WEBVTT
1
00:00:00.580 --> 00:00:08.550
Like right now, this is the intro? No, no, no. No. Oh, God. This is the intro. [laughs] No, no, I'll start it, and then I'll, I'll, I'll set it up. Thank you. [laughs] It's like, intro to what?
2
00:00:09.400 --> 00:00:17.919
Today's episode is brought to you by our friends at Venngage. You spent three weeks on that report. Be honest about how many people actually read past the intro.
3
00:00:18.420 --> 00:00:24.960
Venngage AI is the fix, an AI infographic generator that turns the thing you wrote into the thing people actually look at.
4
00:00:25.720 --> 00:00:32.500
Think about the last thing you understood the second you saw it, a chart, a process diagram, one clean visual.
5
00:00:32.509 --> 00:00:42.520
We read pictures faster than paragraphs and hold onto them longer, which is why the idea you can't get across in three paragraphs usually lands in a single image. The catch has always been time.
6
00:00:42.940 --> 00:00:47.540
A good infographic meant a designer, a queue, and three rounds of feedback.
7
00:00:47.600 --> 00:00:59.500
Venngage AI handles the structure for you, smart layouts, clean hierarchy, your brand colors applied across every asset, so it all looks like it came from the same company, and everything's editable.
8
00:00:59.540 --> 00:01:08.540
Change any element down to the last detail. AI design that's actually yours. Take the report nobody finished reading and turn it into the thing that people share.
9
00:01:08.630 --> 00:01:20.340
And this month only, 30% off any Venngage plan with the code SUPERPATH30 at venngage.com. That's SUPERPATH30 at venngage.com. Hope you enjoy the episode.
10
00:01:21.900 --> 00:01:25.420
Alex here, and welcome back to another Jimmy episode of Content Briefly.
11
00:01:25.640 --> 00:01:33.780
Today, Superpath founder emeritus Jimmy Daly joins us to talk about his recent article for us, The Hardest Part About Content Hasn't Gotten Any Easier.
12
00:01:33.840 --> 00:01:43.100
Jimmy will be the first to admit that the current AI content process he's building at Miro is a ton of fun for him, but he's recently experienced some last-mile content fatigue.
13
00:01:43.160 --> 00:01:48.960
His question, if AI and the writing process only ever gets us at least 90% of the way there, is that enough?
14
00:01:49.160 --> 00:01:58.420
We've got the whole gang together today, including Eric, Ruth, and Rachel, to talk about where we stand on the process of using AI systems to help us write great things. Enjoy the episode.
15
00:01:58.960 --> 00:02:05.680
You know, all the episodes that we've done on AI have been, um, like I've just been so pumped about it, really excited.
16
00:02:05.720 --> 00:02:14.220
I think this is the first one where I, like, really hit a stage in my AI marketing journey where I became really frustrated, but led to this important realization.
17
00:02:14.609 --> 00:02:23.980
And so the article, it's titled The Hardest Part of Content Hasn't Gotten Any Easier, but I almost called it Writing Suffers When Humans Don't, 'cause that's the real idea of it.
18
00:02:24.120 --> 00:02:34.400
The kind of the backstory here is that, so I'm at Miro now. I'm part of the content team, and have spent a, quite a bit of this summer working on content creation systems in cloud code.
19
00:02:34.769 --> 00:02:40.700
And it's a lot of fun, some of the most fun work I've ever done. Maybe I'll do an AI show and tell to walk through, like, how it actually works.
20
00:02:40.800 --> 00:02:50.200
But the idea is that I give the system raw material, which is usually an interview transcript, plus whatever other, like, notes and decks and stuff that I can cobble together on the topic.
21
00:02:50.240 --> 00:02:56.860
That's sort of like step one, and then there's probably 10 different steps and three different human checkpoints.
22
00:02:56.890 --> 00:03:09.460
And the idea is it, it's supposed to kind of mirror a real writing process, except that, one, it's automated by AI, and two, I'm kind of the facilitator, not the actual writer.
23
00:03:09.500 --> 00:03:17.040
One of the things I've sort of learned while doing this is that it's a total abstraction of the type of work that I used to do, and I think many of us used to do.
24
00:03:17.140 --> 00:03:19.720
Instead of actually writing the article, I'm explaining how to write the article.
25
00:03:19.820 --> 00:03:29.710
In that process, there's this whole other thing that I've, like, actually really enjoyed, which is the opportunity to step back and try to articulate, I mean, I think I'm an okay writer, and try to articulate, like, why am I an okay writer?
26
00:03:29.720 --> 00:03:37.950
And what goes into it? And what are all the various steps? It's really hard. It's, like, really, [laughs] really hard to capture all of that and make the system work the way that I want it to work.
27
00:03:38.180 --> 00:03:46.880
So I've run this now hundreds of times. We've published a bunch of stuff too. Eventually, I've gotten quite a bit of it to a place where I felt it was worthy of publishing on social and the blog.
28
00:03:46.920 --> 00:03:54.780
But in doing this, this is kind of where the brewing sense of uncertainty has been coming from. Why is the whole less than the sum of the parts?
29
00:03:54.860 --> 00:04:03.160
Why can't I explain this to Claude in a way that will make the output better? Sometimes it just totally misses, and AI does that sometimes.
30
00:04:03.280 --> 00:04:15.460
Sometimes you end up with this, like, last mile problem, where you can get 80 or 90% of the way there and it feels amazing, but then you spend so much time trying to run that last mile that you, like, give all that time back up.
31
00:04:15.480 --> 00:04:29.909
The realization that I've come to, which is I, I'm sure is not unique among content marketers and certainly not among this group, is that it's too easy [laughs] and therefore the writing will never be as good as it could be if someone endured some misery to create it.
32
00:04:30.130 --> 00:04:39.080
[laughs] In order to create something good, you have to sit there with it, and there's a part of it that's, like, a little bit miserable and you suffer a little bit in, like, this creative sense.
33
00:04:39.400 --> 00:04:46.260
I'm very curious, you know, Rachel, Ruth, Eric, have you all gone through a similar journey? Have you come to accept the limitations of AI?
34
00:04:46.460 --> 00:04:58.500
Or are you more like me, where you're, like, spend so much time tweaking it, hoping you can finally make it work, and that will, like, unlock this giant wave of productivity before coming to this realization that, like, probably I'm actually a little bit late on?
35
00:04:58.600 --> 00:05:01.200
Eric, since we went in this order earlier- Yeah.... y'all said it too first.
