long, so I’ll just stop now.
0:22:22 Do you think there will be best-selling novels that are entirely AI-generated or nearly entirely
0:22:23 in the next few years?
0:22:24 A hundred percent.
0:22:28 I don’t think Stephen King is going to do that, but I think there’ll be some new writer who
0:22:29 will probably write it under a pseudonym.
0:22:34 And a year after the novel is written and has been made into a movie, they’ll say, oh, by
0:22:39 the way, I got the plot idea from a prompt, and then I just started having writing and I
0:22:40 was editing it along the way.
0:22:41 Absolutely.
0:22:45 The copyright suit that follows from training models and stuff, that’s a different issue.
0:22:48 I think there’s two things on this, actually, that are really interesting.
0:22:52 So one is these language models are these averaging machines.
0:22:56 And with art, you almost definitely don’t want the average of all the novels or all the writing
0:22:57 or all the authors.
0:22:59 You want something that’s at the edge.
0:23:04 So how do we actually point them in a direction such that they can be at the edge of culture,
0:23:06 which I think is important for making great art?
0:23:11 I think the other thing is a lot of the artists don’t yet know how to use any new tools, and
0:23:13 we’re going to see artists that are native in the technology.
0:23:17 Instead, what we’re seeing a lot out there, what’s called the slop, has just been a lot
0:23:22 of this low barrier to entry art that’s being created, which is great because it gives people
0:23:24 the sort of fulfillment of creative generation.
0:23:29 I think what we’re talking less about is, hey, how is the ceiling being raised for artists
0:23:30 because they have access to these technologies?
0:23:36 Without going all de champa on what is art, I mean, bad sitcoms are part of society too,
0:23:38 but I think it’s important.
0:23:44 We tend to focus on like the very, very best of things, but most everything isn’t only the
0:23:45 very best.
0:23:47 In business writing, it’s all slop.
0:23:48 Yeah.
0:23:53 I mean, this is why, look, I’ve written a lot of business writing, so I can say this confidently
0:23:56 about what I’ve written and what gets written.
0:24:01 But take something completely mundane that a lot of people in Silicon Valley spend a lot
0:24:04 of time working on, like the enterprise software case study.
0:24:12 I’m telling you, GPT generates a better enterprise case studies faster than the typical marketing
0:24:15 associate does at a company in like one millionth effort.
0:24:16 Yes.
0:24:18 And does the content need to exist?
0:24:19 It actually does.
0:24:21 It’s just an important part of the selling process.
0:24:22 Yes.
0:24:26 And so at the extreme, like with something like medical diagnosis, we tend to think about
0:24:32 the most obscure diseases, the most difficult to understand problems, with the finest hospitals,
0:24:33 with the most resources.
0:24:39 But you have to remember, like 80% of the world has no access to anything.
0:24:39 Right.
0:24:40 Yes.
0:24:48 So wherever you think of medical LLM is, as in the slop scale, most people don’t have access
0:24:50 access to anything average.
0:24:55 So we have to just make sure that the whole debate does not center around, like, what is
0:24:58 Francis Ford Coppola using as the book?
0:24:59 And who are the actors?
0:24:59 Yes.
0:25:01 And who is the cinematographer?
0:25:06 Because that corporate case study, well, they often go and interview the person and film it.
0:25:09 Well, like, all of a sudden, we see it today.
0:25:10 Those things are done over Zoom.
0:25:11 Yes.
0:25:17 So suddenly, flying in or getting a satellite and booking, we’ve changed our view of excellent
0:25:19 because we wanted more access.
0:25:22 And I think that’s absolutely going to happen.
0:25:24 Should you get graded on slop in school?
0:25:25 That’s a different problem.
0:25:27 But most stuff is pretty average.
0:25:29 The world needs more slop, says Stephen.
0:25:33 I feel like this is an oppressed interview where you can put words in my mouth like that.
0:25:33 Yeah.
0:25:34 The world needs more slop.
0:25:35 That’ll be the title.
0:25:37 Well, so actually, Mark makes this point.
0:25:40 I think it’s a really good one, you know, which is, is the bar for success perfection?
0:25:43 Is the bar for success what people can do today?
0:25:46 Or is the bar for success just something that’s better than the alternative?
0:25:50 And in your case of 80% of the world that has access to no medical knowledge,
0:25:54 no medical services, no medical opinion, of course, this is dramatically better.
0:25:54 Yeah.
