WEBVTT - Why the Tech World Is Going Crazy for Claude Code

0:00:02.720 --> 0:00:14.000
<v Speaker 1>Bloomberg Audio Studios, Podcasts, radio News.

0:00:18.360 --> 0:00:21.840
<v Speaker 2>Hello and welcome to another episode of the Odd Lots podcast.

0:00:21.880 --> 0:00:25.400
<v Speaker 2>I'm Joe Wassental and I'm Tracy Alloway. So, Tracy, you're

0:00:25.400 --> 0:00:28.880
<v Speaker 2>cool like if I, like, you know, just start doing

0:00:28.880 --> 0:00:32.000
<v Speaker 2>this part time as I like build out my software business, right, Like,

0:00:32.040 --> 0:00:32.720
<v Speaker 2>you're cool about that.

0:00:32.960 --> 0:00:36.440
<v Speaker 3>I was gonna say, I've been thinking about AI and productivity,

0:00:36.680 --> 0:00:40.720
<v Speaker 3>and so far your productivity has gone down, Joe, because

0:00:40.760 --> 0:00:44.080
<v Speaker 3>instead of doing Odd Lots things, you're coding your own software.

0:00:44.560 --> 0:00:47.639
<v Speaker 2>Except that I'm creating content for the Odd Lots newsletter

0:00:47.680 --> 0:00:53.120
<v Speaker 2>about coding, and that is productivity a creative debatable debatable,

0:00:53.159 --> 0:00:55.880
<v Speaker 2>But but you're cool with that. You're cool with like

0:00:55.960 --> 0:00:57.800
<v Speaker 2>me like, oh, I'm just gonna like check in part

0:00:57.800 --> 0:00:59.480
<v Speaker 2>time on Odd Lots when we have a recording.

0:00:59.520 --> 0:01:01.680
<v Speaker 3>Well, so of course not okay, good, of course not.

0:01:01.840 --> 0:01:03.600
<v Speaker 2>Yeah, that's the right answer. I want you to yes,

0:01:03.680 --> 0:01:05.640
<v Speaker 2>I want you to be really sad. But like a

0:01:05.680 --> 0:01:08.360
<v Speaker 2>few other people, you know, I have like caught the

0:01:08.400 --> 0:01:10.959
<v Speaker 2>sort of like bug of like AI coding and I'm

0:01:11.000 --> 0:01:13.680
<v Speaker 2>totally blown away. I've like played with it from the

0:01:13.720 --> 0:01:16.040
<v Speaker 2>big I started playing around with it last year, but

0:01:16.080 --> 0:01:18.200
<v Speaker 2>then over the holidays. I've been writing about this in

0:01:18.240 --> 0:01:20.839
<v Speaker 2>the newsletter. Suddenly, like my Twitter feed is like clud code,

0:01:20.840 --> 0:01:24.320
<v Speaker 2>clug code, clud code. I do just cursor before, which

0:01:24.319 --> 0:01:27.560
<v Speaker 2>I was very impressed by at the time. And so

0:01:27.680 --> 0:01:29.800
<v Speaker 2>when I got home from vacation, one of the first

0:01:29.840 --> 0:01:31.559
<v Speaker 2>things I did is like figure out how to install

0:01:31.640 --> 0:01:34.920
<v Speaker 2>clug code on my computer. And I was like, oh,

0:01:35.080 --> 0:01:39.080
<v Speaker 2>I am like hooked. This is actually like I see

0:01:39.120 --> 0:01:41.200
<v Speaker 2>why I have my Twitter feed is just like people

0:01:41.200 --> 0:01:41.880
<v Speaker 2>posting about this.

0:01:42.160 --> 0:01:42.520
<v Speaker 4>All right.

0:01:42.560 --> 0:01:44.520
<v Speaker 3>So I have to say I have not tried it

0:01:44.760 --> 0:01:47.319
<v Speaker 3>because I only have a work computer and I can't

0:01:47.360 --> 0:01:51.960
<v Speaker 3>install new software, and I probably definitely cannot install new

0:01:52.000 --> 0:01:53.760
<v Speaker 3>software that then makes changes to.

0:01:53.720 --> 0:01:54.920
<v Speaker 2>Your existing software.

0:01:55.080 --> 0:01:57.680
<v Speaker 3>I don't think Bloomberg would like that, but I have

0:01:57.800 --> 0:02:00.760
<v Speaker 3>seen the hype. Lots of people talk talking about it.

0:02:01.200 --> 0:02:05.000
<v Speaker 3>Have you seen claud Cowork Have you heard.

0:02:04.920 --> 0:02:06.080
<v Speaker 2>Of Oh yeah, yeah, yeah yeah.

0:02:06.160 --> 0:02:08.680
<v Speaker 3>So one of the criticisms of claud code was that

0:02:08.800 --> 0:02:11.560
<v Speaker 3>you know, like, okay code, but you still need some

0:02:11.680 --> 0:02:14.880
<v Speaker 3>background knowledge in coding, because like, you know, the interface

0:02:14.960 --> 0:02:18.200
<v Speaker 3>is kind of like it pies and all of that

0:02:18.360 --> 0:02:22.800
<v Speaker 3>or nineteen nineties Cowork apparently like goes a step further

0:02:23.000 --> 0:02:27.600
<v Speaker 3>for normal people in quoting. It makes it super super easy.

0:02:27.720 --> 0:02:32.400
<v Speaker 3>And the funniest thing is that apparently cloud code actually

0:02:32.440 --> 0:02:33.720
<v Speaker 3>coded Cowork.

0:02:34.080 --> 0:02:37.399
<v Speaker 2>So this is like really relates to my experience last

0:02:37.480 --> 0:02:40.320
<v Speaker 2>year and then this year, which is that even last year,

0:02:40.440 --> 0:02:43.760
<v Speaker 2>like trying to use the AI coding tools, it was

0:02:43.800 --> 0:02:47.280
<v Speaker 2>an annoying process because there are various things that you

0:02:47.320 --> 0:02:49.280
<v Speaker 2>had to do in the actual command line of the

0:02:49.280 --> 0:02:52.320
<v Speaker 2>computer that were like I didn't I don't know command

0:02:52.400 --> 0:02:55.040
<v Speaker 2>line vernacular, and you have to install these libraries and

0:02:55.080 --> 0:02:56.880
<v Speaker 2>stuff like that. So there was this sort of like

0:02:57.440 --> 0:03:03.680
<v Speaker 2>barrier that existed. But what's really changed in the last

0:03:03.760 --> 0:03:06.400
<v Speaker 2>year or with the with Claude code, which has actually

0:03:06.400 --> 0:03:07.920
<v Speaker 2>been around for a while and I should have like

0:03:07.919 --> 0:03:11.080
<v Speaker 2>played with it before, is that like, because it sits

0:03:11.160 --> 0:03:14.480
<v Speaker 2>on your computer, it sort of takes away it de

0:03:14.560 --> 0:03:15.840
<v Speaker 2>abstracts it and so when you talk.

0:03:15.680 --> 0:03:17.880
<v Speaker 3>About like that it actually does the stuff.

0:03:17.560 --> 0:03:19.520
<v Speaker 2>It does it it just like oh, it's like, oh,

0:03:19.520 --> 0:03:22.280
<v Speaker 2>we're gonna need to install this open source natural language

0:03:22.320 --> 0:03:25.160
<v Speaker 2>processing library. It just does it automatically. Instead of me

0:03:25.240 --> 0:03:27.240
<v Speaker 2>trying like figure out like what are the right keystrokes

0:03:27.240 --> 0:03:29.480
<v Speaker 2>to pull that in or why is this not going

0:03:29.480 --> 0:03:32.640
<v Speaker 2>into the right file? Folder or whatever, and so like

0:03:32.639 --> 0:03:35.440
<v Speaker 2>what like Cowork, it's like all like all of these

0:03:35.480 --> 0:03:38.200
<v Speaker 2>sort of like little frictions, like these technical things like

0:03:38.400 --> 0:03:43.280
<v Speaker 2>command line user very rapidly are like dissipating and so

0:03:43.320 --> 0:03:45.120
<v Speaker 2>that like then you have something like Cowork where it's

0:03:45.200 --> 0:03:47.320
<v Speaker 2>just like they know they're taking care of that, and

0:03:47.440 --> 0:03:50.120
<v Speaker 2>so you get this like user interface that's just like

0:03:50.280 --> 0:03:53.960
<v Speaker 2>it's just getting easier and friendlier. There's almost no technical

0:03:54.000 --> 0:03:55.240
<v Speaker 2>frictions at all anymore.

0:03:55.280 --> 0:03:58.040
<v Speaker 3>Also, it feels very iterative, like the code is improving

0:03:58.120 --> 0:04:00.400
<v Speaker 3>upon itself at this point I think was one of

0:04:00.440 --> 0:04:01.680
<v Speaker 3>Claude's main selling points.

0:04:01.880 --> 0:04:04.720
<v Speaker 2>Well, this is like you've seen like people talk about like,

0:04:04.760 --> 0:04:07.720
<v Speaker 2>oh is AGI here? And this is like part of

0:04:07.760 --> 0:04:10.400
<v Speaker 2>the debate because the premp one of the ideas I

0:04:10.400 --> 0:04:12.800
<v Speaker 2>guess behind AGI is like, well what happens when you

0:04:12.840 --> 0:04:15.680
<v Speaker 2>have software that can train itself and so forth? And

0:04:15.760 --> 0:04:17.960
<v Speaker 2>I don't really know if I buy that, but you

0:04:18.040 --> 0:04:20.520
<v Speaker 2>do just see like how fast the iteration cycles are,

0:04:20.880 --> 0:04:23.719
<v Speaker 2>and I think we want to get into this. In part,

0:04:24.000 --> 0:04:26.480
<v Speaker 2>they're fast because a bunch of people are suddenly getting excited.

0:04:26.480 --> 0:04:29.200
<v Speaker 2>So then the human provides this sort of like we're

0:04:29.240 --> 0:04:31.400
<v Speaker 2>sowing the seeds of our own demise because we're so

0:04:31.560 --> 0:04:35.800
<v Speaker 2>enthusiastically participating in the evolution. But I just like it's

0:04:35.800 --> 0:04:38.640
<v Speaker 2>suddenly clear, like, oh, this is going to change I

0:04:38.680 --> 0:04:42.000
<v Speaker 2>think computing. And the other thing is the code works,

0:04:42.240 --> 0:04:44.840
<v Speaker 2>like it creates code that like this is like there's

0:04:44.839 --> 0:04:46.520
<v Speaker 2>no bugs. You know it works.

0:04:47.600 --> 0:04:50.360
<v Speaker 3>Did you see speaking of automating yourself, you see there

0:04:50.400 --> 0:04:52.320
<v Speaker 3>was a post on Reddit from a lawyer who said

0:04:52.360 --> 0:04:55.760
<v Speaker 3>he's basically used claud code to automate like his entire job,

0:04:55.800 --> 0:04:56.960
<v Speaker 3>and he hasn't told anyone.

0:04:57.440 --> 0:04:59.800
<v Speaker 2>I'm not exactly surprised because the other thing that I

0:05:00.000 --> 0:05:04.080
<v Speaker 2>experimented with is and I haven't one hundred percent verified this,

0:05:04.160 --> 0:05:07.120
<v Speaker 2>but on jobs Day last week, I downloaded the full

0:05:07.160 --> 0:05:09.960
<v Speaker 2>pdf and I just typed into the cloud code like

0:05:10.480 --> 0:05:13.560
<v Speaker 2>find the most interesting details and make some charts based

0:05:13.600 --> 0:05:15.440
<v Speaker 2>on and it did it in like a couple of minutes.

0:05:15.480 --> 0:05:17.400
<v Speaker 2>I have no like ability to like I've never like

0:05:17.440 --> 0:05:21.599
<v Speaker 2>built charts myself or whatever, like designing or whatever. And

0:05:21.680 --> 0:05:24.320
<v Speaker 2>I didn't totally confirm yet that the data was all correct,

0:05:24.320 --> 0:05:26.559
<v Speaker 2>but I'm pretty sure it was because everything I spot checked,

0:05:26.680 --> 0:05:28.440
<v Speaker 2>so I didn't just that crucial I didn't tell you

0:05:28.600 --> 0:05:30.200
<v Speaker 2>I know I didn't. That's why I didn't want to like, oh,

0:05:30.320 --> 0:05:32.960
<v Speaker 2>like here's what, here's the today's jobs reporting charts.

0:05:33.080 --> 0:05:33.400
<v Speaker 5>But I did.

0:05:34.200 --> 0:05:37.320
<v Speaker 3>So what application did it actually build it in the charts?

0:05:37.520 --> 0:05:39.200
<v Speaker 2>I don't know. I just had a file like that's

0:05:39.200 --> 0:05:41.120
<v Speaker 2>the thing. I had a file on my computer at

0:05:41.120 --> 0:05:43.920
<v Speaker 2>that point. What kind of file, like a P and

0:05:44.000 --> 0:05:46.479
<v Speaker 2>G file, like an image file, that's the crazy thing.

0:05:46.800 --> 0:05:48.880
<v Speaker 2>I don't know. And so there was just this image

0:05:48.920 --> 0:05:50.760
<v Speaker 2>that had a bunch of charts, and my spot checks

0:05:50.800 --> 0:05:54.360
<v Speaker 2>did suggest like I didn't see anything off. And people

0:05:54.440 --> 0:05:56.839
<v Speaker 2>get paid money to like build that kind of stuff

0:05:56.839 --> 0:05:58.360
<v Speaker 2>for like analysts and stuff like that.

0:05:58.480 --> 0:06:01.160
<v Speaker 3>And right, so this is the other big question. If

0:06:01.200 --> 0:06:05.120
<v Speaker 3>everyone can build their own software, what actually happens to software?

0:06:05.160 --> 0:06:07.640
<v Speaker 3>And I was reading something I forget who it was by,

0:06:07.960 --> 0:06:10.800
<v Speaker 3>but someone used claud code to create they wanted a

0:06:10.839 --> 0:06:13.719
<v Speaker 3>website that would basically make the money for doing nothing,

0:06:13.839 --> 0:06:15.880
<v Speaker 3>and that was the prompt And did.

0:06:15.720 --> 0:06:16.160
<v Speaker 2>They do it?

0:06:16.440 --> 0:06:16.760
<v Speaker 6>Yeah.

0:06:16.839 --> 0:06:20.240
<v Speaker 3>So the idea that the model came up with was

0:06:20.680 --> 0:06:23.920
<v Speaker 3>you can sell prompts, packages of good prompts and sell

0:06:23.960 --> 0:06:26.560
<v Speaker 3>them for like forty bucks and you'll make tons of money.

0:06:27.160 --> 0:06:29.680
<v Speaker 3>And I was thinking about that like, Okay, it's possible

0:06:29.760 --> 0:06:32.760
<v Speaker 3>to make money that way, but also why wouldn't I

0:06:32.839 --> 0:06:35.400
<v Speaker 3>just use claud code to do the same thing.