36
00:05:01.960 --> 00:05:12.650
I, I pretty much, like, the week you wrote this piece, I was having an existential crisis about AI again, where I was like, I, I had so much hope and I feel it all crashing down.
37
00:05:12.900 --> 00:05:22.920
A- and for kind of precisely the reason you talked about, that I, I didn't feel like I was proud of things that were coming out, not because they felt like they were written by AI or they had AI-isms in them.
38
00:05:23.080 --> 00:05:33.140
I just wasn't feeling a connection with what I was making anymore. A lot of it was what you highlighted, because of the lack of suffering for writing the piece. I'll just give the Superpath newsletter as an example.
39
00:05:33.300 --> 00:05:42.960
I never use AI for that, and I think the whole reason it's good is because of how much day-to-day job suffering I do to, like, come up with the topics.
40
00:05:43.120 --> 00:05:53.020
I'm, like, w- writing it as fast as I can, stream of thought the morning I send it out, which produces all these weird things that I would never say if I was coaching an AI writer to write something.
41
00:05:53.190 --> 00:06:01.360
The conditions sort of produce the newsletter, and if I change the conditions too much, the quality wouldn't be there anymore. So like, I- Yeah.... I think it really resonated with me.
42
00:06:01.380 --> 00:06:09.600
That's interesting, because, like, in the case of the newsletter, the suffering does not occur while writing the newsletter. It occurs while you are gaining- A little bit of both. [laughs]... the experience that...
43
00:06:09.660 --> 00:06:18.853
Yeah, yeah. [laughs]But it's like you earned the right to create the newsletter by doing your job, right? And that is hard. And then that- Well, yeah. And then, like, I'm, I'm trying to, like, pull up-...
44
00:06:18.853 --> 00:06:26.664
creates the ideas you write about... I'm trying to pull from things to say, and I think of the weirdest specific examples and weird phrasings because it's sort of off the cuff.
45
00:06:26.764 --> 00:06:34.244
I don't think I could, like, coach an AI- Yeah. Yeah... editor or writer to come up with those same sort of, not just the anecdotes, but also the way it comes across.
46
00:06:34.304 --> 00:06:40.214
Not to say that I'm, like, a unique butterfly, but that it, it comes through in the writing because of how I'm going about writing it. And the writing- Yeah. Yeah...
47
00:06:40.214 --> 00:06:51.044
I do for Doc ends up a very different style because I'm having a different process. So Jimmy, when you were writing this, were you writing this only thinking about the actual content creation process?
48
00:06:51.604 --> 00:06:57.654
I am talking to Claude, and Claude is writing my blog for me. Because- Mm...
49
00:06:57.764 --> 00:07:10.524
I feel like as, as someone who has been in AI for a while, one thing that I have thought about for so long in the content process is I feel like I have now moved to, okay, where am I keeping things separate now?
50
00:07:10.534 --> 00:07:21.724
Where am I so focused on where the human stays in the loop? To your point, a piece of the process has to stay painful [laughs] for the content to be good.
51
00:07:21.784 --> 00:07:31.544
You know, I, I recently joined, um, an AI consulting firm, and I'm their director of marketing now. Part of our content process is we record a p- podcast twice a month.
52
00:07:32.104 --> 00:07:38.944
That podcast, it's still heavily manual, but we are using s- Claude's skills to get a lot of it done.
53
00:07:39.444 --> 00:07:51.044
I'm not gonna expect my content manager to repurpose a data science conversation on her own, but I do expect the data scientists to sit there and talk about these topics.
54
00:07:51.154 --> 00:08:02.784
And if they're gonna do it off the cuff, then that's gonna be the painful process for them. I am making sure that I'm still at least staying the human in one piece of it.
55
00:08:02.924 --> 00:08:14.244
I'm curious what part you were thinking about as you were writing this, or if you were specifically thinking about guiding Claude to write new content and doing that at scale as you were going through this. The latter.
56
00:08:14.504 --> 00:08:19.544
Like, my- Okay... my goal was, I'm not, like, totally giving up on this, but I'm definitely changing the approach.
57
00:08:19.744 --> 00:08:25.664
The goal was to go from I interview a thought leader to I have a thought leadership article that's ready to go.
58
00:08:25.724 --> 00:08:31.504
There is a significant amount of writing in that, but, like, like in my mind, writing is kinda like two distinct steps.
59
00:08:31.624 --> 00:08:41.074
One is understanding what you're trying to say, and then the second is putting the words in the right order so that it's nice for a reader to consume. My goal was to, was to do both.
60
00:08:41.284 --> 00:08:47.054
And there are separate workflows for, like, I guess what I would call marketing hygiene. You know, like, using AI to write show notes.
61
00:08:47.184 --> 00:08:58.104
All the things that, like, we all sort of kn- know we should do, but it's easy to not do, and now AI makes it super easy. That's, like, the best use case. But trying to get AI to do all of it is, um...
62
00:08:58.144 --> 00:09:08.574
Like what I was finding was that I was just sort of working towards this promised land where the system would just work, and then it would all scale super easily, and I now don't believe that's possible, at least not right now.
63
00:09:08.624 --> 00:09:17.604
Coming to that realization made me feel like maybe instead of just banging my head against the wall, what if I just opened up a blank Google Doc and started writing? [laughs] Would it actually take less time?
64
00:09:17.684 --> 00:09:27.584
[laughs] A- and al- additionally, and I feel like this doesn't get talked about all that much with AI, would I enjoy it more? I actually really enjoy the process of writing. Like, it's type two fun.
65
00:09:27.664 --> 00:09:34.574
It's hard, but it's satisfying. The last couple weeks, that's what I've been doing. I, I use the first part of my process. I call it, like, the ingestion phase.
66
00:09:34.784 --> 00:09:43.884
Interview a thought leader, I give it the transcript, and then I have it run through that transcript and identify all of the things that person talked about that could become blog posts.
67
00:09:44.384 --> 00:09:55.124
And then give me like three or four different angles for each idea. So if there's, like, five potential topics and then three ways to approach each, like, now I have, like, 15 things that I can think about.
68
00:09:55.744 --> 00:10:03.804
I find that to be pretty useful, and it kinda like helps me get to the part where I'm like, "Oh, I could take this, option number seven, and I can work with that."
69
00:10:03.824 --> 00:10:10.534
But from there, at least over the past, like I said, like, two weeks, I kinda just went back to doing things the old-fashioned way, and the results are...
70
00:10:11.764 --> 00:10:18.224
I mean, it's not like it's just me, I'm the only judge of these 'cause they haven't, like, hit the world yet, but they're so much better. So much better.
