1
00:00:03,200 --> 00:00:07,560
Speaker 1: Media. Hello and welcome to this week's Better Offline Monologue.

2
00:00:07,600 --> 00:00:17,880
Speaker 1: I'm your host ed ze trn SO. In the last

3
00:00:17,880 --> 00:00:20,480
Speaker 1: few weeks, there's been a dramatic sell off in software stocks,

4
00:00:20,560 --> 00:00:23,400
Speaker 1: driven by the anxiety that companies will, instead of paying

5
00:00:23,440 --> 00:00:27,200
Speaker 1: for someone like Salesforce or Microsoft, simply build their own software.

6
00:00:27,880 --> 00:00:31,200
Speaker 1: It is a genuinely stupid assumption based on analysts and

7
00:00:31,240 --> 00:00:34,159
Speaker 1: reporters that simply do not care about the truth. In

8
00:00:34,200 --> 00:00:37,599
Speaker 1: their mind, one can simply type build me Salesforce now

9
00:00:37,600 --> 00:00:40,080
Speaker 1: into claud code and have it barf out an identical

10
00:00:40,159 --> 00:00:44,360
Speaker 1: functional clone that's compliance secure, stable, or because somebody was

11
00:00:44,400 --> 00:00:46,040
Speaker 1: able to bunk it on the head enough times to

12
00:00:46,040 --> 00:00:49,000
Speaker 1: spit out something that sort of looked like a tool

13
00:00:49,080 --> 00:00:53,239
Speaker 1: like Trello maybe, or a personal website. Look, when you

14
00:00:53,320 --> 00:00:56,560
Speaker 1: pay a software company, even a dogshit mediocre one, a

15
00:00:56,600 --> 00:00:59,680
Speaker 1: monthly fee, you're not just paying them to access the software,

16
00:00:59,720 --> 00:01:01,880
Speaker 1: but to take away the burden of maintenance that comes

17
00:01:01,960 --> 00:01:05,800
Speaker 1: with running a software company. Minor things like currency changes

18
00:01:05,959 --> 00:01:09,480
Speaker 1: or time zone shifting can cause major problems, and systems

19
00:01:09,520 --> 00:01:12,680
Speaker 1: that aren't built with intention, you know, like something in LLM,

20
00:01:12,720 --> 00:01:15,560
Speaker 1: would spit out and things get even more complicated when

21
00:01:15,560 --> 00:01:18,520
Speaker 1: you start connecting other systems of record like billing or

22
00:01:18,520 --> 00:01:21,760
Speaker 1: a customer's personal information, especially if they're in Europe. By

23
00:01:21,800 --> 00:01:24,480
Speaker 1: the way, plan to have any of that information actually

24
00:01:24,520 --> 00:01:27,440
Speaker 1: connect with a customers systems or you're gonna need a

25
00:01:27,520 --> 00:01:29,440
Speaker 1: sock two order and you're definitely going to need to

26
00:01:29,440 --> 00:01:31,800
Speaker 1: make sure it's got rock solid security so that nobody

27
00:01:31,840 --> 00:01:34,240
Speaker 1: can swipe all of that data and then you get sued.

28
00:01:34,640 --> 00:01:37,200
Speaker 1: I also assume you're going to effectively take an engineer

29
00:01:37,240 --> 00:01:40,200
Speaker 1: off of one of your team's probably for good, to

30
00:01:40,240 --> 00:01:44,280
Speaker 1: be honest, to maintain your new internal salesforce Monday Microsoft

31
00:01:44,280 --> 00:01:47,760
Speaker 1: three sixty five and Notion clones you Trello cu clone

32
00:01:47,760 --> 00:01:50,240
Speaker 1: as well. Probably you're a sana What I mean, how

33
00:01:50,320 --> 00:01:53,040
Speaker 1: much software are you gonna build? Good? Thing? You've got

34
00:01:53,120 --> 00:01:55,680
Speaker 1: clawed code to help speed that up, right, Just make

35
00:01:55,680 --> 00:01:58,400
Speaker 1: sure you read everything it writes, because every little fuck

36
00:01:58,480 --> 00:02:01,320
Speaker 1: up just became your problem, and you've got nobody to

37
00:02:01,400 --> 00:02:04,360
Speaker 1: scream at, because while your company is saving fifteen dollars