0:25:58 I look at it like when I had to get permission to use a word processor in college,
0:26:05 one of the stumbling blocks was that my printer was like an Epson MX80 dot matrix printer.
0:26:09 And it looked like a printer, like a computer printer,
0:26:13 which the rules for the papers were they had to be written on a typewriter.
0:26:13 Yes.
0:26:17 And then the Macintosh came out in the spring and only had an image writer,
0:26:19 which is another dot matrix printer.
0:26:25 So all of a sudden the standard changed because the value of being able to revise and edit and
0:26:32 update and copy paste and use fonts was just so much higher than the fidelity of the teacher
0:26:35 reading it on bond paper with courier.
0:26:37 And that’s going to happen with content as well.
0:26:41 What I would love to talk to you about, Stephen, is actually just hearing your take on I.O.
0:26:44 And if you felt like a Google I.O.
0:26:46 Oh, yeah, yeah.
0:26:51 So essentially there was a lot of conversation around Google and how Google had sort of fallen
0:26:54 behind and lost their ability to make new things.
0:26:57 They released a ton of new software at every part of the stack in I.O.
0:27:00 What do you think that says for Google about Google?
0:27:02 Do you think the sort of demise of Google is overstated?
0:27:06 Well, of course, I think the demise of Google is an absurd proposition.
0:27:11 The demise of a giant company is a crazy thing to say.
0:27:17 Driving in, I was listening on CNBC, some investor or whatever talking their book, talking about IBM
0:27:19 is the one to buy.
0:27:24 I almost wanted to pull over to the side of the road and think, what universe am I in
0:27:28 where this company that has died like nine times in my career?
0:27:28 Right.
0:27:30 And so death of is just such a done thing.
0:27:34 Losing a position of influence, however, is a very real thing.
0:27:40 In these platform transitions, big companies have an enormous asset, which is the shock and
0:27:41 awe asset.
0:27:46 And so they have the ability to tell the story called we’re pivoting our whole company around
0:27:49 this and we’re a zillion dollar in whole company.
0:27:59 And here is like a full assault across the board for every single asset we have and every
0:28:04 single category the world is talking about that matters.
0:28:06 And that’s what you could do.
0:28:11 Someone was asking me on Twitter yesterday about this event Microsoft held in 2000 called
0:28:12 Forum 2000.
0:28:18 And it was when we announced like a whole bunch of internet stuff and the early cloud stuff.
0:28:21 It wasn’t called cloud, but early cloud stuff.
0:28:24 And nobody in that room understood what we were talking about.
0:28:25 Not a person.
0:28:32 But they all left like, oh my God, there is so much stuff here, which is a repeat of five
0:28:35 years earlier when we did what was called Internet Strategy Day.
0:28:39 And like the headlines were literally sleeping giant awoke.
0:28:46 And so it was totally predictable that Google would show up with like literally the B2 bombers
0:28:48 of software.
0:28:52 But the question is really much deeper than that.
0:28:59 And it’s really, will they alter their context of how they build products and their go-to-market?
0:29:03 Because that’s really what undermines the big technology companies.
0:29:09 And so with Microsoft, the interesting thing was all those products that got announced over
0:29:13 that five-year span or 10-year span, none of them are around today.
0:29:15 I should be very careful every time I say something like this, I get assaulted.
0:29:22 But like the big announcement at Forum 2000 was the .NET framework in C Sharp, which by
0:29:26 almost any measure, one would call a legacy platform today.
0:29:30 So it like came and went in six or seven years.
0:29:36 And everything was about virtual machines and clustering and all this stuff that VMware was
0:29:36 doing.
0:29:38 And that’s not where anything was.
0:29:42 And on top of all that, the economic model became SaaS.
0:29:49 And so what I’m looking at with Google is not, can they present all the technologies in the
0:29:56 context of Google search and ads, but can they transform the way they think to something
0:29:56 new?
0:29:59 Because that’s really where the disruption is going to happen.
0:30:00 I love that point.
0:30:00 Cool.
0:30:01 Awesome.
0:30:03 Anish, Stephen, thanks so much for this weekend.
0:30:03 Sure thing.
0:30:04 Thank you.
0:30:04 Super fun.
0:30:09 Thanks for listening to the A16Z podcast.
0:30:14 If you enjoyed the episode, let us know by leaving a review at ratethispodcast.com slash
0:30:15 A16Z.
0:30:18 We’ve got more great conversations coming your way.
0:30:19 See you next time.