71
00:10:18.644 --> 00:10:29.144
And I don't find myself doing that thing that we all do when we're editing AI content, which is simultaneously trying to make it not sound like AI and try to make the piece better at the same time.
72
00:10:29.164 --> 00:10:33.823
It's really time-consuming, and it's, like, a very, like, tedious, annoying type of work.
73
00:10:33.884 --> 00:10:42.204
I think it depends on the approach that you're coming at this with because you guys all, you guys all have writing backgrounds, right?
74
00:10:42.324 --> 00:10:51.974
Like, very specifically, you started out as writers, and then you got into content marketing, right? Eric, you're like, kinda. [laughs] Because- More or less...
75
00:10:52.004 --> 00:10:58.823
and I say that because my background from content marketing is in editing. I started out as an- Yeah...
76
00:10:58.834 --> 00:11:09.564
editor, and then I had a mar- a marketing manager who was like, "I'm gonna teach you how to be a writer because I need you to write better than the people whose content you're writing."
77
00:11:09.664 --> 00:11:27.014
In this AI era, I feel like I have adapted well because my process has been, okay, how do I maintain the content creation process as best I can while guiding this agent to create content the way that I would guide a team as an editor?
78
00:11:27.244 --> 00:11:37.204
Instead of how do I go about automating this process to create content in the most creative, true to me way possible.
79
00:11:37.244 --> 00:11:57.914
Jimmy, it sounds like you're trying to maintain, and there has to be some kind of middle of how do I guide the AI in this editor creative way while also maintaining a human in the loop and ensuring that I can scale that out, and I, uh, to your point, I don't think that exists right now.
80
00:11:57.914 --> 00:11:59.254
[laughs] No. No.
81
00:11:59.324 --> 00:12:14.504
If you wanna, if you wanna do it good and, you know, do all of the things within a marketing and sales alignment process of does this content resonate, is it moving our customers down the funnel, I don't think you can do that unless you're gonna create sl- Yes.
82
00:12:14.564 --> 00:12:23.904
I also, one other thing I'll add is the, the work that I'm doing right now is thought leadership content for a Substack, like the Miro Substack, and it'll probably actually be live by the time this episode goes live.
83
00:12:23.944 --> 00:12:35.110
That has been a very, very helpful constraint because The whole point of Substack is subscribers, right? Like, nobody ever says, "Let's publish on Substack, see if we can get organic traffic." The...
84
00:12:35.140 --> 00:12:44.700
My goal in using AI was like, you know, come on, like, why not? Let's see if we can do this. But, like, my boss isn't telling me to use it. Scale isn't actually that important in this situation.
85
00:12:44.760 --> 00:12:54.010
When I sit down and I'm editing an AI-written piece, I'm thinking about, like, would someone subscribe to Substack after reading this? And the answer's just, like, no every single time.
86
00:12:54.040 --> 00:12:58.540
That's when I was like, "Okay, blank doc, start over. I can't work with this anymore. I gotta just do it the way I know how to do it."
87
00:12:58.760 --> 00:13:07.500
Found that I was able to sort of keep that idea of the, the future subscriber in my head, and that guided the writing. I couldn't find any way to articulate that to Claude.
88
00:13:08.360 --> 00:13:15.620
Earlier, you had asked kind of about our own journeys with this experience. I think I've never really adopted AI, to be honest.
89
00:13:15.700 --> 00:13:25.760
I've tried it for a lot of things, and I don't know if it's just the type of work that I do or the way that my brain thinks, but every time I try to use it, I just get so, so, so frustrated.
90
00:13:25.860 --> 00:13:33.440
And it comes back to the point you made earlier about enjoyment. I'm not having fun. The work's not getting done. [laughs] It's not doing good. It's not speeding things up.
91
00:13:33.560 --> 00:13:37.940
I do think I am starting to hear now more people s- having this experience.
92
00:13:38.140 --> 00:13:46.930
At first, it was actually really helpful, and I had this hope that it would continue to expand the amount of things that it could be helpful with.
93
00:13:47.030 --> 00:14:00.500
And so, I continued to really experiment in the hopes that eventually I would figure out what the process was to get it to be able to do these other things. And I think more people now are reaching this inflection point.
94
00:14:00.540 --> 00:14:11.310
We've come to the same point. Other people were just more persistent to keep trying and now are, are also feeling like, "Actually, maybe this just is something that I should just do myself."
95
00:14:11.360 --> 00:14:23.440
There's other ways that these tools can be helpful, and maybe this just isn't one of them. But I really think a big piece of it is something that you just said. As I write these pieces, I can keep the audience in mind.
96
00:14:23.480 --> 00:14:28.800
What do they need? What do they wanna read? What are they gonna subscribe to? What's gonna connect with them?
97
00:14:29.020 --> 00:14:40.120
No matter what or how much information on the audience you give to an AI tool, it can't do those things in the way that a human can do those things.
98
00:14:40.200 --> 00:14:54.480
It can't know what somebody is going to want to read just based off of knowing some things about them. A human can because we can imagine ourselves as that other person. We have empathy. We have human experience.
99
00:14:54.520 --> 00:15:09.520
We have perspective-taking. We're not cognitively thinking about this, but when we write, we are doing this, and I think that is, A, what makes it so hard to explain to a system like this, and B, the system can't do it.
100
00:15:09.780 --> 00:15:14.840
Maybe someday it will be able to, but certainly, at this point, it can't. That's a piece of the editing, too.
101
00:15:14.860 --> 00:15:20.520
When you keep going over it and you're like, "Something's just not working," that's a big piece of the something that doesn't work.
102
00:15:20.580 --> 00:15:26.120
The audience has not been consistently thought about as the piece was written or generated.
103
00:15:26.840 --> 00:15:38.540
With the tools that exist and that keep coming from these tech evolutions, the tools that already exist keep finding ways to combat abusing the new technologies.
104
00:15:38.700 --> 00:15:51.940
Even if you wanted to build out the system, Jimmy, to create your content at scale, we've seen people experiment with that, and then we've seen Google [laughs] be like, "No, no, no, don't do that.
105
00:15:52.160 --> 00:16:04.840
We're not gonna send any traffic to that content because it's not good. It's not gonna resonate. It's not meeting what we want from it." Even if it was generated with AI, are you putting any thought into it?
106
00:16:04.880 --> 00:16:14.200
At that point, you are fighting more with the system and putting more, more thought into guiding it rather than writing it yourself. Yes, totally.
107
00:16:14.940 --> 00:16:21.860
A- anecdotally, I think there's, there's a trend that I've observed where people are sort of hitting this point of fatigue. Like, this is my experience.