0:06:36.400 --> 0:06:38.800
<v Speaker 2>There are many big questions that we use an economy

0:06:38.800 --> 0:06:40.200
<v Speaker 2>are going to have to think about, and I think

0:06:40.279 --> 0:06:42.400
<v Speaker 2>my main takeaway is we're gonna have to think about

0:06:42.400 --> 0:06:44.560
<v Speaker 2>these sooner red than later. But what is clud code?

0:06:44.560 --> 0:06:46.640
<v Speaker 2>Why is everyone so hyped about it? Like what is

0:06:46.680 --> 0:06:50.240
<v Speaker 2>it about this particular piece of software versus what exists

0:06:50.240 --> 0:06:52.800
<v Speaker 2>from open Aye and Gemini and all this stuff, Like

0:06:52.800 --> 0:06:55.719
<v Speaker 2>why is this captured everyone's imagination? We really do have

0:06:55.720 --> 0:06:59.200
<v Speaker 2>the perfect guess because it's someone who, unlike me, has

0:06:59.240 --> 0:07:01.440
<v Speaker 2>been getting their hands and the stuff for longer. One

0:07:01.480 --> 0:07:03.800
<v Speaker 2>of the few people that I know who is into

0:07:04.000 --> 0:07:08.240
<v Speaker 2>lollms before chat GPT existed and was actually using them

0:07:08.320 --> 0:07:11.520
<v Speaker 2>via the API, and was actually talking about their technical

0:07:11.560 --> 0:07:14.880
<v Speaker 2>capacity to do things like coding even before November of

0:07:14.920 --> 0:07:17.640
<v Speaker 2>twenty twenty two. So truly the perfect guests We're going

0:07:17.720 --> 0:07:19.280
<v Speaker 2>to be speaking with Noah Bryer. He is the co

0:07:19.360 --> 0:07:22.720
<v Speaker 2>founder of Alefic, which is a consultancy that helps big

0:07:22.760 --> 0:07:25.920
<v Speaker 2>companies deal with AI stuff. So, uh, Noah, thank you

0:07:25.920 --> 0:07:27.160
<v Speaker 2>so much for coming on odd lots.

0:07:27.240 --> 0:07:28.200
<v Speaker 5>Thank you for having me.

0:07:28.680 --> 0:07:31.160
<v Speaker 2>What is eleven? What's the deal? How are you like

0:07:31.320 --> 0:07:34.960
<v Speaker 2>using llms before chat GPT existed? I don't know. I

0:07:35.000 --> 0:07:36.280
<v Speaker 2>know very few people who were doing that.

0:07:36.880 --> 0:07:39.679
<v Speaker 6>I had the good fortune of shutting down a startup

0:07:39.720 --> 0:07:41.200
<v Speaker 6>in twenty twenty two, and so I had a lot

0:07:41.240 --> 0:07:42.280
<v Speaker 6>of free time on my hands.

0:07:42.400 --> 0:07:43.720
<v Speaker 2>And then how are you using it though?

0:07:43.760 --> 0:07:43.840
<v Speaker 3>Like?

0:07:43.880 --> 0:07:45.720
<v Speaker 2>How did you look at your like how did you

0:07:45.880 --> 0:07:48.080
<v Speaker 2>aware that there was this thing that could be of

0:07:48.160 --> 0:07:49.000
<v Speaker 2>potential used to.

0:07:49.080 --> 0:07:51.000
<v Speaker 6>What was So my very first thing I was doing

0:07:51.120 --> 0:07:53.880
<v Speaker 6>was using gethub Copilot, which at the time was built

0:07:53.920 --> 0:07:57.240
<v Speaker 6>into VS code, and it was autocomplete inside VS code.

0:07:57.240 --> 0:07:59.800
<v Speaker 6>So it was a nice and pretty immediately realized that

0:07:59.800 --> 0:08:03.920
<v Speaker 6>there were certain coding tasks that it could just handle completely.

0:08:04.960 --> 0:08:08.040
<v Speaker 6>Anything that was very pattern based. So if you write code,

0:08:08.080 --> 0:08:10.120
<v Speaker 6>you write a lot of tests. If you write tests,

0:08:10.200 --> 0:08:12.880
<v Speaker 6>every test kind of follows the same pattern, and you

0:08:12.920 --> 0:08:14.520
<v Speaker 6>want it to follow the same pattern. You're looking for

0:08:14.560 --> 0:08:17.160
<v Speaker 6>that structure, and over time, because it was looking at

0:08:17.200 --> 0:08:19.200
<v Speaker 6>your code base, it was able to basically autocomplete it.

0:08:20.000 --> 0:08:24.240
<v Speaker 6>I also started playing with the GPT three API, which

0:08:24.240 --> 0:08:25.880
<v Speaker 6>had come out. I think that came out in November

0:08:25.920 --> 0:08:28.240
<v Speaker 6>of twenty twenty one, and that was the first time

0:08:28.280 --> 0:08:31.640
<v Speaker 6>it was publicly available to everybody, and they had a

0:08:31.720 --> 0:08:35.040
<v Speaker 6>large language, Boddel as we know it today, available to them.

0:08:35.080 --> 0:08:38.000
<v Speaker 6>So I was just testing and building things, and I

0:08:38.040 --> 0:08:41.840
<v Speaker 6>pretty immediately realized the very first thing I did where

0:08:42.000 --> 0:08:44.720
<v Speaker 6>it just blew my mind was I built a web scraper.

0:08:44.800 --> 0:08:47.240
<v Speaker 6>So I was just trying to pull pricing data from

0:08:47.240 --> 0:08:49.520
<v Speaker 6>a website. And I've done a lot of this in

0:08:49.559 --> 0:08:52.199
<v Speaker 6>my career. It's maybe the most annoying task you have

0:08:52.320 --> 0:08:54.360
<v Speaker 6>to do in all of coding, because HTML is the

0:08:54.360 --> 0:08:58.920
<v Speaker 6>most miserable language to have to parse. And I just

0:08:59.000 --> 0:09:00.600
<v Speaker 6>had this thing where I took the page, I took

0:09:00.600 --> 0:09:02.200
<v Speaker 6>the content, I took the text, and I gave it

0:09:02.240 --> 0:09:04.040
<v Speaker 6>to the AI and I asked it to give me

0:09:04.080 --> 0:09:06.600
<v Speaker 6>back the pricing table and give me back the pricing table.

0:09:06.920 --> 0:09:09.480
<v Speaker 6>And I just thought, I'll never do it the other

0:09:09.520 --> 0:09:09.880
<v Speaker 6>way again.

0:09:09.920 --> 0:09:10.320
<v Speaker 5>That's it.

0:09:10.400 --> 0:09:16.040
<v Speaker 3>Yeah, that HTML mentioned just brought up like memories of

0:09:16.080 --> 0:09:19.080
<v Speaker 3>me in like the mid nineties on HTML goodies. Do

0:09:19.080 --> 0:09:21.840
<v Speaker 3>you remember that side? Yeah? I wonder if it's still

0:09:22.200 --> 0:09:26.640
<v Speaker 3>Is it still up? That would be wild? Does cloud code?

0:09:26.679 --> 0:09:29.520
<v Speaker 3>Does that count as AGI? This seems to be the debate,

0:09:29.600 --> 0:09:30.560
<v Speaker 3>right is it AGI?

0:09:31.160 --> 0:09:33.720
<v Speaker 6>I try not to wait into what's AGI and what's not.

0:09:33.880 --> 0:09:36.360
<v Speaker 6>I think my guess on AGI, for what it's worth

0:09:36.480 --> 0:09:39.240
<v Speaker 6>is that it's probably going to be a conversation like

0:09:39.280 --> 0:09:41.640
<v Speaker 6>the Turing Test, where everybody thought it was really really

0:09:41.679 --> 0:09:43.640
<v Speaker 6>important for a really long time. We thought the Turing

0:09:43.640 --> 0:09:46.800
<v Speaker 6>Test was the biggest thing for seventy years or whatever,

0:09:46.880 --> 0:09:50.280
<v Speaker 6>and then CHATCHYBT very clearly passed the Turing test, and

0:09:50.280 --> 0:09:53.360
<v Speaker 6>now everybody pretends, like we it's not just that they forgot,

0:09:53.400 --> 0:09:56.920
<v Speaker 6>they pretend that it never mattered. Oh and so I

0:09:56.960 --> 0:09:59.440
<v Speaker 6>am kind of guessing that that's going to be what

0:09:59.480 --> 0:10:00.960
<v Speaker 6>the conversation. It's like, it's just going to be a

0:10:00.960 --> 0:10:04.120
<v Speaker 6>sort of forever moving goalpost, because it turns out that

0:10:04.160 --> 0:10:07.000
<v Speaker 6>the idea we had for what general intelligence looks like

0:10:07.720 --> 0:10:11.200
<v Speaker 6>is not quite that. But I also think, you know,

0:10:11.360 --> 0:10:14.600
<v Speaker 6>the computer scientists and the sort of serious AI researchers

0:10:14.800 --> 0:10:17.600
<v Speaker 6>would say that much of what's going on inside claud

0:10:17.640 --> 0:10:20.720
<v Speaker 6>code is not the model itself. It's the model paired

0:10:20.760 --> 0:10:23.720
<v Speaker 6>with a human and I think that is a pretty

0:10:23.760 --> 0:10:24.480
<v Speaker 6>important distinction.

0:10:24.559 --> 0:10:25.760
<v Speaker 5>But I don't know about AGI.

0:10:41.600 --> 0:10:46.640
<v Speaker 2>Well, okay, so you were using GPT to code prior

0:10:46.720 --> 0:10:49.920
<v Speaker 2>to the release of jag GUPT. So therefore, coding models

0:10:49.920 --> 0:10:53.120
<v Speaker 2>have been around a long time. So what is for

0:10:53.200 --> 0:10:56.120
<v Speaker 2>those who haven't played around with it? What is claude code?

0:10:56.160 --> 0:10:58.760
<v Speaker 2>Because again, coding models have been around for a long time.

0:11:00.040 --> 0:11:02.599
<v Speaker 2>Well maybe have heard of cursor, copilot or some of

0:11:02.679 --> 0:11:05.760
<v Speaker 2>these other harnesses, et cetera. What is cloud code?

0:11:05.840 --> 0:11:08.920
<v Speaker 5>So if we back up first and we go to copilot.

0:11:09.160 --> 0:11:12.560
<v Speaker 6>So Copilot was the first sort of commercial application of

0:11:12.559 --> 0:11:17.120
<v Speaker 6>a largelanguage model by most accounts, And what Copilot did

0:11:17.160 --> 0:11:19.360
<v Speaker 6>in its initial instantiation was just.

0:11:19.440 --> 0:11:20.600
<v Speaker 2>Auto Microsoft product.

0:11:20.640 --> 0:11:23.400
<v Speaker 6>It's a Microsoft product, so Microsoft doones, get hub, getthub,

0:11:23.520 --> 0:11:28.120
<v Speaker 6>develop copilot. It was Microsoft had the partnership with open Ai,

0:11:28.280 --> 0:11:30.120
<v Speaker 6>and so they built it in and what it was

0:11:30.160 --> 0:11:33.280
<v Speaker 6>doing was doing autocomplete. So if you're writing code, a

0:11:33.320 --> 0:11:35.679
<v Speaker 6>lot of writing code is boiler plate or trying to

0:11:35.760 --> 0:11:38.320
<v Speaker 6>remember the name of a function. And you know, the

0:11:38.400 --> 0:11:41.400
<v Speaker 6>reason stack overflow existed was because you can never remember

0:11:41.480 --> 0:11:45.000
<v Speaker 6>the exact name of that function or the exact rejects

0:11:45.080 --> 0:11:47.320
<v Speaker 6>that you need to use in order to find and

0:11:47.360 --> 0:11:50.520
<v Speaker 6>replace something, and so you would go search for it.

0:11:50.559 --> 0:11:53.160
<v Speaker 6>And they realized that you could just build that into

0:11:53.200 --> 0:11:56.959
<v Speaker 6>the ID your code editor and and have it autocomplete

0:11:57.000 --> 0:12:02.640
<v Speaker 6>for you, and it was pretty amazing. Then, uh, Chatchipt

0:12:02.800 --> 0:12:04.640
<v Speaker 6>came out, and even before that, I had built a

0:12:04.679 --> 0:12:07.199
<v Speaker 6>simple chatbot for myself because I realized that, hey, I

0:12:07.200 --> 0:12:09.360
<v Speaker 6>could just ask this and instead of going and searching

0:12:09.400 --> 0:12:13.559
<v Speaker 6>stack overflow. It was totally capable of answering code questions,

0:12:13.559 --> 0:12:15.800
<v Speaker 6>and it was capable of writing rejects or doing these

0:12:15.800 --> 0:12:18.480
<v Speaker 6>things and didn't make mistakes. Yes, but like there's famous

0:12:18.520 --> 0:12:22.120
<v Speaker 6>mistakes on stack overflow of incorrect rejects that now exists

0:12:22.120 --> 0:12:25.520
<v Speaker 6>in every code base in the world, and so you know,

0:12:25.559 --> 0:12:26.920
<v Speaker 6>I think there were a lot of us just kind

0:12:26.920 --> 0:12:29.480
<v Speaker 6>of playing with these things and realizing they were a

0:12:29.559 --> 0:12:33.000
<v Speaker 6>huge boon. And so I think, really the next step

0:12:33.040 --> 0:12:35.080
<v Speaker 6>is Cursor comes out. And the thing Curser realized that

0:12:35.240 --> 0:12:37.719
<v Speaker 6>co Pilot didn't was that it wasn't good enough to

0:12:37.720 --> 0:12:40.160
<v Speaker 6>have autocomplete. You also needed the Q and A because

0:12:40.160 --> 0:12:42.640
<v Speaker 6>you have these things that you can't just auto complete.

0:12:42.640 --> 0:12:43.880
<v Speaker 6>You want to be able to ask the question and

0:12:43.920 --> 0:12:46.079
<v Speaker 6>answer it. And then Chatchipt came out and everybody was

0:12:46.080 --> 0:12:50.280
<v Speaker 6>switching between ID and then I think, really the next

0:12:50.280 --> 0:12:52.840
<v Speaker 6>big piece is that claud code came out. And what

0:12:52.920 --> 0:12:56.040
<v Speaker 6>claud Code did that was so remarkable was they took

0:12:56.080 --> 0:12:59.079
<v Speaker 6>the same set of models really and they took them

0:12:59.080 --> 0:13:01.600
<v Speaker 6>out of the chatbot and they really just gave it

0:13:01.679 --> 0:13:06.880
<v Speaker 6>some very basic functionality to operate within your machine.

0:13:06.960 --> 0:13:07.160
<v Speaker 5>Right.