38
00:02:04,360 --> 00:02:07,080
Speaker 1: per seep per month, you've also fired the people whose

39
00:02:07,120 --> 00:02:09,320
Speaker 1: job it is to make sure your nasty little software

40
00:02:09,320 --> 00:02:13,120
Speaker 1: subscriptions actually fucking function. And while you may fear that

41
00:02:13,160 --> 00:02:15,079
Speaker 1: a boss might try and force this down your throat

42
00:02:15,120 --> 00:02:17,600
Speaker 1: to save money. I must be clear how impossible this

43
00:02:17,720 --> 00:02:21,120
Speaker 1: task is. Even the most annoying, frustrating suffer as a

44
00:02:21,160 --> 00:02:24,880
Speaker 1: service contract protects you from the grueling, underlying maintenance and

45
00:02:24,919 --> 00:02:28,000
Speaker 1: infrastructural bullshit that goes into making sure the thing you're

46
00:02:28,040 --> 00:02:30,920
Speaker 1: paying for actually loads and functions wherever you load in

47
00:02:31,320 --> 00:02:33,200
Speaker 1: and even on the browser that you want to load

48
00:02:33,200 --> 00:02:36,760
Speaker 1: it in. In most cases, thank you riverside. The people

49
00:02:36,800 --> 00:02:39,880
Speaker 1: pushing this narrative are either fundamentally disconnected from how the

50
00:02:39,919 --> 00:02:44,800
Speaker 1: world works or actively incentivized to mislead you. I've seen

51
00:02:44,880 --> 00:02:49,120
Speaker 1: this narrative propagated on multiple different business television shows and

52
00:02:49,160 --> 00:02:53,120
Speaker 1: supposedly respectable outlets, and it makes me genuinely worried that

53
00:02:53,160 --> 00:02:56,960
Speaker 1: we don't have a media industry prepared to dissect fundamentally

54
00:02:56,960 --> 00:03:01,160
Speaker 1: deceptive narratives. Just because it's possible for a non CODA

55
00:03:01,200 --> 00:03:04,200
Speaker 1: to cobble together a website that looks near identical to

56
00:03:04,240 --> 00:03:07,120
Speaker 1: a model's training data in the space of an hour

57
00:03:07,440 --> 00:03:10,920
Speaker 1: doesn't mean that we're replacing every software company, nor is

58
00:03:10,960 --> 00:03:14,239
Speaker 1: its ability to do so. Any indicator that will be

59
00:03:14,280 --> 00:03:17,280
Speaker 1: able to do more than that in the future. I

60
00:03:17,440 --> 00:03:20,320
Speaker 1: plead with the media, please stop filling in the gaps.

61
00:03:20,639 --> 00:03:24,480
Speaker 1: Please stop seeing every incremental improvement as proof of whatever

62
00:03:24,600 --> 00:03:28,000
Speaker 1: marketing slot Warrio Ama Day or Clammy Sam Altman is

63
00:03:28,040 --> 00:03:31,520
Speaker 1: trying to cram down your throats. You're being played, You're

64
00:03:31,560 --> 00:03:35,200
Speaker 1: being conned, and by extension, you're conning your listeners, your readers,

65
00:03:35,240 --> 00:03:37,960
Speaker 1: and your viewers. And once the bubble pops, I believe

66
00:03:38,000 --> 00:03:41,840
Speaker 1: they will demand an explanation from you. I certainly will,

67
00:03:42,600 --> 00:03:46,280
Speaker 1: and really, in this era, I think people are under

68
00:03:46,440 --> 00:03:49,080
Speaker 1: selling how big a reckoning there will be when the

69
00:03:49,120 --> 00:03:53,040
Speaker 1: bubble collapses. How are we meant to trust anybody who

70
00:03:53,360 --> 00:03:57,640
Speaker 1: vociferously pushed this or who got obsessed with AI once

71
00:03:57,680 --> 00:03:59,880
Speaker 1: all of this falls apart. When I say falls apart,

72
00:04:00,200 --> 00:04:03,000
Speaker 1: I mean that the current rates that you are paying

73
00:04:03,280 --> 00:04:06,520
Speaker 1: are not even close to being sustainable. I'm currently working