108
00:16:21.920 --> 00:16:27.460
Like, I'm like, "Oh, okay, I could, like, spend a month building this amazing system, and then it will just spit out articles, and then I can do something else."
109
00:16:27.540 --> 00:16:35.660
It's, like, kind of exhausting to do it and keep getting results you don't like, and then, um, eventually you hit a point where you're like, "I just... This is, like, a waste of my time." [laughs] You know what I mean?
110
00:16:35.740 --> 00:16:45.260
I personally have spent an enormous amount of time tinkering, and I'm just sort of tired of tinkering. I was having this LinkedIn conversation actually just yesterday with Tim Metz from Animals.
111
00:16:45.539 --> 00:16:53.800
I was venting my frustration that, like, Claude gives me something, but then I... if I wanna edit it, I have to tell Claude what I want. I can't just highlight the text and change it.
112
00:16:53.860 --> 00:17:03.370
That drives me insane because then you're just, like, constantly playing charades, where you're, like, dancing around and jumping and, and like, it like kinda gets what you mean but not exactly, and I'm like, "I could just do this myself in two seconds."
113
00:17:03.560 --> 00:17:09.220
And he was like, "Oh, well, you could just, like, vibe code up a text editor." And I was like, "Oh my God, absolutely not." [laughs] I will... I'm not... It...
114
00:17:09.280 --> 00:17:16.240
Like, a year ago, I'd have been like, "That's an amazing idea. I'm gonna spend three days doing that." [laughs] But I've just, like, hit a point now... Not that I think it's Tim.
115
00:17:16.340 --> 00:17:22.020
Tim actually probably could do it in five minutes. It would work 'cause he's, he's, like, amazing. I personally am just like, there's a million text editors.
116
00:17:22.060 --> 00:17:29.080
I'm not gonna rebuild Google Docs so that I can have this one little thing. I think a lot of it is a moving target of what's impressive.
117
00:17:29.440 --> 00:17:41.420
When any of us got our hands on ChatGPT two, three years ago, it was like, "Oh my God, a computer can write something that's 80% good enough?" That was absolutely mind-blowing at the time.
118
00:17:41.840 --> 00:17:49.800
Now, we're like, "Oh, I see it again. There's the ChatGPT essay." So we've- Yeah... like, as Rachel pointed out, as humans, we consume a lot of stuff. We internalize what's good.
119
00:17:49.860 --> 00:17:52.770
We internalize what's not, and like, the, the target has moved now.
120
00:17:52.960 --> 00:18:06.340
What felt, like, really close, "Oh, if the model gets 10% better, then I'll be able to make stuff with AI at this massive scale, and everything will be brilliant, and I'll be happy," the quality target has, like, shifted over time as a reader as well.
121
00:18:06.380 --> 00:18:11.300
We develop filters or heuristics for us to, like, notice what's AI and what's not, and we get sick of it.
122
00:18:11.440 --> 00:18:20.840
Part of me wonders sometimes, too, like, is this a case of Wayne Gretzky was a great hockey player and a terrible coach? Am I just a bad coach of AI? Is this a me problem?
123
00:18:20.900 --> 00:18:31.500
Like, is there someone out there who's actually, like, very good at abstracting their process in a way that they can give Claude this waterfall of steps to do? 'Cause I definitely don't think of my work in that way.
124
00:18:31.640 --> 00:18:42.980
I pull up a Google Doc and I'm like, "I'm gonna write a case study," and I hammer away at it over the course of two weeks, 15 minutes at a time, and it's rough, and it's terrible, and then eventually it's good, versus I'm asking Claude to be like, "Step one."
125
00:18:43.716 --> 00:18:52.476
Read the transcript. Step two, write the whole case study. And then, I mean, think about it. Step three. Like, that's not really how writing works, right? And I don't know, like, am I just explaining the task wrong?
126
00:18:53.286 --> 00:19:01.706
Have you... But have you ever had to handhold somebody and walk them through that process in guiding them in how to create something like that?
127
00:19:01.796 --> 00:19:15.016
To an extent, but I think, like, people can fill in the gaps a lot better than Claude is able to right now for whatever reason. Do marketers understand that you are s- literally talking to a large language model?
128
00:19:15.436 --> 00:19:27.576
The technicalities underneath how it actually works, I think, is really important to understand so that you know how to guide them, how you have to in order to get them to work the way you need to.
129
00:19:27.856 --> 00:19:30.516
That only comes with practice, unfortunately.
130
00:19:30.596 --> 00:19:41.616
And yeah, Jimmy, I think a year ago, a lot of us would've been like, "Yeah, let me spend all weekend doing this," but we can only tinker so much and execute our jobs at the same time.
131
00:19:41.716 --> 00:19:52.906
That's a great point, because it's two jobs. Yeah. It's- Right? It's two jobs. Yeah. A quick follow-up point on the idea of coaching writers, like, this is my first job at Animals in, like, I guess it was 2016.
132
00:19:53.256 --> 00:20:00.896
There's, like, seven people there. My job was to coach up this team of, of really young writers. The, the coaching piece had almost nothing to do with writing.
133
00:20:01.076 --> 00:20:11.976
The conversation is not about, like, how do we write a good article about product analytics? The coaching conversation is, like, what are product analytics? Who is installing these events to be tracked? For what purpose?
134
00:20:12.096 --> 00:20:18.936
What's usage-based pricing? Anyway, you sort of, like, go through this whole thing, and then the writing flows so much more easily.
135
00:20:18.976 --> 00:20:22.196
And, like, I can't even imagine trying to do that with Cla- Like, I wouldn't even bother.
136
00:20:22.376 --> 00:20:30.636
That person now has this bank of knowledge, so then when they pick up the next customer, they, like, already have so much context around the B2B SaaS industry and how these tools work and blah, blah, blah.
137
00:20:30.736 --> 00:20:39.696
Of course, you get into the writing bit of it, but that part just kind of flowed naturally. For all the reasons that people become good writers, like, they're suffering on their own. The coach doesn't have to suffer.
138
00:20:39.756 --> 00:20:46.086
The writer always has to suffer, for the most part. And, like, that, uh, that worked well, and, like, it's not that it was easy, but it was...
139
00:20:46.136 --> 00:20:56.446
It felt worth doing, where it feels, at least, at least right now, it does, it does not feel worth doing that type of work with AI. It's like AI tools can't do the whole trial by fire, you know?
140
00:20:56.616 --> 00:21:05.826
But they tell you that you- [laughs] You, you could give it, like, an insane task, and two seconds later it'd be like, "Great, done. Here's, like, an app I built that looks amazing," and is, like, is complete nonsense.