0:13:07.200 --> 0:13:08.760
<v Speaker 6>And so you know, if you really look at kind

0:13:08.760 --> 0:13:11.079
<v Speaker 6>of what exists within claud code, you're calling out to

0:13:11.120 --> 0:13:13.760
<v Speaker 6>a model, and they gave it capability around sort of

0:13:13.880 --> 0:13:16.000
<v Speaker 6>two big things. One is you can read and write

0:13:16.000 --> 0:13:19.320
<v Speaker 6>files on your computer, and then two is that you

0:13:19.400 --> 0:13:24.040
<v Speaker 6>can operate Unix the base commands, the bashed commands that

0:13:24.200 --> 0:13:29.240
<v Speaker 6>exist in your environment. And again, because these models were

0:13:29.280 --> 0:13:31.600
<v Speaker 6>trained on the Internet, and there's no greater source of

0:13:31.600 --> 0:13:33.840
<v Speaker 6>information on the Internet than how to make the Internet,

0:13:34.080 --> 0:13:38.040
<v Speaker 6>they know how to use Unix commands incredibly well, right,

0:13:38.040 --> 0:13:40.680
<v Speaker 6>because Unix has existed for whatever it is, sixty years,

0:13:40.679 --> 0:13:43.120
<v Speaker 6>and the way these commands were designed, they're all designed

0:13:43.120 --> 0:13:45.920
<v Speaker 6>to be very very simple. There's a find command, and

0:13:46.000 --> 0:13:47.640
<v Speaker 6>you know, there's some thing called re rep and it

0:13:47.679 --> 0:13:50.520
<v Speaker 6>can search to a code base. And Unix has this

0:13:50.600 --> 0:13:53.240
<v Speaker 6>sort of beautiful way of tying one command to another,

0:13:53.280 --> 0:13:55.120
<v Speaker 6>so you can take the output of one command and

0:13:55.120 --> 0:13:58.120
<v Speaker 6>send it to another. And they kind of just gave

0:13:58.200 --> 0:14:00.839
<v Speaker 6>the model access to these two or three very simple things,

0:14:01.400 --> 0:14:03.800
<v Speaker 6>and it kind of turned out that it unlocked a

0:14:03.840 --> 0:14:06.720
<v Speaker 6>whole bunch of functionality. I don't think even the people

0:14:06.800 --> 0:14:10.320
<v Speaker 6>who built it fully realized. Like one example that I

0:14:10.320 --> 0:14:13.199
<v Speaker 6>think about a lot is just the challenge you have

0:14:13.360 --> 0:14:15.040
<v Speaker 6>with all of these.

0:14:14.880 --> 0:14:16.880
<v Speaker 5>AI models is that they're stateless.

0:14:17.200 --> 0:14:20.920
<v Speaker 6>So every time you talk to chat GPT, it's sending

0:14:20.960 --> 0:14:23.960
<v Speaker 6>your entire conversation history back to chat GPT because it

0:14:23.960 --> 0:14:28.600
<v Speaker 6>has no saved history of that chat, right, And that's fine,

0:14:28.640 --> 0:14:30.560
<v Speaker 6>it's the way it works. It's just fact. But it

0:14:30.680 --> 0:14:32.960
<v Speaker 6>means that you know, it forgets things. It doesn't no

0:14:33.040 --> 0:14:37.160
<v Speaker 6>conversation to conversation. And one very easy way to save

0:14:37.280 --> 0:14:41.080
<v Speaker 6>your state is just write it to a file. And

0:14:41.160 --> 0:14:44.560
<v Speaker 6>so you give it right access and it can create files.

0:14:44.560 --> 0:14:47.440
<v Speaker 6>And now all of a sudden you've overcome this, like

0:14:47.680 --> 0:14:51.520
<v Speaker 6>probably the single biggest challenge that exists inside these large

0:14:51.560 --> 0:14:54.600
<v Speaker 6>language models, which is that they're fundamentally stateless.

0:14:55.000 --> 0:14:58.400
<v Speaker 3>So Claude writes itself little like memory notes right to

0:14:58.920 --> 0:15:01.960
<v Speaker 3>remember the entire ca context of the conversation, and that's

0:15:01.960 --> 0:15:03.200
<v Speaker 3>how it solved that problem.

0:15:03.480 --> 0:15:05.760
<v Speaker 6>No, so there's sort of two things going on in

0:15:05.840 --> 0:15:06.560
<v Speaker 6>claude code.

0:15:06.600 --> 0:15:07.280
<v Speaker 5>Beneath the hood.

0:15:07.520 --> 0:15:10.080
<v Speaker 6>There's one thing that works exactly like chetchipt or any

0:15:10.080 --> 0:15:13.240
<v Speaker 6>of these other ones, which is it's maintaining a conversation history.

0:15:13.280 --> 0:15:18.840
<v Speaker 6>So every message you send it and every action it takes,

0:15:18.880 --> 0:15:22.200
<v Speaker 6>it's recording to a log, which is just one big file.

0:15:22.400 --> 0:15:25.240
<v Speaker 6>That's really no different than what chet ChiPT can do.

0:15:25.520 --> 0:15:28.160
<v Speaker 6>Where it gets really interesting though, is it can also

0:15:28.440 --> 0:15:32.600
<v Speaker 6>write files that it can then read, so whereas that

0:15:32.640 --> 0:15:36.280
<v Speaker 6>conversation history is all saved off, and eventually that conversation

0:15:36.320 --> 0:15:37.600
<v Speaker 6>gets too long and it needs to do a thing

0:15:37.640 --> 0:15:41.120
<v Speaker 6>called compaction. And when it compacts it, it tries to

0:15:41.160 --> 0:15:44.560
<v Speaker 6>sort of just remember the bits because there the total.

0:15:44.320 --> 0:15:44.880
<v Speaker 5>Window is.

0:15:47.440 --> 0:15:50.160
<v Speaker 6>Large. But I mean it's like one hundred thousand tokens

0:15:50.840 --> 0:15:51.160
<v Speaker 6>I mean.

0:15:51.040 --> 0:15:54.320
<v Speaker 3>By memory notes, right. It compacts the information into the

0:15:54.400 --> 0:15:56.400
<v Speaker 3>important stuff, then retreat.

0:15:56.520 --> 0:15:56.960
<v Speaker 5>It does that.

0:15:57.720 --> 0:15:59.800
<v Speaker 6>It only does that at the end, like once it

0:15:59.840 --> 0:16:02.680
<v Speaker 6>runs out of space, once it runs out of context window.

0:16:02.880 --> 0:16:04.600
<v Speaker 6>So it has two hundred thousand tokens I think, and

0:16:04.600 --> 0:16:06.840
<v Speaker 6>two hundred thousand tokens in rough terms is probably one

0:16:06.880 --> 0:16:10.920
<v Speaker 6>hundred and fifty thousand words. It says, okay, it's time

0:16:10.960 --> 0:16:13.600
<v Speaker 6>for me to compact all of this stuff, and so

0:16:13.640 --> 0:16:16.280
<v Speaker 6>it still saves your whole history.

0:16:15.960 --> 0:16:16.600
<v Speaker 5>On your computer.

0:16:16.640 --> 0:16:19.320
<v Speaker 6>You still have the entire message, but for that session,

0:16:19.720 --> 0:16:23.280
<v Speaker 6>it just compacts it down to this, you know, maybe

0:16:23.360 --> 0:16:27.920
<v Speaker 6>twenty five thousand token memory of what it was.

0:16:29.400 --> 0:16:32.240
<v Speaker 2>And is this like something that was not obvious before

0:16:32.360 --> 0:16:36.400
<v Speaker 2>as a solution like this compaction? How important is it

0:16:36.920 --> 0:16:40.320
<v Speaker 2>for this being? Like okay? As a human I can

0:16:40.320 --> 0:16:42.600
<v Speaker 2>work on this on a project for a long time,

0:16:42.680 --> 0:16:44.920
<v Speaker 2>like how much of an an unlock was there?

0:16:45.120 --> 0:16:47.640
<v Speaker 6>I'm not sure compaction was the okay walk. I think

0:16:47.680 --> 0:16:52.280
<v Speaker 6>the compaction functionality is helpful. The way chat GPT does it,

0:16:52.320 --> 0:16:54.320
<v Speaker 6>for what it's worth is they don't do compaction. They

0:16:54.360 --> 0:16:57.600
<v Speaker 6>just forget your messages eventually. So if you're in one chat,

0:16:58.000 --> 0:17:01.000
<v Speaker 6>eventually your oldest message is going to fall off the back.

0:17:02.200 --> 0:17:06.680
<v Speaker 6>For coding, that's probably less helpful, but there are trade offs.

0:17:06.720 --> 0:17:10.399
<v Speaker 6>I both techniques work. I think fundamentally, the thing that

0:17:10.520 --> 0:17:14.440
<v Speaker 6>is special about cloud code is not the compaction. It's

0:17:14.680 --> 0:17:17.199
<v Speaker 6>the ability to write and read files on your computer,

0:17:17.560 --> 0:17:19.520
<v Speaker 6>which means you can always write off memories.

0:17:19.960 --> 0:17:21.600
<v Speaker 2>And then what does that mean? Write off memory?

0:17:21.760 --> 0:17:24.359
<v Speaker 6>So you could say, hey, it's really important that I

0:17:24.400 --> 0:17:26.800
<v Speaker 6>remember this thing for future sessions. I want to always

0:17:26.880 --> 0:17:30.040
<v Speaker 6>work this way. So in a code base of mine,

0:17:30.720 --> 0:17:34.119
<v Speaker 6>I have a set of documentation that explains how I

0:17:34.200 --> 0:17:37.200
<v Speaker 6>like to do things. And cloud code makes a mistake,

0:17:37.240 --> 0:17:39.320
<v Speaker 6>and so the next time I can write a memory,

0:17:39.680 --> 0:17:42.159
<v Speaker 6>essentially it's written as a thing they call a skill,

0:17:43.000 --> 0:17:44.640
<v Speaker 6>and you can write it off and you say, hey,

0:17:44.760 --> 0:17:46.840
<v Speaker 6>whenever you run into this, I want you to operate

0:17:46.880 --> 0:17:50.800
<v Speaker 6>in this kind of way. And that existing across every

0:17:50.840 --> 0:17:53.960
<v Speaker 6>session is really a thing you can only do when

0:17:53.960 --> 0:17:56.600
<v Speaker 6>you can store it as a file. It's a thing

0:17:56.640 --> 0:17:58.760
<v Speaker 6>you can't do in quite the same way when you're

0:17:58.800 --> 0:18:00.840
<v Speaker 6>operating in this environment or it's just going back and

0:18:00.880 --> 0:18:03.840
<v Speaker 6>forth to the API. So this access to the file

0:18:03.920 --> 0:18:06.000
<v Speaker 6>system is one really big piece. And then the second

0:18:06.040 --> 0:18:09.640
<v Speaker 6>is is just the Unix commands, I mean computers. Every

0:18:09.640 --> 0:18:13.359
<v Speaker 6>computer program lives on top of these sort of baseline

0:18:14.400 --> 0:18:18.720
<v Speaker 6>functions and the way that the designers of Unix built

0:18:18.760 --> 0:18:22.600
<v Speaker 6>them is really elegant, and they're very small. They all

0:18:22.640 --> 0:18:27.480
<v Speaker 6>do one thing, and they're all composable, and in coding terms,

0:18:27.480 --> 0:18:30.359
<v Speaker 6>composable means they can be chained together, right, and so

0:18:31.240 --> 0:18:34.920
<v Speaker 6>you can say, hey, look for files that mentioned this word,

0:18:35.119 --> 0:18:37.560
<v Speaker 6>and then from those files, I want you to take

0:18:37.600 --> 0:18:40.280
<v Speaker 6>this second action, and then from the output of that action,

0:18:40.320 --> 0:18:41.400
<v Speaker 6>I want you to take a third action.

0:18:41.440 --> 0:18:43.119
<v Speaker 5>And that's just built into Unix.

0:18:43.160 --> 0:18:45.760
<v Speaker 6>You literally just put a little pipe in between and

0:18:45.840 --> 0:18:48.320
<v Speaker 6>you just pipe them from one to another and that's it.

0:18:48.400 --> 0:18:51.199
<v Speaker 6>And so you give it access to write these commands,

0:18:51.200 --> 0:18:52.960
<v Speaker 6>and all of a sudden, it gets these sort of

0:18:52.960 --> 0:18:55.600
<v Speaker 6>second and third order effects that are just incredibly powerful

0:18:55.600 --> 0:18:57.720
<v Speaker 6>and built over a really long time.

0:18:58.320 --> 0:19:01.159
<v Speaker 3>So how much of claud code, the way it's different

0:19:01.200 --> 0:19:04.400
<v Speaker 3>to other models, how much of that was overcoming technological

0:19:04.480 --> 0:19:08.760
<v Speaker 3>challenges versus just having a good idea, because hearing you

0:19:08.840 --> 0:19:12.280
<v Speaker 3>describe it, I mean, giving access to a computer seems

0:19:12.359 --> 0:19:15.760
<v Speaker 3>like kind of obvious, like let's just do that.

0:19:17.640 --> 0:19:19.480
<v Speaker 6>I don't have a good answer to that. I think

0:19:19.560 --> 0:19:22.000
<v Speaker 6>that it was kind of just a good idea. Yeah,

0:19:22.200 --> 0:19:25.040
<v Speaker 6>I think they did some patterns really well. They're clearly

0:19:25.119 --> 0:19:30.119
<v Speaker 6>incredibly talented, not just engineers, but kind of thinkers about

0:19:30.119 --> 0:19:33.680
<v Speaker 6>how to structure it. Like the primitives inside cloud code

0:19:33.720 --> 0:19:36.280
<v Speaker 6>are just smart. And then the thing that they've done

0:19:36.640 --> 0:19:41.880
<v Speaker 6>and Boris Cherney, who's the lead developer on cloud coded anthropic,

0:19:41.920 --> 0:19:44.920
<v Speaker 6>he talks about light and demand a lot right, and

0:19:45.000 --> 0:19:47.280
<v Speaker 6>Layton Demand is basically just, hey, look at the ways

0:19:47.320 --> 0:19:49.520
<v Speaker 6>people are using these systems and then figure out ways

0:19:49.520 --> 0:19:52.080
<v Speaker 6>to make that a part of the product itself. I

0:19:52.119 --> 0:19:54.600
<v Speaker 6>think what they've done brilliantly, and this is kind of

0:19:54.640 --> 0:19:56.800
<v Speaker 6>easy when you have a community of developers who are

0:19:56.840 --> 0:19:58.440
<v Speaker 6>nerds who want to go talk about all the ways

0:19:58.480 --> 0:20:01.399
<v Speaker 6>that they're using these things. Is they have I am

0:20:01.480 --> 0:20:03.679
<v Speaker 6>amazed at the speed in which you know, I have

0:20:03.720 --> 0:20:07.199
<v Speaker 6>a small community of fifteen CTOs who all use this

0:20:07.240 --> 0:20:10.600
<v Speaker 6>stuff religiously. And you know, when we first started that community,

0:20:10.720 --> 0:20:12.879
<v Speaker 6>it took them a month to I would see it

0:20:12.920 --> 0:20:14.480
<v Speaker 6>in the chat, and then a month later it would

0:20:14.480 --> 0:20:17.399
<v Speaker 6>get built into cloud code, and then increasingly.

0:20:16.920 --> 0:20:18.959
<v Speaker 5>It's like a day later. It feels like they're just

0:20:19.160 --> 0:20:19.960
<v Speaker 5>they're just listening to it.