74
00:04:06,600 --> 00:04:08,400
Speaker 1: on a piece called the Hater's Guide to Anthropic on

75
00:04:08,400 --> 00:04:12,120
Speaker 1: my premium newsletter, and I found a mathematical study that

76
00:04:12,200 --> 00:04:16,240
Speaker 1: found that on a twenty dollars a month CLAWED subscription,

77
00:04:16,680 --> 00:04:20,760
Speaker 1: you can spend over one hundred dollars of anthropics API tokens,

78
00:04:20,960 --> 00:04:23,600
Speaker 1: and on the two hundred dollars a month about two

79
00:04:23,720 --> 00:04:28,320
Speaker 1: three hundred to seven hundred dollars. You got to I

80
00:04:28,320 --> 00:04:31,279
Speaker 1: guess it takes money to lose money. But this is

81
00:04:31,520 --> 00:04:34,599
Speaker 1: the reality of the AI bubble. Everything you were using

82
00:04:34,680 --> 00:04:39,359
Speaker 1: is being subsidized. It's being subsidized by these companies in

83
00:04:39,400 --> 00:04:41,400
Speaker 1: the hopes that it kind of bakes into your life,

84
00:04:41,440 --> 00:04:44,800
Speaker 1: except it's not good enough to do that. The frenzied

85
00:04:44,920 --> 00:04:49,680
Speaker 1: media push under clawed Code existed to kind of beguile you,

86
00:04:50,000 --> 00:04:52,200
Speaker 1: to make you think that there's more happening than a

87
00:04:52,240 --> 00:04:54,000
Speaker 1: tool that might be able to kind of build a

88
00:04:54,000 --> 00:04:56,960
Speaker 1: website sort of or build you some half functional software

89
00:04:57,000 --> 00:04:59,719
Speaker 1: that maybe works sometimes, and when it comes to actually

90
00:04:59,720 --> 00:05:02,120
Speaker 1: building that into something you could sell your shit out

91
00:05:02,120 --> 00:05:04,360
Speaker 1: of luck. You actually do need to code. And even

92
00:05:04,560 --> 00:05:08,160
Speaker 1: the things that you vibe code weren't really secure or functional.

93
00:05:09,279 --> 00:05:13,039
Speaker 1: And even that mediocre web slop you're seeing spooched out

94
00:05:13,040 --> 00:05:17,240
Speaker 1: by claud code is heavily subsidized. If people had to

95
00:05:17,240 --> 00:05:20,120
Speaker 1: pay the real rates, those people you see jacking off

96
00:05:20,160 --> 00:05:22,640
Speaker 1: on Twitter about claud code, they'd be paying two hundred

97
00:05:22,680 --> 00:05:25,680
Speaker 1: plus dollars a day. Do you think they'd actually pay that?

98
00:05:26,040 --> 00:05:28,680
Speaker 1: Do you think that's actually happening because I'm sick and

99
00:05:28,720 --> 00:05:31,200
Speaker 1: fucking tired of hearing all the people rambling about claud

100
00:05:31,240 --> 00:05:34,400
Speaker 1: code writing all the goddamn code. I'm sick and tired

101
00:05:34,440 --> 00:05:36,440
Speaker 1: of it because I don't see what the end point is.

102
00:05:36,800 --> 00:05:40,160
Speaker 1: I don't hear an actual result from this. I think

103
00:05:40,160 --> 00:05:42,919
Speaker 1: people should be laid off. But if this was doing

104
00:05:42,960 --> 00:05:45,200
Speaker 1: the thing that they were saying it would that al

105
00:05:45,240 --> 00:05:47,960
Speaker 1: would be happening, and that would be definitively connected to AI,

106
00:05:48,200 --> 00:05:51,640
Speaker 1: or more software would be being shipped. That's actually good.

107
00:05:52,520 --> 00:05:56,680
Speaker 1: Just more software existing doesn't mean the software's functional or useful.