141
00:21:06.356 --> 00:21:13.666
And it's just, like, very confusing. And it'll gaslight you into thinking it's the best thing in the world. [laughs] Yeah, yes. Kind of, like, always having to calibrate. I'm like, "Wait a minute. It says...
142
00:21:13.916 --> 00:21:20.896
Is this good? I don't think this is good," but it looks like it's good, but it's actually not. It's, it's really... You use a lot of mental energy on that.
143
00:21:20.976 --> 00:21:39.386
Jimmy, one of my favorite parts of the piece, you were saying that you spend a lot less time vetting ideas now because I, I think the argument is because it's so easy to build things with AI that there's less pressure to vet the idea, whereas, like, you used to have to spend thousands of dollars to write something or produce something.
144
00:21:39.856 --> 00:21:47.436
Now you don't have to, and so you, you kind of skip that filter step. Yes. Can you just- I hate that. [laughs]... talk about that? [laughs] Yeah, yeah.
145
00:21:47.536 --> 00:21:59.396
I mean, um, yeah, like, so at Reforge and also now at Miro, we sell to product managers. In the PM space, like, something that comes up a lot lately is AI allows you to explore quite a bit.
146
00:21:59.556 --> 00:22:08.556
Building a prototype used to be, like, insanely expensive. Like, you would never build a pro- You would never have a developer write code until you were so confident that you were building the right thing.
147
00:22:08.996 --> 00:22:18.836
But now you can use Lovable and just make, uh, 10 prototypes in f- in a couple minutes. But the, the cool part is, like, you can explore and then try to feel out, like, oh, could I imagine this in our product?
148
00:22:19.096 --> 00:22:27.525
I've kind of tried to, like, take that and apply it to writing, and it just does not work. It... I kind of imagined it would be, like, a one-to-one substitute, but it's just not.
149
00:22:27.536 --> 00:22:33.916
Because what happens is you just waste all this time chasing around half-baked ideas and seeing if you can
150
00:22:34.936 --> 00:22:50.496
write them into something good, and I'm finding that that just is just a total waste of time, and it's way better to spend that time vetting in the exact same way that we/I used to vet, where there is a real cost, there is real time involved.
151
00:22:50.556 --> 00:23:01.406
If you don't get it right, then there's a, a real consequence in terms of, like, time or money lost. One of the interesting things about writing is Rachel left a comment in the article, uh, in a, in a draft about how...
152
00:23:02.176 --> 00:23:07.686
You were saying, Rachel, like, the way that I vet ideas is by writing, and that's sort of how I find out is this any good.
153
00:23:07.696 --> 00:23:15.676
And I was thinking about that 'cause I, like, sort of do a similar thing, but it's, like, a super messy... I'm, like, copying and pasting stuff and putting links in a Google Doc and, like, trying to think it out.
154
00:23:15.856 --> 00:23:24.876
I wouldn't really call it writing. It's, it's kind of like my, like, research step before I do that. But it's- Scribbling is what I call it to myself. Yeah. Yes, scribbling. That's exactly what it is.
155
00:23:24.956 --> 00:23:31.236
And, like, sometimes you do that, or at least I do that, and I have a doc full of stuff, and I'm like, "Well, [laughs] it's just not gonna work."
156
00:23:31.456 --> 00:23:39.056
But I save it, and then I almost always end up coming back to it later, even if it's not that like, oh, like, light bulb went off. I can use it now.
157
00:23:39.086 --> 00:23:59.656
It's more like, oh, there was, like, a little thing I found, and that's, like, a super useful cognitive exercise where I'm, like, filling my brain and then, like, using my own brain [laughs] to try to make a judgment on whether this is worth pursuing or not, and whether you pursue it or not, you actually still get a ton of value from it, even if that value is not an article that you can publish.
158
00:23:59.776 --> 00:24:03.476
Yeah, I love that. I have that same experience with the Superpath newsletter all the time.
159
00:24:03.696 --> 00:24:11.376
I have a list of ideas, and I, I start writing it, and then if in, like, five minutes, like, the light bulb hasn't clicked, this idea was nothing.
160
00:24:11.556 --> 00:24:21.076
If it didn't immediately turn into, like, 1,000 words in five minutes, then it wasn't a good enough idea. Yeah, yeah. I, I will say I've had more... not fun.
161
00:24:21.196 --> 00:24:32.195
I've had more sort of satisfaction from work over the last, even just the last couple weeks, than I've had over the last year because I just gave up [laughs] on trying to do it a new way and just going back to doing it.
162
00:24:32.576 --> 00:24:38.256
I keep calling it the old way, but just kind of like the way that I know how and the way also that I, that I trust will get me a good result.
163
00:24:38.596 --> 00:24:48.306
That shift of focus from can I make this faster and almost as good to I don't care if this is a little bit slower because I know I'm gonna get the outcome I want, has been...
164
00:24:48.516 --> 00:24:58.664
I just feel like I'm not, like, trapped in this AI world, and I feel like I have- Some say again I feel like there's a lot of people who will be listening to this that are like, "Duh."
165
00:24:58.864 --> 00:25:02.924
Like, "I've, I have felt this the whole time. Like, I appreciate- Yeah...
166
00:25:02.954 --> 00:25:13.324
the art of writing, but the problem I have is the execs on my team say AI exists, and therefore, I need to churn out three times the amount of content that I did in the past."
167
00:25:13.844 --> 00:25:24.484
I'm curious, Jimmy, as you think about transitioning your process, and you're saying, "I'm comfortable pulling back a little bit," do you foresee that being a challenge, how you communicate that?
168
00:25:24.544 --> 00:25:32.584
Like, "Hey, you know, I've realized this about my process, and I'm gonna deliver you fewer things, but they're gonna be better, and you'll see the value in that."
169
00:25:32.804 --> 00:25:44.644
I think it's an easier case to make if you're launching a Substack, for example. I think it's a harder case to make if you are tasked with creating content meant for SEO or AEO.
170
00:25:44.904 --> 00:25:53.104
Also, though, content marketers have been dealing with short-sighted management forever. And so it's not like a... That's like a n- not really a new problem.
171
00:25:53.404 --> 00:26:05.664
If you're in a situation where someone is saying like, "No, you, you have to double your output," or you're a freelancer and the client is saying like, "We're gonna pay you half as much as we used to," like, that's toxic, in my opinion.
172
00:26:05.704 --> 00:26:12.764
Like, truly toxic. I'm imagining that those same types of things happened five years ago. The context was just a little bit different.