0:20:20.000 --> 0:20:22.040
<v Speaker 6>But I think they're just not only tapped in, but

0:20:22.040 --> 0:20:26.080
<v Speaker 6>they're really fundamentally, you know, they're dog fooding it. They

0:20:26.200 --> 0:20:29.080
<v Speaker 6>use their own products when you you know, they talk

0:20:29.119 --> 0:20:33.480
<v Speaker 6>about the productivity engineering productivity at Anthropic. You know, despite

0:20:33.560 --> 0:20:36.600
<v Speaker 6>growing at a crazy clip, it continues to go up.

0:20:36.640 --> 0:20:40.639
<v Speaker 6>And you know, anybody who's built had to manage large

0:20:40.680 --> 0:20:44.720
<v Speaker 6>scale pieces of software, large scale code basis knows that's

0:20:44.880 --> 0:20:46.119
<v Speaker 6>not the norm.

0:20:46.840 --> 0:20:52.000
<v Speaker 2>So VS code and cursor, these are IDEs. Cloud code

0:20:52.040 --> 0:20:54.520
<v Speaker 2>is not an ID. What it's called a CLI is.

0:20:54.520 --> 0:20:56.160
<v Speaker 6>A CLI a command line interface.

0:20:56.320 --> 0:20:59.600
<v Speaker 2>Got it and the other labs now they also have CLIs.

0:21:00.080 --> 0:21:02.800
<v Speaker 2>So why are we all talking about claud code and

0:21:02.640 --> 0:21:06.520
<v Speaker 2>I chattypts is called codex. I don't know what Geminis

0:21:06.560 --> 0:21:06.960
<v Speaker 2>is called.

0:21:07.080 --> 0:21:08.000
<v Speaker 5>I think it's just called the gem.

0:21:08.760 --> 0:21:11.720
<v Speaker 2>Why are we all talking about claud code rather than

0:21:11.840 --> 0:21:14.480
<v Speaker 2>the other cla is that kind of have the same thing? Like,

0:21:14.520 --> 0:21:15.320
<v Speaker 2>what is the difference?

0:21:15.880 --> 0:21:18.040
<v Speaker 6>I think, first and foremost they were first, okay, so

0:21:18.560 --> 0:21:20.600
<v Speaker 6>and I think they've they've had a lot more and

0:21:20.920 --> 0:21:24.399
<v Speaker 6>you know, from my very personal opinion, I think they've

0:21:24.680 --> 0:21:28.879
<v Speaker 6>done something smarter and better as far as the permissioning model.

0:21:29.000 --> 0:21:30.720
<v Speaker 6>So you know, one of the really dangerous things is

0:21:30.720 --> 0:21:32.360
<v Speaker 6>you've got the same running on your computer. You don't

0:21:32.400 --> 0:21:36.080
<v Speaker 6>want it to go and delete everything. And they have

0:21:36.119 --> 0:21:38.679
<v Speaker 6>a very fine grain permissioning model where you can say, hey,

0:21:38.720 --> 0:21:40.520
<v Speaker 6>I want to allow this just this one time. I

0:21:40.520 --> 0:21:41.560
<v Speaker 6>want to always allow it.

0:21:41.680 --> 0:21:43.840
<v Speaker 2>You know, I always click always allow. I'm living on

0:21:43.840 --> 0:21:44.240
<v Speaker 2>the edge.

0:21:44.480 --> 0:21:47.040
<v Speaker 6>You can you can next time you run it. You

0:21:47.040 --> 0:21:50.080
<v Speaker 6>can just do a flag that says dangerously skip permissions

0:21:50.920 --> 0:21:54.840
<v Speaker 6>and and it'll just They call it yolo mode. I

0:21:54.880 --> 0:21:58.560
<v Speaker 6>think I think more fundamentally though, if I look at

0:21:58.760 --> 0:22:02.480
<v Speaker 6>at Codex versus Code, I think it's a difference in

0:22:02.480 --> 0:22:07.760
<v Speaker 6>philosophy around what you want AI to do to me. Codex,

0:22:08.400 --> 0:22:12.800
<v Speaker 6>which is excellent, is very focused on building an agent

0:22:12.880 --> 0:22:15.199
<v Speaker 6>that you could just give something to and it'll just

0:22:15.280 --> 0:22:17.600
<v Speaker 6>go do it. So I want to give it that task.

0:22:17.680 --> 0:22:19.560
<v Speaker 6>I don't want to intercede. I don't want to give

0:22:19.600 --> 0:22:24.680
<v Speaker 6>it any more feedback. And claud code is much more

0:22:24.720 --> 0:22:28.159
<v Speaker 6>designed to be kind of a pair programmer. And so

0:22:28.320 --> 0:22:31.040
<v Speaker 6>you know, in engineering, pair programming has existed for a while.

0:22:31.280 --> 0:22:34.320
<v Speaker 6>It's a really weird sort of productivity thing where you

0:22:34.400 --> 0:22:36.480
<v Speaker 6>put two engineers on the same problem and it turns

0:22:36.480 --> 0:22:40.120
<v Speaker 6>out that you can get better code and multiplier Yeah,

0:22:40.160 --> 0:22:41.800
<v Speaker 6>and it sort of makes up for the fact that

0:22:41.800 --> 0:22:44.240
<v Speaker 6>obviously you know you're doubling the staff on it. But

0:22:44.359 --> 0:22:47.240
<v Speaker 6>because of how many fewer bugs because you've both said svies,

0:22:47.720 --> 0:22:50.200
<v Speaker 6>it has seemed to work out for many folks. Most

0:22:50.280 --> 0:22:53.840
<v Speaker 6>companies don't practice it, but I think cloud code fundamentally

0:22:53.880 --> 0:22:56.120
<v Speaker 6>is much more designed in that way.

0:22:56.200 --> 0:22:58.280
<v Speaker 5>It's a pair programmer. It's they. You know.

0:22:58.320 --> 0:23:01.560
<v Speaker 6>Whenever I start a project, I start in plan mode.

0:23:01.600 --> 0:23:03.440
<v Speaker 6>So you start in plan mode. You put together a plan.

0:23:03.480 --> 0:23:04.639
<v Speaker 6>I really, I mean it has spent a lot of

0:23:04.720 --> 0:23:07.159
<v Speaker 6>time in plan mode. You go through, it gives you

0:23:07.160 --> 0:23:09.000
<v Speaker 6>a plan, back, it asks you how you feel you

0:23:09.000 --> 0:23:11.320
<v Speaker 6>can give it a whole bunch of direction, and then

0:23:11.359 --> 0:23:13.760
<v Speaker 6>it's only then that it goes off and it goes

0:23:13.800 --> 0:23:16.359
<v Speaker 6>into it. So you know, we're working together. And I

0:23:16.359 --> 0:23:18.560
<v Speaker 6>actually have a whole system now that I've designed where

0:23:19.680 --> 0:23:22.280
<v Speaker 6>I use a task management system called Linear, So I

0:23:22.280 --> 0:23:25.359
<v Speaker 6>have claud Code write tasks off to Linear, and then

0:23:25.640 --> 0:23:29.520
<v Speaker 6>I've worked with claud Code to write a document that

0:23:29.600 --> 0:23:31.919
<v Speaker 6>helps sort of decide a set of heuristics to decide

0:23:31.920 --> 0:23:34.240
<v Speaker 6>when you should assign it to Codex versus when you

0:23:34.280 --> 0:23:36.120
<v Speaker 6>should give it to claud Code. And so if it's

0:23:36.119 --> 0:23:38.240
<v Speaker 6>tightly defined enough and simple enough, I just send it

0:23:38.240 --> 0:23:41.200
<v Speaker 6>off to Codex and it does it totally independently. And

0:23:41.240 --> 0:23:44.280
<v Speaker 6>then if it's complicated enough that I think it requires

0:23:44.320 --> 0:23:47.879
<v Speaker 6>my time and attention, then it saves it for me

0:23:48.920 --> 0:23:52.840
<v Speaker 6>us to do together, and we'll work on it together.

0:23:52.880 --> 0:23:55.520
<v Speaker 6>And so if it's sort of touching a kind of

0:23:55.600 --> 0:23:58.520
<v Speaker 6>important enough if it's changing some part of the data model.

0:23:58.880 --> 0:24:01.440
<v Speaker 6>There's these other kind of you know, fairly basic extet

0:24:01.480 --> 0:24:05.080
<v Speaker 6>of criteria that I use, But that to me is

0:24:05.080 --> 0:24:09.480
<v Speaker 6>the fundamental distinction. And you know, I find cloud code

0:24:09.480 --> 0:24:12.800
<v Speaker 6>in that way to be just it sort of fits

0:24:12.920 --> 0:24:15.359
<v Speaker 6>what I want to do and how I want to

0:24:15.400 --> 0:24:17.080
<v Speaker 6>work much better.

0:24:17.640 --> 0:24:21.320
<v Speaker 3>Talk a little bit more about how it actually impacts

0:24:21.440 --> 0:24:24.960
<v Speaker 3>the workflow of an engineer, because you know, my impression

0:24:25.160 --> 0:24:29.760
<v Speaker 3>was people can code, right, Like, the coding problem is

0:24:29.880 --> 0:24:31.919
<v Speaker 3>kind of solved at this point, and even if you

0:24:32.000 --> 0:24:34.760
<v Speaker 3>can't code, even if you're not a professional engineer, you

0:24:34.800 --> 0:24:38.880
<v Speaker 3>can hire someone from like India or Indonesia or wherever

0:24:39.040 --> 0:24:40.920
<v Speaker 3>to just write you a code. Maybe it'll take them

0:24:40.920 --> 0:24:44.120
<v Speaker 3>a week instead of like two days with claud code.

0:24:44.240 --> 0:24:49.040
<v Speaker 3>But how much does this actually change the workflow for

0:24:49.080 --> 0:24:50.399
<v Speaker 3>an engineer.

0:24:50.880 --> 0:24:53.399
<v Speaker 6>As completely as it could be changed. I mean, I

0:24:53.440 --> 0:24:58.879
<v Speaker 6>would say that over the last three months, I've written, personally,

0:24:59.640 --> 0:25:01.679
<v Speaker 6>I don't know a few hundred lines of code. Like

0:25:01.840 --> 0:25:05.920
<v Speaker 6>I am mostly a manager of a set of agents

0:25:05.960 --> 0:25:08.879
<v Speaker 6>who are writing code on my behalf. And you know,

0:25:08.960 --> 0:25:11.680
<v Speaker 6>increasingly what I think is interesting. I've been thinking about

0:25:11.680 --> 0:25:14.400
<v Speaker 6>this a bunch lately, is like, in some ways, it's

0:25:14.400 --> 0:25:16.479
<v Speaker 6>just bringing me back to the core challenge that has

0:25:16.480 --> 0:25:19.639
<v Speaker 6>always existed in software development, which is how do you

0:25:19.720 --> 0:25:23.320
<v Speaker 6>manage all large scale software.

0:25:23.000 --> 0:25:24.680
<v Speaker 5>Development project acrossation.

0:25:25.200 --> 0:25:28.680
<v Speaker 6>It has become a coordination problem, and I spend a

0:25:28.760 --> 0:25:33.000
<v Speaker 6>lot of time sort of now designing my CLAUD code

0:25:33.040 --> 0:25:35.879
<v Speaker 6>system to ensure that code goes through all the proper

0:25:35.960 --> 0:25:39.280
<v Speaker 6>spec checks and that it has all these things. The

0:25:39.359 --> 0:25:43.040
<v Speaker 6>other thing that you know makes code a particularly good

0:25:43.080 --> 0:25:45.480
<v Speaker 6>place to do this is that code is verifiable in

0:25:45.520 --> 0:25:47.879
<v Speaker 6>a way that you know most other work is not.

0:25:48.440 --> 0:25:51.720
<v Speaker 6>So you know, with code, you can verify that the

0:25:51.800 --> 0:25:53.840
<v Speaker 6>build works right. So you can say, hey, I want

0:25:53.880 --> 0:25:56.040
<v Speaker 6>to build this, I want to build this package. I

0:25:56.040 --> 0:25:57.560
<v Speaker 6>want to make sure that it's actually going to build

0:25:57.560 --> 0:25:59.240
<v Speaker 6>and that there's going to be no failures. That's a

0:25:59.320 --> 0:26:01.520
<v Speaker 6>very easy check. It's either true or it's not true.

0:26:01.880 --> 0:26:06.639
<v Speaker 6>There's also coders use linting, and so linting is a

0:26:06.680 --> 0:26:10.280
<v Speaker 6>way to kind of look at it's static code analysis,

0:26:10.280 --> 0:26:11.960
<v Speaker 6>so it basically tries to sort.

0:26:11.760 --> 0:26:12.680
<v Speaker 5>Of find.

0:26:15.200 --> 0:26:17.119
<v Speaker 6>Things in your code base that are not going to

0:26:17.160 --> 0:26:19.840
<v Speaker 6>work ahead of time where you can predict that. Obviously,

0:26:19.880 --> 0:26:23.840
<v Speaker 6>you can't predict Alan Turing prove that you can't predict

0:26:24.560 --> 0:26:26.280
<v Speaker 6>with certainty whether code is going to run, but there's

0:26:26.280 --> 0:26:28.879
<v Speaker 6>certain patterns and things that it can find. It's essentially

0:26:28.960 --> 0:26:32.159
<v Speaker 6>does static pattern analysis, and so you know, you have

0:26:32.200 --> 0:26:35.000
<v Speaker 6>it run all these things. But the more kind of

0:26:35.000 --> 0:26:38.040
<v Speaker 6>opinionated you can be about that, and the more steps

0:26:38.080 --> 0:26:39.800
<v Speaker 6>you can have it go through, so I find, you know,

0:26:39.880 --> 0:26:42.359
<v Speaker 6>now I'm kind of the designer, which, honestly, as an

0:26:42.440 --> 0:26:45.360
<v Speaker 6>entrepreneur and as a CEO of company is like that's

0:26:45.480 --> 0:26:48.240
<v Speaker 6>kind of always been my job. Like I've been not

0:26:48.440 --> 0:26:50.160
<v Speaker 6>I have less and less been a person who writes

0:26:50.160 --> 0:26:52.240
<v Speaker 6>code and more and more been a person who designs

0:26:52.640 --> 0:26:54.440
<v Speaker 6>a system in that case of company with a bunch

0:26:54.440 --> 0:26:56.840
<v Speaker 6>of people who write code.

0:26:57.440 --> 0:27:00.199
<v Speaker 2>One of the funny things it seems to me is

0:27:00.240 --> 0:27:05.280
<v Speaker 2>that setting aside Claud code, Claude itself has a reputation

0:27:05.560 --> 0:27:08.720
<v Speaker 2>for it's a nicer chat bot to talk to. People

0:27:08.840 --> 0:27:13.280
<v Speaker 2>find it, and you know, chat gbt seems to really

0:27:13.359 --> 0:27:16.040
<v Speaker 2>be psycho fantic. I still think it's I know it's improved,

0:27:16.040 --> 0:27:17.600
<v Speaker 2>but I actually don't think it's improved or not. I

0:27:17.680 --> 0:27:22.719
<v Speaker 2>still people like the prose style of Claud Claude and

0:27:22.760 --> 0:27:26.360
<v Speaker 2>I'm curious that in the pair trading pair trading, I'm

0:27:26.359 --> 0:27:30.199
<v Speaker 2>thinking about finance, the pair engineering model, whether there is

0:27:30.320 --> 0:27:33.159
<v Speaker 2>also an edge there which is like, here is a

0:27:33.280 --> 0:27:36.440
<v Speaker 2>chat bot that is not annoying to talk to while

0:27:36.440 --> 0:27:39.280
<v Speaker 2>you're iterating, and whether that is like a meaningful distinction

0:27:39.720 --> 0:27:43.000
<v Speaker 2>between you know, coding with codex or whatever.