108
00:05:56,920 --> 00:05:58,960
Speaker 1: And if you look at VIBE coded apps, they all

109
00:05:59,040 --> 00:06:01,560
Speaker 1: kind of look the same. And that's the result of

110
00:06:01,240 --> 00:06:04,719
Speaker 1: the more using the same training data. And it's why

111
00:06:04,960 --> 00:06:07,919
Speaker 1: these things are not really good at building nuanced or

112
00:06:08,440 --> 00:06:11,400
Speaker 1: unique software, because that's not what they do. They copy

113
00:06:11,480 --> 00:06:14,040
Speaker 1: software that's already been written, and they do so with

114
00:06:14,120 --> 00:06:18,279
Speaker 1: no intention, no real plan, and nothing you can look

115
00:06:18,320 --> 00:06:20,560
Speaker 1: back on and say, oh, right, that's why they built

116
00:06:20,600 --> 00:06:23,280
Speaker 1: it this way. And perhaps that's not a problem as

117
00:06:23,320 --> 00:06:25,599
Speaker 1: you're a solo person building your own deadly little lap,

118
00:06:25,800 --> 00:06:28,120
Speaker 1: but expand that to one thousand people or ten thousand

119
00:06:28,200 --> 00:06:31,760
Speaker 1: people let's say five ten years passes, and let's take

120
00:06:31,800 --> 00:06:35,720
Speaker 1: one of these what I think are bullshit statements about how, oh,

121
00:06:35,839 --> 00:06:39,120
Speaker 1: ninety percent of our code's written by AI. Great, what

122
00:06:39,240 --> 00:06:41,080
Speaker 1: happens in three years or four years when you go

123
00:06:41,120 --> 00:06:43,120
Speaker 1: and look back and that, you go, shit, why did

124
00:06:43,120 --> 00:06:46,320
Speaker 1: we do that? Are the person left or the person

125
00:06:46,320 --> 00:06:48,800
Speaker 1: didn't leave. They've just been drinking heavily and they don't

126
00:06:48,800 --> 00:06:50,919
Speaker 1: really remember what they were prompting the model with. And

127
00:06:51,320 --> 00:06:53,440
Speaker 1: oh man, how do we actually see why they built

128
00:06:53,440 --> 00:06:56,599
Speaker 1: this or what the reason was. This becomes a huge

129
00:06:56,600 --> 00:06:59,040
Speaker 1: problem when things start breaking, when you have to make

130
00:06:59,080 --> 00:07:01,599
Speaker 1: things more efficient, and then people will say, oh, we

131
00:07:01,720 --> 00:07:05,240
Speaker 1: use AI to fix them. At some point, you're just

132
00:07:05,360 --> 00:07:08,679
Speaker 1: looking at a kind of a rat's nest of crap

133
00:07:09,560 --> 00:07:16,240
Speaker 1: of code written unintentionally spooched out meaningless. How do you

134
00:07:16,280 --> 00:07:19,560
Speaker 1: fix that long term? How do you build an efficient

135
00:07:20,000 --> 00:07:22,440
Speaker 1: software package out of that? How do you build software

136
00:07:22,480 --> 00:07:24,840
Speaker 1: that continues to run well into the future. Is the

137
00:07:24,880 --> 00:07:28,400
Speaker 1: idea that you just build AI on AI? God? No, anyway,

138
00:07:28,440 --> 00:07:30,840
Speaker 1: I think once the real costs are charge, people are

139
00:07:30,840 --> 00:07:34,119
Speaker 1: gonna drop this shit like a bad habit. I can't wait,

140
00:07:34,280 --> 00:07:36,000
Speaker 1: and I'm gonna be smug about it. I'm gonna be

141
00:07:36,000 --> 00:07:39,280
Speaker 1: annoying about it, not gonna lie. But I think based

142
00:07:39,280 --> 00:07:42,360
Speaker 1: on the YouTube comments, you're all gonna join me. Now.

143
00:07:42,440 --> 00:07:44,600
Speaker 1: Next week we've got more Hater season, and I'll be

144
00:07:44,640 --> 00:07:46,960
Speaker 1: back with one of these monologues. Maybe orerhaps I'll give

145
00:07:46,960 --> 00:07:50,000
Speaker 1: you multiple interviews, done some weird ones. Got Victorious Song

146
00:07:50,080 --> 00:07:53,320
Speaker 1: on wellness, Got Adam Becker on Epstein. I mean, got

147
00:07:53,360 --> 00:07:56,560
Speaker 1: all sorts of shit coming. Looking forward to bringing you more,

148
00:07:56,800 --> 00:07:57,680
Speaker 1: and I love you all