173
00:26:13.084 --> 00:26:26.264
It's relatively easy to show version A that you wrote and version B that Claude wrote, and even without explaining it, I think the difference would be, like, pretty obvious to most people that one is going to be better than the other.
174
00:26:26.324 --> 00:26:35.244
I mean, it depends on, like, the skill of the writer and a million kind of subjective things too, but I think it's pretty easy to beat AI, and I think it's pretty easy to prove that you can beat AI also.
175
00:26:35.344 --> 00:26:45.544
You know, maybe the conversation then is like, "Okay, I can use AI to, like, make everything else faster and easier, but, like, I'm going to protect this one thing because if we don't, then it's like we don't even respect our readers."
176
00:26:45.664 --> 00:26:51.504
I think there's, like, a hard line you can draw there while still leaning on AI to do, like, all kinds of other useful, productive things.
177
00:26:51.544 --> 00:27:01.384
I also think your own personal capacity is something you have to proactively communicate at all times. You can sort of just pull AI usage into that. You know, like, "I tried AI for this.
178
00:27:01.564 --> 00:27:08.764
I really thought we could, like, scale this up to this m- massive scale that I promised. It's not working as I expected, therefore, I have to pull back channel A."
179
00:27:08.944 --> 00:27:16.884
Uh, or like, "We decided to write all of our LinkedIn posts now with AI. They're clearly not working. I, I think we should take that back, sacrifice this other thing we're doing."
180
00:27:17.164 --> 00:27:25.694
I, I don't think any manager is like, "I want you to do these 15 channels, and I want you to use AI for all of them." What they want is business results, right? Mm-hmm. They, they think that that's a means to an end.
181
00:27:25.904 --> 00:27:28.424
If you can communicate that it's not, like, they'll get there.
182
00:27:28.524 --> 00:27:36.524
I just try to sort of keep an open dialogue with, with our CEO at Dock on, like, "Look at this cool thing I did with AI," but then you're like, "Ooh, look at how badly this failed."
183
00:27:36.564 --> 00:27:44.604
And I, I think, like, trickling it- [laughs]... into, like, day-to-day conversations is working better than, like, making grand statements about, like, when I'm going to use AI or not.
184
00:27:44.764 --> 00:27:54.684
I, I don't know if that's, like, a standard relationship people have with it. Maybe this is a hot take, but a good leader is going to respect you asking, "Why are we using this?
185
00:27:54.904 --> 00:28:09.524
What is the goal here, and what is, what is your goal? Why are you having me use this?" I think a lot of teams now are really trying to figure out, how do we actually get a return out of these tools that we're using?
186
00:28:09.604 --> 00:28:21.804
And this is kind of how you figure that out. Don't create content just to create content. Are you really gonna post that just to post that? Are you gonna use AI just to use AI to get that up? Come on.
187
00:28:22.084 --> 00:28:36.964
[laughs] At some point, this is all gonna change because OpenAI and Antropic will go public, they'll raise their prices, and then every company's gonna be like, "We have to rein in our token spend because this is insanely expensive, and the company is not any more profitable than it used to be."
188
00:28:37.694 --> 00:28:47.424
[laughs] And then we'll all have this, like, much easier case. And they, they're already doing that. Yeah, totally. Like- Yeah... like, I have zero sense how much money I have spent on Miro's behalf this summer.
189
00:28:47.584 --> 00:28:54.664
Not a clue. And like- You don't wanna, and you don't wanna know. [laughs] I don't really, I don't really, I don't really care, but I bet it's a lot. [laughs] You know?
190
00:28:54.824 --> 00:29:06.804
I mean- Like, companies are now having to look at, like, okay great, that list pull that just happened was actually so much more money than it could've been had we set parameters on which models our teams are using.
191
00:29:06.944 --> 00:29:10.184
But I, I don't wanna get too technical here. [laughs] Yeah, yeah.
192
00:29:10.224 --> 00:29:18.124
My, my brother's a software engineer at Apple, and he was saying that his boss told them, "You are not allowed to use Claude Code to change the color of anything.
193
00:29:18.424 --> 00:29:27.984
You have to go into the code and change the hex code because we are burning a ton of tokens with these tiny little requests that would probably just be easier to go and do manually anyways."
194
00:29:28.044 --> 00:29:31.084
Retraining them a little bit to then, like, use less of it. It's weird.
195
00:29:31.244 --> 00:29:41.204
I think it's ironic because AI was kind of initially, and I think still is, question mark, branded as this tool that, like, you're not gonna have to do any tedious, menial work anymore.
196
00:29:41.244 --> 00:29:50.164
You're just gonna be able to work on these amazing, big, blue sky projects. So it's ironic to tell your senior devs- That would be the tedious work... go in there and change all the hex codes manually.
197
00:29:50.174 --> 00:29:56.754
Like, you can't get any more tedious than that. [laughs] These gi- these giant areas where, like, it just makes everything so much easier, and then there's these- Mm-hmm...
198
00:29:56.754 --> 00:30:05.184
other giant areas where it just makes everything so much harder. Mm-hmm. Like, so much harder. In this, in this way that I feel like, at least for me, like, I'm only just now appreciating.
199
00:30:05.244 --> 00:30:13.004
As my, like, excitement wanes a bit, I'm like, "Oh, I actually hate doing all these little tedious things that I do as a result of AI usage."
200
00:30:13.144 --> 00:30:22.164
I guess I've been there for a bit now, but I'm kinda only just, like, kind of, uh, like waking up. There's always going to be tedium and suffering in the work that we do.
201
00:30:22.264 --> 00:30:28.664
We can try to escape that, and we can try to figure out how to do it, but, like, that's just kinda how work is.
202
00:30:28.853 --> 00:30:36.314
It's kind of about figuring out where we wanna apply that suffering and also, like, where that suffering counts for something.
203
00:30:36.684 --> 00:30:47.004
Most of us feel happy to do hard work that does something, but if it's just, like, making all these hex codes, you know? Like, that's not purposeful suffering.
204
00:30:47.444 --> 00:30:55.554
Your big idea in this piece is, is like, direct your suffering to a place where it will matter, where it will be meaningful and purposeful. Love that. I love that.
205
00:30:55.604 --> 00:31:04.324
I actually feel like that's, like, a perfect place to end. If you wanna come and suffer with us- [laughs]... join Superpath Pro. Join Superpath Pro to suffer purposefully with us.
206
00:31:04.444 --> 00:31:09.164
I would say the Reading List channel has been really good lately. Ronnie Fiegens, today or yesterday, dropped an article.