0:27:43.840 --> 0:27:47.800
<v Speaker 6>Yeah, I I don't know. It still can be very annoying,

0:27:47.840 --> 0:27:52.560
<v Speaker 6>I can tell you. And it'll still sometimes be overly,

0:27:54.480 --> 0:27:57.239
<v Speaker 6>overly effusive with me about a design choice, I mean,

0:27:57.680 --> 0:28:00.640
<v Speaker 6>or sort of notice something which I could live without.

0:28:01.000 --> 0:28:04.160
<v Speaker 2>So I work on this project that's doing this linguistic

0:28:04.200 --> 0:28:07.040
<v Speaker 2>things and I eventually had to say, like, give it

0:28:07.040 --> 0:28:09.399
<v Speaker 2>to me straight, how bad is this? And then so

0:28:09.440 --> 0:28:12.000
<v Speaker 2>I said, I said, Actually, what I said was assume

0:28:12.080 --> 0:28:14.359
<v Speaker 2>for a moment that you are a quantitative linguistics for

0:28:14.440 --> 0:28:17.840
<v Speaker 2>the PhD. Give me your honest assessment of where we

0:28:17.880 --> 0:28:21.480
<v Speaker 2>are with this. And it said like you've developed a

0:28:21.560 --> 0:28:24.680
<v Speaker 2>nice toy and there's no evidence that it actually does.

0:28:24.800 --> 0:28:26.639
<v Speaker 2>And I was like, okay, that's nice to hear. I

0:28:26.680 --> 0:28:29.040
<v Speaker 2>actually like, you know, I appreciate that, and I thank

0:28:29.040 --> 0:28:32.280
<v Speaker 2>you very blunt, not you know, it's still like polite,

0:28:32.480 --> 0:28:35.320
<v Speaker 2>but it was like, this, doesn't you haven't really shown anything.

0:28:35.400 --> 0:28:37.560
<v Speaker 2>You haven't really established at all that your saltware does

0:28:37.600 --> 0:28:38.280
<v Speaker 2>what it claims to.

0:28:38.600 --> 0:28:42.560
<v Speaker 6>Yeah, I think so. I think stylistically I kind of

0:28:42.600 --> 0:28:46.920
<v Speaker 6>personally agree. My theory by the way, on Claude versus

0:28:47.920 --> 0:28:50.800
<v Speaker 6>Opening Eye Chat GPT models is I think Claude is

0:28:50.840 --> 0:28:54.000
<v Speaker 6>actually better at sort of reflecting what you give it.

0:28:54.640 --> 0:28:56.240
<v Speaker 6>And so I think part of why we think it's

0:28:56.280 --> 0:28:57.920
<v Speaker 6>better is it it's better at.

0:28:57.800 --> 0:28:59.080
<v Speaker 5>Pretending it's us.

0:28:59.280 --> 0:29:01.520
<v Speaker 6>Yeah, And so we tend to like that is this

0:29:01.600 --> 0:29:05.160
<v Speaker 6>is purely speculation, but that's always been my theory on so.

0:29:05.200 --> 0:29:06.440
<v Speaker 2>It flatters you in a different way.

0:29:06.720 --> 0:29:09.520
<v Speaker 5>I think it's flattering you in a much more way.

0:29:09.600 --> 0:29:09.840
<v Speaker 1>Yeah.

0:29:09.960 --> 0:29:15.360
<v Speaker 6>Interest But for a long time, just Anthropic has been

0:29:15.400 --> 0:29:18.040
<v Speaker 6>producing the best coding models, you know. I mean there's

0:29:18.240 --> 0:29:21.160
<v Speaker 6>there can be some debate there now, but you know,

0:29:21.200 --> 0:29:24.280
<v Speaker 6>there's a great curse story from Cursor actually where Cursor

0:29:24.480 --> 0:29:27.920
<v Speaker 6>basically wasn't that good, and then SunNet three point five

0:29:27.960 --> 0:29:30.760
<v Speaker 6>came out and all of a sudden, Cursor was amazing,

0:29:30.960 --> 0:29:33.760
<v Speaker 6>and Cursor became a tool that everybody started using. But

0:29:33.800 --> 0:29:35.840
<v Speaker 6>it wasn't until this other model came out and they

0:29:35.880 --> 0:29:38.800
<v Speaker 6>made that the default model, and you know, I for

0:29:38.840 --> 0:29:40.720
<v Speaker 6>what it's worth. I think the other takeaway from that,

0:29:40.840 --> 0:29:43.480
<v Speaker 6>which is a kind of big theme we see in

0:29:43.520 --> 0:29:46.160
<v Speaker 6>the market as a thing that the cloud Code team

0:29:46.200 --> 0:29:49.160
<v Speaker 6>has talked about, is you constantly have to be building

0:29:49.320 --> 0:29:52.600
<v Speaker 6>ahead with AI in a way that is very unique

0:29:52.640 --> 0:29:55.280
<v Speaker 6>in the world of software, where you kind of always

0:29:55.320 --> 0:29:58.160
<v Speaker 6>want to build things that are working at like seventy

0:29:58.240 --> 0:30:00.480
<v Speaker 6>or eighty percent, because if you really spend the time

0:30:00.520 --> 0:30:02.640
<v Speaker 6>to get it up to ninety or one hundred, you're

0:30:02.680 --> 0:30:04.240
<v Speaker 6>going to lose all the games you get when the

0:30:04.280 --> 0:30:07.160
<v Speaker 6>next model comes out, and the you know, with the

0:30:07.200 --> 0:30:09.560
<v Speaker 6>amount of capex being spent on these models, like there's

0:30:09.600 --> 0:30:11.480
<v Speaker 6>a next model that's going to come out that's going

0:30:11.520 --> 0:30:13.240
<v Speaker 6>to be awesome, and you just kind of want to

0:30:13.240 --> 0:30:15.480
<v Speaker 6>be downstream from that, and you don't want to waste

0:30:15.520 --> 0:30:18.400
<v Speaker 6>six months getting an extra three percent when that new

0:30:18.400 --> 0:30:19.960
<v Speaker 6>model is going to give you an extra seven.

0:30:20.840 --> 0:30:23.200
<v Speaker 3>Yeah, this is the only certainty with AI is like

0:30:23.280 --> 0:30:24.680
<v Speaker 3>there's always going to be a new model.

0:30:25.440 --> 0:30:27.920
<v Speaker 2>The worst model ever use is the one that we're using.

0:30:28.240 --> 0:30:40.080
<v Speaker 4>That's right, That's right.

0:30:44.560 --> 0:30:47.800
<v Speaker 3>Are we all going to become coding illiterate? Are we

0:30:47.880 --> 0:30:50.360
<v Speaker 3>just going to forget how to code. If everyone's using

0:30:50.640 --> 0:30:52.440
<v Speaker 3>you know, general language to do.

0:30:52.480 --> 0:30:54.600
<v Speaker 2>Forget, I've never learned. Yeah, okay, you know what I've

0:30:54.600 --> 0:30:55.240
<v Speaker 2>been thinking about.

0:30:55.560 --> 0:30:56.040
<v Speaker 1>You know that.

0:30:57.680 --> 0:31:01.160
<v Speaker 2>Scott Karp, the CEO Palanteer, here's that line. He's like,

0:31:01.400 --> 0:31:03.120
<v Speaker 2>when I was young, I was too poor to have

0:31:03.160 --> 0:31:04.760
<v Speaker 2>a car, or so I didn't get a so I

0:31:04.800 --> 0:31:06.960
<v Speaker 2>never learned to drive. And now I'm too rich, so

0:31:07.000 --> 0:31:09.240
<v Speaker 2>I never learned to drive. I feel like when I

0:31:09.280 --> 0:31:11.240
<v Speaker 2>was young, I was too dumb to learn to code,

0:31:11.400 --> 0:31:12.320
<v Speaker 2>and now.

0:31:12.480 --> 0:31:13.280
<v Speaker 3>You leaped ahead.

0:31:13.440 --> 0:31:16.080
<v Speaker 2>Yeah, now I'm too smart to learn Python or HTML

0:31:16.160 --> 0:31:16.560
<v Speaker 2>or whatever.

0:31:17.440 --> 0:31:19.920
<v Speaker 6>I have a couple of takes on this person one personally.

0:31:19.960 --> 0:31:23.880
<v Speaker 6>So first one is I just think like this is

0:31:23.920 --> 0:31:25.160
<v Speaker 6>the worry of all technology.

0:31:25.240 --> 0:31:25.440
<v Speaker 5>Ever.

0:31:25.600 --> 0:31:28.440
<v Speaker 6>There was a paper that came out that showed that

0:31:28.720 --> 0:31:32.400
<v Speaker 6>people were, uh, you know, they were forgetting more things

0:31:32.480 --> 0:31:34.880
<v Speaker 6>or something because they were using chat GBD. But you know, uh,

0:31:35.800 --> 0:31:39.440
<v Speaker 6>in Phadrious Plato was worried that people were gonna forget

0:31:39.480 --> 0:31:42.440
<v Speaker 6>things because they started writing things down. And you know,

0:31:42.480 --> 0:31:44.160
<v Speaker 6>I think the trade off there was pretty good. We

0:31:44.240 --> 0:31:47.840
<v Speaker 6>got the scientific revolution a couple of other things, So, uh,

0:31:48.400 --> 0:31:51.040
<v Speaker 6>you know, I think that's the sort of natural knee jerk.

0:31:52.720 --> 0:31:57.200
<v Speaker 6>With that said, it is. I it's very strange when

0:31:57.240 --> 0:32:00.160
<v Speaker 6>you have people, you know, the Cloud Code team is

0:32:00.200 --> 0:32:04.080
<v Speaker 6>talking about how little code they write. Now, I draw

0:32:04.080 --> 0:32:06.800
<v Speaker 6>a distinction between the sort of Vibe coding and the

0:32:07.080 --> 0:32:10.479
<v Speaker 6>kind of amateur people who have never written code. And

0:32:10.840 --> 0:32:12.800
<v Speaker 6>I think that is amazing, by the way, And I

0:32:12.800 --> 0:32:15.800
<v Speaker 6>think there's a lot of software developers who are really

0:32:15.800 --> 0:32:19.680
<v Speaker 6>mad about that because there they claim it's for safety

0:32:19.680 --> 0:32:21.800
<v Speaker 6>reasons or whatever, but I think fundamentally it's just they've

0:32:21.800 --> 0:32:25.480
<v Speaker 6>got people on their turf. But I think that's incredible.

0:32:25.480 --> 0:32:29.240
<v Speaker 6>I mean, my hat, my my nine year old Vibe

0:32:29.240 --> 0:32:33.880
<v Speaker 6>coded a website wo and for Secret Santa. She's now ten.

0:32:34.080 --> 0:32:35.520
<v Speaker 6>She would get bad at me if I called her nine,

0:32:35.560 --> 0:32:37.600
<v Speaker 6>but I think she vibe good at when she was nine.

0:32:38.040 --> 0:32:40.440
<v Speaker 6>But that that's awesome, right, I don't know, that's amazing.

0:32:40.480 --> 0:32:43.000
<v Speaker 6>That's a way for people to express themselves in a

0:32:43.040 --> 0:32:46.440
<v Speaker 6>way that they couldn't before you did your your linguistics process.

0:32:46.480 --> 0:32:51.480
<v Speaker 6>That's that's fun and interesting. But yeah, I also think

0:32:51.520 --> 0:32:54.720
<v Speaker 6>the other the the thing that's happening with professional software

0:32:54.720 --> 0:32:56.960
<v Speaker 6>developers when you hear from anthropic or you know, when

0:32:56.960 --> 0:33:00.600
<v Speaker 6>I'm talking about it's you know, the code going through

0:33:00.600 --> 0:33:04.360
<v Speaker 6>this process, and you know, all the code still gets

0:33:04.400 --> 0:33:06.080
<v Speaker 6>reviewed by people. We're not letting it get out the

0:33:06.120 --> 0:33:08.400
<v Speaker 6>door if it's not at the same level as human

0:33:08.560 --> 0:33:11.480
<v Speaker 6>and it's just But what's amazing is I'm I'm running

0:33:11.520 --> 0:33:13.400
<v Speaker 6>five of these sessions at a time, right, and so

0:33:13.440 --> 0:33:16.000
<v Speaker 6>I've got like software being developed in parallel in a

0:33:16.040 --> 0:33:19.280
<v Speaker 6>way that is unimaginable. And you know the other thing

0:33:19.320 --> 0:33:23.080
<v Speaker 6>is just now, the best software engineers wrote the least

0:33:23.080 --> 0:33:26.800
<v Speaker 6>code anyway. You know the sort of classic story of

0:33:26.840 --> 0:33:29.640
<v Speaker 6>like the difference between a junior developer or senior developers

0:33:29.640 --> 0:33:32.720
<v Speaker 6>that a junior developer gets a problem and they sit

0:33:32.760 --> 0:33:34.320
<v Speaker 6>down and they put their fingers on the keyboard and

0:33:34.320 --> 0:33:37.320
<v Speaker 6>they start writing code. And a senior developer gets a

0:33:37.320 --> 0:33:39.920
<v Speaker 6>problem and sits there for three hours and tries to

0:33:39.920 --> 0:33:41.360
<v Speaker 6>figure out what the best way to solve it is

0:33:41.400 --> 0:33:43.720
<v Speaker 6>and then spends five minutes writing code to get it done.

0:33:43.960 --> 0:33:46.400
<v Speaker 3>True elegance is restraint. That's what I say.

0:33:46.840 --> 0:33:47.280
<v Speaker 4>What are you.

0:33:47.280 --> 0:33:50.440
<v Speaker 2>Seeing in the companies you're working for? Like, I find

0:33:50.480 --> 0:33:53.200
<v Speaker 2>it hard to believe and I was maybe skeptical of this,

0:33:53.320 --> 0:33:55.800
<v Speaker 2>but it feels like right now we're here with technology,

0:33:55.800 --> 0:33:58.640
<v Speaker 2>where like if viral like company is like, like I said,

0:33:58.640 --> 0:34:00.920
<v Speaker 2>you can build charts of data in a way that

0:34:01.040 --> 0:34:02.800
<v Speaker 2>used to be like someone would have had to get

0:34:02.800 --> 0:34:04.760
<v Speaker 2>their hands dirty, ear, et cetera. And the companies that

0:34:04.840 --> 0:34:07.720
<v Speaker 2>you talk to is right now that having an effect

0:34:07.800 --> 0:34:09.920
<v Speaker 2>on how they think about what positions they're hiring for

0:34:09.960 --> 0:34:11.680
<v Speaker 2>and the skills they're looking for and so forth.