207
00:31:09.244 --> 00:31:19.544
If you wanna, like, really get fired up about AI and writing, go to the Reading List channel and read what Ronnie posted about the Claude watermarking thing. It's wild.
208
00:31:19.654 --> 00:31:22.044
We could do a whole episode on Claude watermarking alone.
209
00:31:22.084 --> 00:31:32.524
But anyways, we're all, like, dealing with the same, more or less the same things in our day-to-day work, and it's an awesome place, um, to see what other people are doing and also see what they're frustrated with and vent about it and then, uh, get support.
210
00:31:32.584 --> 00:31:42.104
That's like a, kind of a microcosm of what's happening in the community. Jimmy, thanks for joining us. Rachel, Ruth, Eric, thanks for being here. Talk to y'all soon. Always a great time. Bye, guys. Bye. Thanks, guys.
211
00:31:42.114 --> 00:31:49.204
See you. [outro music]
1
00:00:00.580 --> 00:00:08.550
Like right now, this is the intro? No, no, no. No. Oh, God. This is the intro. [laughs] No, no, I'll start it, and then I'll, I'll, I'll set it up. Thank you. [laughs] It's like, intro to what?
2
00:00:09.400 --> 00:00:17.919
Today's episode is brought to you by our friends at Venngage. You spent three weeks on that report. Be honest about how many people actually read past the intro.
3
00:00:18.420 --> 00:00:24.960
Venngage AI is the fix, an AI infographic generator that turns the thing you wrote into the thing people actually look at.
4
00:00:25.720 --> 00:00:32.500
Think about the last thing you understood the second you saw it, a chart, a process diagram, one clean visual.
5
00:00:32.509 --> 00:00:42.520
We read pictures faster than paragraphs and hold onto them longer, which is why the idea you can't get across in three paragraphs usually lands in a single image. The catch has always been time.
6
00:00:42.940 --> 00:00:47.540
A good infographic meant a designer, a queue, and three rounds of feedback.
7
00:00:47.600 --> 00:00:59.500
Venngage AI handles the structure for you, smart layouts, clean hierarchy, your brand colors applied across every asset, so it all looks like it came from the same company, and everything's editable.
8
00:00:59.540 --> 00:01:08.540
Change any element down to the last detail. AI design that's actually yours. Take the report nobody finished reading and turn it into the thing that people share.
9
00:01:08.630 --> 00:01:20.340
And this month only, 30% off any Venngage plan with the code SUPERPATH30 at venngage.com. That's SUPERPATH30 at venngage.com. Hope you enjoy the episode.
10
00:01:21.900 --> 00:01:25.420
Alex here, and welcome back to another Jimmy episode of Content Briefly.
11
00:01:25.640 --> 00:01:33.780
Today, Superpath founder emeritus Jimmy Daly joins us to talk about his recent article for us, The Hardest Part About Content Hasn't Gotten Any Easier.
12
00:01:33.840 --> 00:01:43.100
Jimmy will be the first to admit that the current AI content process he's building at Miro is a ton of fun for him, but he's recently experienced some last-mile content fatigue.
13
00:01:43.160 --> 00:01:48.960
His question, if AI and the writing process only ever gets us at least 90% of the way there, is that enough?
14
00:01:49.160 --> 00:01:58.420
We've got the whole gang together today, including Eric, Ruth, and Rachel, to talk about where we stand on the process of using AI systems to help us write great things. Enjoy the episode.
15
00:01:58.960 --> 00:02:05.680
You know, all the episodes that we've done on AI have been, um, like I've just been so pumped about it, really excited.
16
00:02:05.720 --> 00:02:14.220
I think this is the first one where I, like, really hit a stage in my AI marketing journey where I became really frustrated, but led to this important realization.
17
00:02:14.609 --> 00:02:23.980
And so the article, it's titled The Hardest Part of Content Hasn't Gotten Any Easier, but I almost called it Writing Suffers When Humans Don't, 'cause that's the real idea of it.
18
00:02:24.120 --> 00:02:34.400
The kind of the backstory here is that, so I'm at Miro now. I'm part of the content team, and have spent a, quite a bit of this summer working on content creation systems in cloud code.
19
00:02:34.769 --> 00:02:40.700
And it's a lot of fun, some of the most fun work I've ever done. Maybe I'll do an AI show and tell to walk through, like, how it actually works.
20
00:02:40.800 --> 00:02:50.200
But the idea is that I give the system raw material, which is usually an interview transcript, plus whatever other, like, notes and decks and stuff that I can cobble together on the topic.
21
00:02:50.240 --> 00:02:56.860
That's sort of like step one, and then there's probably 10 different steps and three different human checkpoints.
22
00:02:56.890 --> 00:03:09.460
And the idea is it, it's supposed to kind of mirror a real writing process, except that, one, it's automated by AI, and two, I'm kind of the facilitator, not the actual writer.
23
00:03:09.500 --> 00:03:17.040
One of the things I've sort of learned while doing this is that it's a total abstraction of the type of work that I used to do, and I think many of us used to do.
24
00:03:17.140 --> 00:03:19.720
Instead of actually writing the article, I'm explaining how to write the article.
25
00:03:19.820 --> 00:03:29.710
In that process, there's this whole other thing that I've, like, actually really enjoyed, which is the opportunity to step back and try to articulate, I mean, I think I'm an okay writer, and try to articulate, like, why am I an okay writer?
26
00:03:29.720 --> 00:03:37.950
And what goes into it? And what are all the various steps? It's really hard. It's, like, really, [laughs] really hard to capture all of that and make the system work the way that I want it to work.
27
00:03:38.180 --> 00:03:46.880
So I've run this now hundreds of times. We've published a bunch of stuff too. Eventually, I've gotten quite a bit of it to a place where I felt it was worthy of publishing on social and the blog.
28
00:03:46.920 --> 00:03:54.780
But in doing this, this is kind of where the brewing sense of uncertainty has been coming from. Why is the whole less than the sum of the parts?
29
00:03:54.860 --> 00:04:03.160
Why can't I explain this to Claude in a way that will make the output better? Sometimes it just totally misses, and AI does that sometimes.
30
00:04:03.280 --> 00:04:15.460
Sometimes you end up with this, like, last mile problem, where you can get 80 or 90% of the way there and it feels amazing, but then you spend so much time trying to run that last mile that you, like, give all that time back up.
31
00:04:15.480 --> 00:04:29.909
The realization that I've come to, which is I, I'm sure is not unique among content marketers and certainly not among this group, is that it's too easy [laughs] and therefore the writing will never be as good as it could be if someone endured some misery to create it.