0:34:12.200 --> 0:34:17.360
<v Speaker 6>I think that it's hard to answer, right, really, I

0:34:17.440 --> 0:34:22.239
<v Speaker 6>think that certainly. I do think I personally think if

0:34:22.280 --> 0:34:24.440
<v Speaker 6>I look at the sort of layoffs in the technology

0:34:24.480 --> 0:34:26.080
<v Speaker 6>industry of the last couple of years, I think some

0:34:26.200 --> 0:34:28.600
<v Speaker 6>part of that is just looking at the output of

0:34:28.640 --> 0:34:33.000
<v Speaker 6>these models and saying, hey, these models are able to

0:34:33.000 --> 0:34:35.880
<v Speaker 6>produce it, you know, the median, and I have a

0:34:35.880 --> 0:34:38.120
<v Speaker 6>whole bunch of sort of middle managers who are producing

0:34:38.160 --> 0:34:41.040
<v Speaker 6>it the sixty fifth percentile. And it's like I can

0:34:41.120 --> 0:34:44.440
<v Speaker 6>produce median for a dollar fifty per million tokens, or

0:34:44.480 --> 0:34:48.440
<v Speaker 6>I can produce sixty fifth percentile for hover many hundreds

0:34:48.480 --> 0:34:50.160
<v Speaker 6>of thousands dollars a year. It's it's a sort of

0:34:50.160 --> 0:34:53.399
<v Speaker 6>fairly simple trade off, I think. So I do think

0:34:53.400 --> 0:34:55.000
<v Speaker 6>there's a lot of downstream effects. I think the other

0:34:55.000 --> 0:34:57.480
<v Speaker 6>thing that's happening, is is kind of like middle management

0:34:57.560 --> 0:35:00.239
<v Speaker 6>is under threat because it's the realization that hey, like,

0:35:00.520 --> 0:35:02.880
<v Speaker 6>part of what these models are amazing at is is

0:35:03.360 --> 0:35:05.080
<v Speaker 6>I think of them as like a fuzzy interface. They

0:35:05.120 --> 0:35:07.920
<v Speaker 6>can sort of turn any data into any other data, right.

0:35:07.960 --> 0:35:11.200
<v Speaker 6>You can sort of transform data from one format to another.

0:35:11.360 --> 0:35:13.000
<v Speaker 6>You can take a PDF and you can turn it

0:35:13.040 --> 0:35:15.279
<v Speaker 6>into charts, right, And there's whole people who exist or

0:35:15.400 --> 0:35:17.120
<v Speaker 6>you know, if you think about what product managers do

0:35:17.160 --> 0:35:19.000
<v Speaker 6>a lot of what product managers do is they take

0:35:19.239 --> 0:35:21.040
<v Speaker 6>how people are using a product and they try to

0:35:21.080 --> 0:35:25.319
<v Speaker 6>transform it into a format that engineers can then use

0:35:25.400 --> 0:35:27.000
<v Speaker 6>to figure out what to do. And I think a

0:35:27.040 --> 0:35:31.080
<v Speaker 6>lot of those kind of a lot of those pieces

0:35:31.239 --> 0:35:35.880
<v Speaker 6>that used to just be kind of transferring knowledge.

0:35:36.080 --> 0:35:39.520
<v Speaker 2>I've always said, Tracy, I think one of the most

0:35:39.560 --> 0:35:42.920
<v Speaker 2>important roles in any organization is essentially translation work. And

0:35:42.960 --> 0:35:46.120
<v Speaker 2>you see it in the newsroom where it's like here's

0:35:46.160 --> 0:35:50.919
<v Speaker 2>a team specialized in emerging market currencies, and then they

0:35:50.960 --> 0:35:53.280
<v Speaker 2>have to like they have to then tell the senior

0:35:53.400 --> 0:35:56.040
<v Speaker 2>editors what they're working on. But the senior editors, who

0:35:56.040 --> 0:35:58.920
<v Speaker 2>are maybe more generalists, don't really know like why like

0:35:59.000 --> 0:36:02.360
<v Speaker 2>some sort of like you know, one yen carry is

0:36:02.400 --> 0:36:05.600
<v Speaker 2>important and that a really important role within any organization

0:36:05.719 --> 0:36:08.680
<v Speaker 2>is essentially the team that can translate between the generalist

0:36:08.680 --> 0:36:11.319
<v Speaker 2>team and the specialist team. Absolutely, and so that's an

0:36:11.320 --> 0:36:14.120
<v Speaker 2>interesting observation and the sort of engineering world of like, okay,

0:36:14.120 --> 0:36:17.000
<v Speaker 2>these are tools that are in sometimes translation tools.

0:36:17.480 --> 0:36:19.840
<v Speaker 3>So we talked I agree completely, by the way, but

0:36:20.160 --> 0:36:23.359
<v Speaker 3>we talked about vibe coding and Joe has this application

0:36:24.160 --> 0:36:26.879
<v Speaker 3>that I don't think you're looking to monetize.

0:36:26.400 --> 0:36:27.960
<v Speaker 2>No, it's I'm just trying to make it for the

0:36:28.000 --> 0:36:30.160
<v Speaker 2>good of the world, right, Okay, so when did that

0:36:30.160 --> 0:36:33.640
<v Speaker 2>become a crown? I'm not a lotizing it, But.

0:36:33.719 --> 0:36:37.400
<v Speaker 3>Like this opens up massive questions for software as a service,

0:36:37.480 --> 0:36:41.239
<v Speaker 3>right versas because if everyone can write their own software,

0:36:42.480 --> 0:36:45.320
<v Speaker 3>you can replicate anything that's out there that is currently

0:36:45.480 --> 0:36:47.400
<v Speaker 3>charging money. What's going to happen to software?

0:36:48.200 --> 0:36:50.839
<v Speaker 6>I think software is pretty screwed. A lot of it,

0:36:50.880 --> 0:36:53.120
<v Speaker 6>at least not all of it. You know, you still

0:36:53.480 --> 0:36:55.880
<v Speaker 6>it depends on whether you call that cloud provider software

0:36:55.960 --> 0:36:57.640
<v Speaker 6>or not. You know, you still need to run this

0:36:57.640 --> 0:37:00.360
<v Speaker 6>stuff somewhere. And I think there's there's certain kind of

0:37:00.400 --> 0:37:02.920
<v Speaker 6>software that you know, you just don't really want to

0:37:02.960 --> 0:37:06.160
<v Speaker 6>be in the business of writing, you know, I as

0:37:06.200 --> 0:37:09.040
<v Speaker 6>someone who's tried to build a project management system, I'd

0:37:09.080 --> 0:37:11.840
<v Speaker 6>really rather I don't think anybody should be in that business.

0:37:13.400 --> 0:37:15.960
<v Speaker 6>But I do think fundamentally, I mean, we see this

0:37:16.120 --> 0:37:19.759
<v Speaker 6>every day inside enterprises. The sort of build versus buy

0:37:20.840 --> 0:37:23.960
<v Speaker 6>pendulum has just swung. And you know, I mean I

0:37:24.120 --> 0:37:26.960
<v Speaker 6>used to run a SaaS company and we sold to enterprises,

0:37:27.000 --> 0:37:29.919
<v Speaker 6>and you know, for a long time that I think

0:37:29.960 --> 0:37:32.520
<v Speaker 6>that made a lot of sense, right because like, hey,

0:37:32.880 --> 0:37:34.440
<v Speaker 6>it just didn't make sense to try to build this

0:37:34.480 --> 0:37:37.400
<v Speaker 6>thing on your own. And so but the price of

0:37:37.400 --> 0:37:39.960
<v Speaker 6>that was, you know, won the price, right, like, and

0:37:40.040 --> 0:37:41.759
<v Speaker 6>it got to be more and more expensive. The other

0:37:41.800 --> 0:37:43.320
<v Speaker 6>price was that you were paying for a lot of

0:37:43.360 --> 0:37:45.600
<v Speaker 6>stuff you didn't need, right, because the whole job of

0:37:45.640 --> 0:37:49.719
<v Speaker 6>building SaaS is you need to generalize problems, and so

0:37:49.920 --> 0:37:52.719
<v Speaker 6>you build things that are going to work for everybody.

0:37:53.040 --> 0:37:54.680
<v Speaker 5>And that means either you.

0:37:54.600 --> 0:37:56.279
<v Speaker 6>Have to sort of adapt or you have to build

0:37:56.320 --> 0:38:00.880
<v Speaker 6>this sort of very configurable software. And I think, and

0:38:01.400 --> 0:38:03.560
<v Speaker 6>what I see you just you know firsthand, is that

0:38:05.080 --> 0:38:10.440
<v Speaker 6>inside these organizations you can now solve very specific problems

0:38:10.440 --> 0:38:14.680
<v Speaker 6>that are highly valuable and not only can you solve

0:38:14.680 --> 0:38:18.279
<v Speaker 6>them better than generic software, but you can actually, in

0:38:18.320 --> 0:38:20.360
<v Speaker 6>a lot of ways do it for less money because

0:38:20.360 --> 0:38:23.000
<v Speaker 6>you're trying to tackle less stuff. You didn't need the

0:38:23.080 --> 0:38:25.319
<v Speaker 6>sixteen other features. You bought it for the one that

0:38:25.400 --> 0:38:29.480
<v Speaker 6>you really really cared about. And so I think that

0:38:29.640 --> 0:38:31.400
<v Speaker 6>part of it, you know, I don't like there's I

0:38:31.680 --> 0:38:34.120
<v Speaker 6>definitely think there are pieces of the software industry that

0:38:34.239 --> 0:38:36.200
<v Speaker 6>are gonna, you know, come out the other side. You're

0:38:36.200 --> 0:38:38.799
<v Speaker 6>gonna nobody wants to deal with payroll, right like you know,

0:38:38.840 --> 0:38:41.480
<v Speaker 6>somebody you're still gonna buy some payroll software and you're

0:38:41.520 --> 0:38:44.000
<v Speaker 6>still gonna have that. But you know, I do think

0:38:44.000 --> 0:38:46.840
<v Speaker 6>there are a lot of pieces where the software existed

0:38:46.960 --> 0:38:50.000
<v Speaker 6>essentially as a kind of wrapper around a database, and

0:38:50.040 --> 0:38:53.120
<v Speaker 6>now you're just gonna, you know, with just the database,

0:38:53.160 --> 0:38:55.120
<v Speaker 6>you can do that. And then you know, the other

0:38:55.160 --> 0:38:57.040
<v Speaker 6>piece I'd say here is it's this is not this

0:38:57.160 --> 0:38:59.239
<v Speaker 6>is a kind of confluence of circumstances where it's not

0:38:59.360 --> 0:39:03.880
<v Speaker 6>just the coding, it's also the fact that you have

0:39:04.080 --> 0:39:06.200
<v Speaker 6>ai to do a whole bunch of work. So you know,

0:39:06.200 --> 0:39:08.719
<v Speaker 6>if we pick on CRM for a second, right, like

0:39:09.120 --> 0:39:13.960
<v Speaker 6>you know, Salesforce dot Com Salesforce dot com. We can

0:39:14.239 --> 0:39:16.400
<v Speaker 6>you know, you look at what the interface of that is.

0:39:16.440 --> 0:39:20.120
<v Speaker 6>And essentially it has existed to get salespeople to take

0:39:20.239 --> 0:39:22.839
<v Speaker 6>unstructured data, which is sales meetings, and turn it into

0:39:22.880 --> 0:39:25.839
<v Speaker 6>structured data that so can be stored into database. And

0:39:25.920 --> 0:39:29.960
<v Speaker 6>now you have AI, and AI is very capable of

0:39:30.040 --> 0:39:33.680
<v Speaker 6>taking unstructured data directly from the source, so you have

0:39:33.719 --> 0:39:37.160
<v Speaker 6>people recording meetings, and then it can structure it into

0:39:37.200 --> 0:39:39.400
<v Speaker 6>any data that you want. This is one of the

0:39:39.480 --> 0:39:42.320
<v Speaker 6>very first sort of mind blowing moments I had was

0:39:42.360 --> 0:39:46.399
<v Speaker 6>that I could give it a Jason interface. I could

0:39:46.400 --> 0:39:49.480
<v Speaker 6>describe exactly what I wanted the data structure to be,

0:39:49.920 --> 0:39:52.480
<v Speaker 6>and it would give me back that information in that

0:39:52.560 --> 0:39:55.080
<v Speaker 6>data structure. And we've just basically been having a bunch

0:39:55.080 --> 0:39:56.799
<v Speaker 6>of humans do that work for a very long time,

0:39:56.840 --> 0:39:58.880
<v Speaker 6>whether it's in CRM or project management or any of

0:39:58.880 --> 0:40:00.800
<v Speaker 6>these other places. And the ability to just kind of

0:40:00.840 --> 0:40:02.640
<v Speaker 6>get rid of that whole thing, I think it really

0:40:02.640 --> 0:40:05.480
<v Speaker 6>does bring into question the value of a lot of

0:40:05.480 --> 0:40:06.400
<v Speaker 6>these software companies.

0:40:06.520 --> 0:40:08.840
<v Speaker 2>Well, so we have seen like a lot of software sucks.

0:40:08.880 --> 0:40:11.960
<v Speaker 2>They look like melting ice cubes right now. Maybe they

0:40:12.440 --> 0:40:14.799
<v Speaker 2>So what is it I want to talk. I mean,

0:40:14.840 --> 0:40:17.000
<v Speaker 2>this is like you know, our listeners who are investors,

0:40:17.080 --> 0:40:19.800
<v Speaker 2>there's a pretty high stakes question of like what residual

0:40:19.960 --> 0:40:22.960
<v Speaker 2>value there is. But talk a little bit more about Salesforce.

0:40:23.000 --> 0:40:24.480
<v Speaker 2>Maybe there would be a time to learn what sale

0:40:24.840 --> 0:40:28.040
<v Speaker 2>what it actually does as it's massively being disrupted. Now

0:40:28.040 --> 0:40:30.560
<v Speaker 2>we get around to learning what Salesforce is. But I

0:40:30.600 --> 0:40:32.560
<v Speaker 2>know it's like many things there are apps that people

0:40:32.560 --> 0:40:35.279
<v Speaker 2>built onto Salesforce. But this sounds like we're hitting on

0:40:35.360 --> 0:40:37.160
<v Speaker 2>when I think probably one of the crucial questions for

0:40:37.239 --> 0:40:39.600
<v Speaker 2>like the future of the software industry. So talk a

0:40:39.600 --> 0:40:42.120
<v Speaker 2>little bit more about like the current approach and what

0:40:42.160 --> 0:40:45.879
<v Speaker 2>people are buying when they buy a package or subscribe

0:40:45.960 --> 0:40:49.000
<v Speaker 2>to a service from Salesforce, and then what the unlocked

0:40:49.000 --> 0:40:53.319
<v Speaker 2>opportunity is from having AI like live in the same

0:40:53.320 --> 0:40:54.520
<v Speaker 2>world as all your files.

0:40:55.040 --> 0:40:58.759
<v Speaker 6>Yeah, so I think if we take CRM as the

0:40:58.800 --> 0:41:01.640
<v Speaker 6>general category, know the biggest players there are.

0:41:01.640 --> 0:41:02.719
<v Speaker 2>That's customer relationship.

0:41:02.719 --> 0:41:05.920
<v Speaker 6>Customer relationship management, that's like what you know, Salesforce does it,

0:41:05.960 --> 0:41:08.600
<v Speaker 6>SVP does it, HubSpot does it for the mid market.

0:41:10.160 --> 0:41:12.040
<v Speaker 6>You know, when I think about that product and I

0:41:12.040 --> 0:41:15.600
<v Speaker 6>think about the way we've used it inside enterprise sales organizations. Essentially,

0:41:15.680 --> 0:41:18.280
<v Speaker 6>you know, it's a database of companies, it's a database

0:41:18.280 --> 0:41:20.360
<v Speaker 6>of contacts, it's a database of deals you have in

0:41:20.400 --> 0:41:22.799
<v Speaker 6>the pipeline, and it's a way to track all those

0:41:22.840 --> 0:41:25.200
<v Speaker 6>deals you guys hit on something before that I think

0:41:25.239 --> 0:41:30.680
<v Speaker 6>is really it, which is like inside companies, there is

0:41:30.719 --> 0:41:33.400
<v Speaker 6>a huge group of people and who exist to answer

0:41:33.440 --> 0:41:36.000
<v Speaker 6>the question from management of what is the status of

0:41:36.080 --> 0:41:38.720
<v Speaker 6>something right? And you know that can be sales management,

0:41:38.760 --> 0:41:40.360
<v Speaker 6>it can be product management, it doesn't matter, right, it

0:41:40.400 --> 0:41:43.520
<v Speaker 6>could be within a newsroom. Somebody wants to know what

0:41:43.560 --> 0:41:45.879
<v Speaker 6>the status is and somebody else exists to go figure

0:41:45.920 --> 0:41:49.200
<v Speaker 6>out what the answer to that question is. And so fundamentally,

0:41:49.719 --> 0:41:52.880
<v Speaker 6>I think those CRM tools are bought first and foremost

0:41:53.120 --> 0:41:55.760
<v Speaker 6>to answer what is the status right? What's my pipeline

0:41:55.760 --> 0:41:58.040
<v Speaker 6>look like? And to answer what your pipeline looks like,

0:41:58.040 --> 0:42:01.319
<v Speaker 6>you need a bunch of salespeople putting deals in and

0:42:01.360 --> 0:42:03.520
<v Speaker 6>those deals are associated with contacts and companies and they

0:42:03.520 --> 0:42:07.240
<v Speaker 6>say when is that deal going to close? And essentially

0:42:07.239 --> 0:42:09.880
<v Speaker 6>you were asking the sales people to make the updates

0:42:09.880 --> 0:42:14.560
<v Speaker 6>in the system to do that and just very tactically,

0:42:14.680 --> 0:42:17.319
<v Speaker 6>I mean, you know, I run a company. Now we

0:42:17.440 --> 0:42:18.880
<v Speaker 6>talk to a lot of we have a lot of

0:42:18.880 --> 0:42:23.400
<v Speaker 6>sales calls. We record those calls and they get transcribed,

0:42:23.440 --> 0:42:26.440
<v Speaker 6>and the AI then looks through them and makes decisions

0:42:26.480 --> 0:42:30.440
<v Speaker 6>about where this deal should be in the process. And

0:42:31.400 --> 0:42:34.080
<v Speaker 6>it's much better than having somebody try to go update it,

0:42:34.120 --> 0:42:36.600
<v Speaker 6>because those people never updated anywhere. The secret of all

0:42:36.640 --> 0:42:39.239
<v Speaker 6>of this enterprise software is that nobody was using it

0:42:39.239 --> 0:42:42.520
<v Speaker 6>the way that anybody wanted to anyway. And so you know,

0:42:43.120 --> 0:42:46.680
<v Speaker 6>I think that that is sort of you know, a

0:42:46.680 --> 0:42:48.359
<v Speaker 6>lot of what's happening there. Again, it's sort of some

0:42:48.400 --> 0:42:50.640
<v Speaker 6>of it's the coding, some of it's just the core capabilities.

0:42:50.680 --> 0:42:52.759
<v Speaker 6>And then you know, you still need databases, right, so

0:42:52.800 --> 0:42:54.320
<v Speaker 6>it's like, you know, you look at what data bricks

0:42:54.320 --> 0:42:56.400
<v Speaker 6>and stuff like. You know, I think those folks are

0:42:56.440 --> 0:42:58.840
<v Speaker 6>still sort of genuinely sitting in a pretty good place

0:42:58.880 --> 0:43:01.919
<v Speaker 6>where you know, all software has to sit on side

0:43:02.320 --> 0:43:05.000
<v Speaker 6>on top of some database that you can sort of

0:43:05.040 --> 0:43:08.080
<v Speaker 6>read and write to. But you know, I think some

0:43:08.120 --> 0:43:13.239
<v Speaker 6>of those categories that were specifically focused on kind of

0:43:13.320 --> 0:43:16.799
<v Speaker 6>like human input. Now, of course, you know, Salesforce has

0:43:16.800 --> 0:43:18.880
<v Speaker 6>a whole AI thing and they're saying, hey, you shouldn't

0:43:18.880 --> 0:43:22.320
<v Speaker 6>have humans inputting in salesforce. You know, at sales is

0:43:22.400 --> 0:43:24.799
<v Speaker 6>just one small piece. They have a whole customer support thing,

0:43:25.080 --> 0:43:28.680
<v Speaker 6>which obviously also has an interesting implication where you know

0:43:28.680 --> 0:43:31.160
<v Speaker 6>you're doing support with AI agents and so some of

0:43:31.200 --> 0:43:32.880
<v Speaker 6>it comes back to seats. I mean, you know, it

0:43:32.920 --> 0:43:36.000
<v Speaker 6>gets to be fairly complicated, but I do think I

0:43:36.000 --> 0:43:40.280
<v Speaker 6>think the fundamental underlying thing is anybody who buys software

0:43:41.080 --> 0:43:44.920
<v Speaker 6>that is, you know, uh SaaS, you're always buying for

0:43:44.960 --> 0:43:48.480
<v Speaker 6>a subset of the functionality that's nobody is using one

0:43:48.520 --> 0:43:50.839
<v Speaker 6>hundred percent of the functionality of SaaS, and so there's

0:43:50.840 --> 0:43:52.880
<v Speaker 6>always a trade off that's happening there where you know

0:43:53.000 --> 0:43:55.399
<v Speaker 6>you're spending more money than you need to because you're

0:43:55.440 --> 0:43:58.040
<v Speaker 6>not using all of these pieces. And so you know,

0:43:58.160 --> 0:44:00.600
<v Speaker 6>if you can more narrowly focus that, that is where

0:44:00.680 --> 0:44:02.560
<v Speaker 6>you could say, hey, we could solve this kind of

0:44:02.560 --> 0:44:04.400
<v Speaker 6>more narrow problem. And not only can we solve it

0:44:04.400 --> 0:44:07.320
<v Speaker 6>more narrowly, we can solve it way more effectively because

0:44:07.640 --> 0:44:10.080
<v Speaker 6>you know, the trick with AI is that the more

0:44:10.520 --> 0:44:13.520
<v Speaker 6>specific you are with it, the better the output is. Right,

0:44:13.560 --> 0:44:15.879
<v Speaker 6>So it's like, if you know, if outside of coding,

0:44:15.920 --> 0:44:18.480
<v Speaker 6>if you just ask chet gpz to write you a story.

0:44:18.600 --> 0:44:21.680
<v Speaker 6>It's going to write you a very very median story, right,

0:44:22.200 --> 0:44:25.600
<v Speaker 6>sort of exactly the median. But if you work with

0:44:25.640 --> 0:44:28.439
<v Speaker 6>it and you you know, then you're going to get it.

0:44:28.560 --> 0:44:31.640
<v Speaker 6>The more of your own expertise, you imbuing it, the

0:44:32.280 --> 0:44:34.319
<v Speaker 6>further up above the media, and it's going to be

0:44:34.360 --> 0:44:35.840
<v Speaker 6>and it's going to be you know. Of course that

0:44:35.920 --> 0:44:39.399
<v Speaker 6>also means it's less where the line is between what's

0:44:39.440 --> 0:44:41.640
<v Speaker 6>AI and what's not AI is going to continue to

0:44:41.640 --> 0:44:43.759
<v Speaker 6>get lorier, Joe, how.

0:44:43.719 --> 0:44:45.760
<v Speaker 3>Much does Claude Code actually cost?

0:44:46.280 --> 0:44:46.640
<v Speaker 4>Do you know?

0:44:47.280 --> 0:44:47.640
<v Speaker 5>Well?

0:44:48.600 --> 0:44:51.920
<v Speaker 2>I paid for the two two hundred dollars a month version,

0:44:52.360 --> 0:44:56.080
<v Speaker 2>but like high roller, yeah, I know, but uh, you know,

0:44:56.160 --> 0:44:57.640
<v Speaker 2>I think it's you can get it with the pro

0:44:57.760 --> 0:44:59.920
<v Speaker 2>version of like or whatever. The sub version of that

0:45:00.040 --> 0:45:02.200
<v Speaker 2>blow twenty dollars. But I hit a limit fairly quickly,

0:45:02.280 --> 0:45:03.719
<v Speaker 2>and I was like, I didn't have my website up

0:45:03.800 --> 0:45:06.600
<v Speaker 2>so like, and then I bought the five. Then I

0:45:06.640 --> 0:45:09.520
<v Speaker 2>paid five dollars for the extra compute, and I was like,

0:45:09.560 --> 0:45:10.160
<v Speaker 2>this is dumb.

0:45:10.239 --> 0:45:13.640
<v Speaker 3>I think, yeah, okay, so we're going out to.

0:45:13.800 --> 0:45:16.560
<v Speaker 2>Two nice dinners right month, that's not you know, when

0:45:16.560 --> 0:45:18.080
<v Speaker 2>I think about that way, it doesn't see that big.

0:45:17.920 --> 0:45:19.680
<v Speaker 3>Of a deal, it's worth it to you. Yeah, okay,

0:45:19.719 --> 0:45:21.600
<v Speaker 3>so I think we can all agree this is like

0:45:21.680 --> 0:45:26.319
<v Speaker 3>a valuable service that claud Code is providing. But we

0:45:26.400 --> 0:45:29.080
<v Speaker 3>touched on this in the intro. It seems like the

0:45:29.160 --> 0:45:33.320
<v Speaker 3>models just keep replicating themselves really really quickly. So anything

0:45:33.320 --> 0:45:36.520
<v Speaker 3>that claud Code can do, I would expect another model

0:45:36.560 --> 0:45:39.120
<v Speaker 3>will come in in like a month maybe less and

0:45:39.160 --> 0:45:41.880
<v Speaker 3>do the exact same thing. What does that mean for

0:45:41.960 --> 0:45:46.279
<v Speaker 3>the actual like valuations of these companies and the models, Like,

0:45:46.360 --> 0:45:48.360
<v Speaker 3>how are they going to monetize it when it seems

0:45:48.440 --> 0:45:53.239
<v Speaker 3>so difficult to actually differentiate yourself, especially for like a

0:45:53.280 --> 0:45:55.719
<v Speaker 3>substantial portion of time.

0:45:55.800 --> 0:45:58.279
<v Speaker 6>Yeah, well, so again here I think we have to

0:45:58.440 --> 0:46:02.480
<v Speaker 6>distinguish between claud Code and the Claude model. So in

0:46:02.760 --> 0:46:06.040
<v Speaker 6>claud Code's case, if you're using you know, the latest version,

0:46:06.040 --> 0:46:08.680
<v Speaker 6>you're using Opus four point five, which is the model.

0:46:08.760 --> 0:46:11.640
<v Speaker 6>Opus four point five has a price of something in

0:46:11.680 --> 0:46:14.480
<v Speaker 6>the dollar fifty to two dollars for a million input

0:46:14.560 --> 0:46:16.880
<v Speaker 6>tokens and whatever it is on the output, which is

0:46:16.920 --> 0:46:19.640
<v Speaker 6>like roughly the going rate for cutting edge modles. Gemini

0:46:19.800 --> 0:46:25.160
<v Speaker 6>three pro is the same price open a Chachipt five

0:46:25.200 --> 0:46:27.560
<v Speaker 6>point two is. They're all the same price. So the

0:46:27.560 --> 0:46:29.759
<v Speaker 6>first thing is is you have to differentiate between those.

0:46:30.760 --> 0:46:32.960
<v Speaker 6>And so I think a big part of what Anthropic

0:46:33.040 --> 0:46:34.880
<v Speaker 6>is trying to do is they're trying to lock people

0:46:34.880 --> 0:46:39.200
<v Speaker 6>into claud Code. In fact, there's just some controversy amongst

0:46:39.480 --> 0:46:43.560
<v Speaker 6>some nerds where open Code, which is a competitor to

0:46:43.640 --> 0:46:47.879
<v Speaker 6>claud Code, used to let you use your Claude Max

0:46:47.880 --> 0:46:50.120
<v Speaker 6>two hundred dollars. So the trick with the Claude Max

0:46:50.160 --> 0:46:53.320
<v Speaker 6>plan is if you're just buying those that number of tokens,

0:46:53.320 --> 0:46:56.000
<v Speaker 6>it would cost you significantly more than two hundred dollars.

0:46:56.040 --> 0:47:00.400
<v Speaker 6>It is a super super discounted plan. So like you,

0:47:00.400 --> 0:47:03.560
<v Speaker 6>you are probably you have the access, I have the

0:47:03.600 --> 0:47:06.239
<v Speaker 6>access to use I would guess in the thousand or

0:47:06.239 --> 0:47:10.640
<v Speaker 6>two thousand dollars of tokens for my two hundred dollars

0:47:10.640 --> 0:47:13.359
<v Speaker 6>a month. So it's a very very heavily subsidized plan.

0:47:13.640 --> 0:47:15.840
<v Speaker 6>And open Code, which is an open source version of

0:47:16.000 --> 0:47:19.920
<v Speaker 6>Claude code, a sort of competitor. They had found a

0:47:19.920 --> 0:47:22.600
<v Speaker 6>way that they would let you use your Claude Max

0:47:22.640 --> 0:47:27.680
<v Speaker 6>plan with open code, and Anthropic last week shut that down,

0:47:28.360 --> 0:47:31.520
<v Speaker 6>and some open Code people got very upset because they said,

0:47:31.560 --> 0:47:34.160
<v Speaker 6>like this is not what you're supposed to do or

0:47:34.520 --> 0:47:37.400
<v Speaker 6>I'm not sure exactly what they said. I never felt

0:47:37.440 --> 0:47:39.400
<v Speaker 6>like I got a particularly good argument out of it.

0:47:40.280 --> 0:47:42.600
<v Speaker 6>But you know, I do think part of what they're

0:47:42.600 --> 0:47:47.040
<v Speaker 6>trying to get at because is that, you know, at

0:47:47.040 --> 0:47:49.839
<v Speaker 6>the very top, models like these are all amazing, like

0:47:50.040 --> 0:47:56.040
<v Speaker 6>the Google Opening and Anthropic, their best models are all

0:47:56.080 --> 0:47:58.319
<v Speaker 6>on par with each other. I mean, I would move

0:47:58.320 --> 0:48:00.920
<v Speaker 6>them around a little bit. I still think Opus four

0:48:01.000 --> 0:48:03.120
<v Speaker 6>point five is the best model out there, but you know,

0:48:03.160 --> 0:48:08.439
<v Speaker 6>I mean that might change tomorrow. Like and that's where

0:48:08.480 --> 0:48:10.560
<v Speaker 6>something like cloud code is really interesting because it's a

0:48:11.200 --> 0:48:14.080
<v Speaker 6>product that is very it's just theirs. It's a piece

0:48:14.080 --> 0:48:17.120
<v Speaker 6>of software. It's not an AI model, and so it's

0:48:17.160 --> 0:48:21.000
<v Speaker 6>sort of it's less able to be disrupted. Now again,

0:48:21.120 --> 0:48:24.080
<v Speaker 6>I think if if somebody else wanted to copy that exactly,

0:48:24.800 --> 0:48:28.319
<v Speaker 6>they could. Codex has one, Gemini has one. I just

0:48:28.320 --> 0:48:30.439
<v Speaker 6>think they take a very different tact with it where

0:48:30.440 --> 0:48:32.360
<v Speaker 6>it's much less and so, you know, I think what

0:48:32.360 --> 0:48:34.120
<v Speaker 6>they're trying to do is get developers like me to

0:48:34.160 --> 0:48:36.440
<v Speaker 6>feel very comfortable inside that so that when we go

0:48:36.560 --> 0:48:39.640
<v Speaker 6>open I still open Codex or tried Gemini, or I

0:48:39.719 --> 0:48:42.840
<v Speaker 6>was playing with open code the other day, and it

0:48:42.960 --> 0:48:44.960
<v Speaker 6>just doesn't feel familiar in the same way that you know,

0:48:45.000 --> 0:48:47.120
<v Speaker 6>if you're trying to move somebody from a PC to

0:48:47.160 --> 0:48:48.480
<v Speaker 6>a MAC, it doesn't feel familiar. Right.

0:48:48.480 --> 0:48:50.680
<v Speaker 3>They want to own like the ecosystem.

0:48:50.080 --> 0:48:54.160
<v Speaker 2>The environment, that environment or what a world. Noah, thank

0:48:54.200 --> 0:48:56.200
<v Speaker 2>you so much for coming on odd Laws. I was

0:48:56.239 --> 0:48:58.560
<v Speaker 2>like dying to do an episode about this topic.

0:48:58.600 --> 0:48:59.359
<v Speaker 5>Thanks for having me.

0:49:00.080 --> 0:49:03.160
<v Speaker 2>Way, I don't have AI psychosis. I have a claud complex.

0:49:03.600 --> 0:49:05.239
<v Speaker 3>Why is everyone making that joke?

0:49:05.360 --> 0:49:06.439
<v Speaker 2>Wait, which joke?

0:49:07.000 --> 0:49:08.080
<v Speaker 3>The psychosis joke?

0:49:08.600 --> 0:49:10.160
<v Speaker 2>I think you were going to be proud of me

0:49:10.200 --> 0:49:11.480
<v Speaker 2>for saying Claude complex.

0:49:11.760 --> 0:49:13.040
<v Speaker 3>Oh oh that is very good.

0:49:13.680 --> 0:49:16.400
<v Speaker 2>I do one pun finally for Tracy And why was

0:49:16.400 --> 0:49:17.239
<v Speaker 2>it for making that joke?

0:49:17.320 --> 0:49:18.640
<v Speaker 5>Well, I was a joke.

0:49:18.840 --> 0:49:20.759
<v Speaker 2>I was handing your sirt. I finally make a pun,

0:49:20.880 --> 0:49:21.879
<v Speaker 2>and you just jump right over.

0:49:22.160 --> 0:49:25.520
<v Speaker 3>Everyone keeps saying that claud code is AI psychosis for

0:49:25.600 --> 0:49:28.960
<v Speaker 3>smart people, right, Like, how did that beck come a thing? Yeah?

0:49:29.080 --> 0:49:30.439
<v Speaker 2>All right, but there's a good pun.

0:49:30.560 --> 0:49:33.160
<v Speaker 3>Also very pro coded. I find you think so all

0:49:33.160 --> 0:49:34.400
<v Speaker 3>of AI is pro coded.

0:49:34.880 --> 0:49:37.279
<v Speaker 2>Uh, this is true. We should talk more about this,

0:49:37.320 --> 0:49:39.120
<v Speaker 2>you know, we should have David Shorn he's been doing

0:49:39.120 --> 0:49:41.319
<v Speaker 2>a lot of polling about various demographics and how they

0:49:41.320 --> 0:49:45.319
<v Speaker 2>feel about AI. We should need some interesting that. Yeah,

0:49:45.320 --> 0:49:47.719
<v Speaker 2>we should do that anyway. No, thank you so much

0:49:47.719 --> 0:49:48.480
<v Speaker 2>for coming on offline.

0:49:48.520 --> 0:49:49.319
<v Speaker 5>Thanks for having me.

0:50:01.480 --> 0:50:03.960
<v Speaker 2>Well that was fun, Tracy. I really like I It's

0:50:04.200 --> 0:50:07.399
<v Speaker 2>obvious to anyone who's been within five minutes five feet

0:50:07.440 --> 0:50:09.320
<v Speaker 2>of me for the last two weeks. I'm like totally

0:50:09.400 --> 0:50:12.720
<v Speaker 2>addicted and gone down, I know, gone down the rabbit

0:50:12.760 --> 0:50:16.840
<v Speaker 2>hole and stuff. But like I for the first time, unironically,

0:50:17.000 --> 0:50:21.160
<v Speaker 2>I'm like, Okay, this is transformative technology beyond being very

0:50:21.160 --> 0:50:22.120
<v Speaker 2>impressive technology.

0:50:22.200 --> 0:50:26.000
<v Speaker 3>Right, So I've been coming to a conclusion, which is

0:50:26.040 --> 0:50:31.040
<v Speaker 3>that you know, AI can be both under hyped and

0:50:31.160 --> 0:50:34.920
<v Speaker 3>overvalued simultaneously. Like and I feel like that's kind of

0:50:34.920 --> 0:50:36.480
<v Speaker 3>where we are at the moment.

0:50:36.239 --> 0:50:37.560
<v Speaker 2>That were you're making your stock call.

0:50:37.840 --> 0:50:42.040
<v Speaker 3>Yeah no, but seriously, like it it's a big deal.

0:50:42.360 --> 0:50:45.400
<v Speaker 3>It's going to change the way we work, But is

0:50:45.400 --> 0:50:48.839
<v Speaker 3>it monetizable? Can you differentiate the actual models? The better

0:50:48.920 --> 0:50:52.440
<v Speaker 3>the technology gets, like, the easier it is to just

0:50:52.480 --> 0:50:54.960
<v Speaker 3>do what everyone else is doing. And also like the

0:50:55.040 --> 0:50:57.520
<v Speaker 3>compute gets cheaper and cheaper. So I just don't know

0:50:57.560 --> 0:50:58.879
<v Speaker 3>how you monetize this.

0:50:59.040 --> 0:51:01.799
<v Speaker 2>Well, so that's very interesting his point, which is that

0:51:01.840 --> 0:51:05.560
<v Speaker 2>it's the tokens are heavily subsidized still, and so that

0:51:05.560 --> 0:51:08.080
<v Speaker 2>if you're paying and actually using that two hundred dollars

0:51:08.080 --> 0:51:10.880
<v Speaker 2>max program and you actually use it to the limit,

0:51:11.400 --> 0:51:14.640
<v Speaker 2>Claude is going to lose money on this and then

0:51:14.680 --> 0:51:18.040
<v Speaker 2>the prices keep dropping. And I know, like claud code

0:51:18.200 --> 0:51:21.400
<v Speaker 2>is okay, they're attempting to create something that resembles a

0:51:21.440 --> 0:51:24.400
<v Speaker 2>traditional software ecosystem that you feel as a user that

0:51:24.440 --> 0:51:28.719
<v Speaker 2>you're locked into. But so far, in my various like

0:51:29.000 --> 0:51:32.200
<v Speaker 2>since November twenty twenty two when I started playing with AI,

0:51:32.680 --> 0:51:35.920
<v Speaker 2>it hasn't felt like anyone has established lock in with anything.

0:51:36.160 --> 0:51:41.120
<v Speaker 2>And it's very it's very movable, and I suspect even

0:51:41.160 --> 0:51:43.319
<v Speaker 2>though I have this file now on my desktop that

0:51:43.400 --> 0:51:47.400
<v Speaker 2>has a file called claud md that gives instructions, et cetera,

0:51:47.800 --> 0:51:50.320
<v Speaker 2>I'm certain that if I open this file with Codex

0:51:50.400 --> 0:51:52.800
<v Speaker 2>or Googles, I could probably just pick it up the same.

0:51:52.920 --> 0:51:55.640
<v Speaker 3>Yeah. I also think there's a fundamental issue with the

0:51:55.640 --> 0:51:58.840
<v Speaker 3>lock in strategy because when you're talking about technology on

0:51:58.880 --> 0:52:02.719
<v Speaker 3>the Internet, it just feels very against the grain to

0:52:02.920 --> 0:52:06.600
<v Speaker 3>try to lock people into anything, and we've seen various

0:52:06.600 --> 0:52:10.319
<v Speaker 3>projects over the years, and it's a lot harder than

0:52:10.320 --> 0:52:10.840
<v Speaker 3>it looks.

0:52:11.840 --> 0:52:13.840
<v Speaker 2>Yeah, I mean, I guess I would say it's a

0:52:13.840 --> 0:52:15.600
<v Speaker 2>lot harder than it looks. But then we also know

0:52:15.640 --> 0:52:17.680
<v Speaker 2>the flip side, which is that tons of people are

0:52:17.719 --> 0:52:19.920
<v Speaker 2>locked into software that they hate, right, Yeah, people are

0:52:20.200 --> 0:52:22.120
<v Speaker 2>Oh I hate people. How many times have you Oh,

0:52:22.320 --> 0:52:26.080
<v Speaker 2>I hate Outlook right? Or I hate Microsoft teams and

0:52:26.160 --> 0:52:28.040
<v Speaker 2>I hate this and I spend money on it every

0:52:28.040 --> 0:52:30.239
<v Speaker 2>month and my organization can't move off of it, or

0:52:30.239 --> 0:52:32.200
<v Speaker 2>we can't migrate off of it. So I do think

0:52:32.239 --> 0:52:35.160
<v Speaker 2>that cuts both ways. I do think he offered the

0:52:35.200 --> 0:52:40.520
<v Speaker 2>best explanation I've heard of why the AI coding models

0:52:40.960 --> 0:52:43.840
<v Speaker 2>are a threat to a lot of pretty big software businesses,

0:52:44.160 --> 0:52:47.680
<v Speaker 2>especially especially the point about how the user never uses

0:52:47.760 --> 0:52:50.680
<v Speaker 2>all of the features that they actually that the software

0:52:50.680 --> 0:52:53.759
<v Speaker 2>got built for, and therefore maybe the build versus by

0:52:53.920 --> 0:52:57.560
<v Speaker 2>calculation really starts to shift when they can just design

0:52:57.600 --> 0:52:58.920
<v Speaker 2>that one feature very quickly.

0:52:59.560 --> 0:53:02.200
<v Speaker 3>I totally agree on the software side, it seems like

0:53:02.239 --> 0:53:05.720
<v Speaker 3>an existential threat, but just like the locked in ecosystem

0:53:05.960 --> 0:53:09.480
<v Speaker 3>of a particular model. I know he said it's not

0:53:09.520 --> 0:53:12.360
<v Speaker 3>actually a model, but that seems like a bigger issue

0:53:12.360 --> 0:53:14.120
<v Speaker 3>to me. I don't know. I guess we'll see.

0:53:14.280 --> 0:53:15.799
<v Speaker 2>We're going to see and I don't know. I kind

0:53:15.840 --> 0:53:16.880
<v Speaker 2>of think we're going to see quickly.

0:53:17.200 --> 0:53:20.680
<v Speaker 3>Yeah, that's again, that's the only certainty is like stuff

0:53:20.719 --> 0:53:21.160
<v Speaker 3>is happening.

0:53:21.360 --> 0:53:22.120
<v Speaker 2>What is happening now?

0:53:22.239 --> 0:53:22.439
<v Speaker 4>Yeah?

0:53:22.520 --> 0:53:23.520
<v Speaker 3>Okay, shall we leave it there?

0:53:23.600 --> 0:53:24.359
<v Speaker 2>Let's leave it there.

0:53:24.480 --> 0:53:26.760
<v Speaker 3>This has been another episode of the Odd Lots podcast.

0:53:26.840 --> 0:53:30.040
<v Speaker 3>I'm Tracy Alloway. You can follow me at Tracy Alloway.

0:53:29.680 --> 0:53:31.799
<v Speaker 2>And I'm Jolly Isn't all. You can follow me at

0:53:31.840 --> 0:53:34.719
<v Speaker 2>the Stalwart. Follow our guest Noah Brier, He's at Hey,

0:53:34.760 --> 0:53:38.600
<v Speaker 2>It's Noah. Follow our producers Kerman, Rodriguez at Carman Arman, Dashill,

0:53:38.600 --> 0:53:41.400
<v Speaker 2>ben Att at Dashbot, and Kilbrooks at Kilbrooks. And for

0:53:41.440 --> 0:53:43.960
<v Speaker 2>more odd Laws content, go to Bloomberg dot com slash

0:53:43.960 --> 0:53:46.640
<v Speaker 2>odd Lots with a daily newsletter and all of our episodes,

0:53:46.800 --> 0:53:48.680
<v Speaker 2>and you can chat about all of these topics twenty

0:53:48.719 --> 0:53:51.800
<v Speaker 2>four to seven in our discord Discord dot gg slash

0:53:51.800 --> 0:53:52.280
<v Speaker 2>odd Lots.

0:53:52.520 --> 0:53:54.560
<v Speaker 3>And if you enjoy odd Lots, if you like it

0:53:54.600 --> 0:53:57.719
<v Speaker 3>when we talk about advances in AI, then please leave

0:53:57.800 --> 0:54:01.400
<v Speaker 3>us a positive review on your favorite podcast platform and remember,

0:54:01.440 --> 0:54:03.960
<v Speaker 3>if you are a Bloomberg subscriber, you can listen to

0:54:04.000 --> 0:54:07.080
<v Speaker 3>all of our episodes absolutely ad free. All you need

0:54:07.120 --> 0:54:09.800
<v Speaker 3>to do is find the Bloomberg channel on Apple Podcast

0:54:09.880 --> 0:54:28.400
<v Speaker 3>and follow the instructions there. Thanks for listening.