32
00:04:30.130 --> 00:04:39.080
[laughs] In order to create something good, you have to sit there with it, and there's a part of it that's, like, a little bit miserable and you suffer a little bit in, like, this creative sense.
33
00:04:39.400 --> 00:04:46.260
I'm very curious, you know, Rachel, Ruth, Eric, have you all gone through a similar journey? Have you come to accept the limitations of AI?
34
00:04:46.460 --> 00:04:58.500
Or are you more like me, where you're, like, spend so much time tweaking it, hoping you can finally make it work, and that will, like, unlock this giant wave of productivity before coming to this realization that, like, probably I'm actually a little bit late on?
35
00:04:58.600 --> 00:05:01.200
Eric, since we went in this order earlier- Yeah.... y'all said it too first.
36
00:05:01.960 --> 00:05:12.650
I, I pretty much, like, the week you wrote this piece, I was having an existential crisis about AI again, where I was like, I, I had so much hope and I feel it all crashing down.
37
00:05:12.900 --> 00:05:22.920
A- and for kind of precisely the reason you talked about, that I, I didn't feel like I was proud of things that were coming out, not because they felt like they were written by AI or they had AI-isms in them.
38
00:05:23.080 --> 00:05:33.140
I just wasn't feeling a connection with what I was making anymore. A lot of it was what you highlighted, because of the lack of suffering for writing the piece. I'll just give the Superpath newsletter as an example.
39
00:05:33.300 --> 00:05:42.960
I never use AI for that, and I think the whole reason it's good is because of how much day-to-day job suffering I do to, like, come up with the topics.
40
00:05:43.120 --> 00:05:53.020
I'm, like, w- writing it as fast as I can, stream of thought the morning I send it out, which produces all these weird things that I would never say if I was coaching an AI writer to write something.
41
00:05:53.190 --> 00:06:01.360
The conditions sort of produce the newsletter, and if I change the conditions too much, the quality wouldn't be there anymore. So like, I- Yeah.... I think it really resonated with me.
42
00:06:01.380 --> 00:06:09.600
That's interesting, because, like, in the case of the newsletter, the suffering does not occur while writing the newsletter. It occurs while you are gaining- A little bit of both. [laughs]... the experience that...
43
00:06:09.660 --> 00:06:18.853
Yeah, yeah. [laughs]But it's like you earned the right to create the newsletter by doing your job, right? And that is hard. And then that- Well, yeah. And then, like, I'm, I'm trying to, like, pull up-...
44
00:06:18.853 --> 00:06:26.664
creates the ideas you write about... I'm trying to pull from things to say, and I think of the weirdest specific examples and weird phrasings because it's sort of off the cuff.
45
00:06:26.764 --> 00:06:34.244
I don't think I could, like, coach an AI- Yeah. Yeah... editor or writer to come up with those same sort of, not just the anecdotes, but also the way it comes across.
46
00:06:34.304 --> 00:06:40.214
Not to say that I'm, like, a unique butterfly, but that it, it comes through in the writing because of how I'm going about writing it. And the writing- Yeah. Yeah...
47
00:06:40.214 --> 00:06:51.044
I do for Doc ends up a very different style because I'm having a different process. So Jimmy, when you were writing this, were you writing this only thinking about the actual content creation process?
48
00:06:51.604 --> 00:06:57.654
I am talking to Claude, and Claude is writing my blog for me. Because- Mm...
49
00:06:57.764 --> 00:07:10.524
I feel like as, as someone who has been in AI for a while, one thing that I have thought about for so long in the content process is I feel like I have now moved to, okay, where am I keeping things separate now?
50
00:07:10.534 --> 00:07:21.724
Where am I so focused on where the human stays in the loop? To your point, a piece of the process has to stay painful [laughs] for the content to be good.
51
00:07:21.784 --> 00:07:31.544
You know, I, I recently joined, um, an AI consulting firm, and I'm their director of marketing now. Part of our content process is we record a p- podcast twice a month.
52
00:07:32.104 --> 00:07:38.944
That podcast, it's still heavily manual, but we are using s- Claude's skills to get a lot of it done.
53
00:07:39.444 --> 00:07:51.044
I'm not gonna expect my content manager to repurpose a data science conversation on her own, but I do expect the data scientists to sit there and talk about these topics.
54
00:07:51.154 --> 00:08:02.784
And if they're gonna do it off the cuff, then that's gonna be the painful process for them. I am making sure that I'm still at least staying the human in one piece of it.
55
00:08:02.924 --> 00:08:14.244
I'm curious what part you were thinking about as you were writing this, or if you were specifically thinking about guiding Claude to write new content and doing that at scale as you were going through this. The latter.
56
00:08:14.504 --> 00:08:19.544
Like, my- Okay... my goal was, I'm not, like, totally giving up on this, but I'm definitely changing the approach.
57
00:08:19.744 --> 00:08:25.664
The goal was to go from I interview a thought leader to I have a thought leadership article that's ready to go.
58
00:08:25.724 --> 00:08:31.504
There is a significant amount of writing in that, but, like, like in my mind, writing is kinda like two distinct steps.
59
00:08:31.624 --> 00:08:41.074
One is understanding what you're trying to say, and then the second is putting the words in the right order so that it's nice for a reader to consume. My goal was to, was to do both.
60
00:08:41.284 --> 00:08:47.054
And there are separate workflows for, like, I guess what I would call marketing hygiene. You know, like, using AI to write show notes.
61
00:08:47.184 --> 00:08:58.104
All the things that, like, we all sort of kn- know we should do, but it's easy to not do, and now AI makes it super easy. That's, like, the best use case. But trying to get AI to do all of it is, um...
62
00:08:58.144 --> 00:09:08.574
Like what I was finding was that I was just sort of working towards this promised land where the system would just work, and then it would all scale super easily, and I now don't believe that's possible, at least not right now.
63
00:09:08.624 --> 00:09:17.604
Coming to that realization made me feel like maybe instead of just banging my head against the wall, what if I just opened up a blank Google Doc and started writing? [laughs] Would it actually take less time?
64
00:09:17.684 --> 00:09:27.584
[laughs] A- and al- additionally, and I feel like this doesn't get talked about all that much with AI, would I enjoy it more? I actually really enjoy the process of writing. Like, it's type two fun.
65
00:09:27.664 --> 00:09:34.574
It's hard, but it's satisfying. The last couple weeks, that's what I've been doing. I, I use the first part of my process. I call it, like, the ingestion phase.
66
00:09:34.784 --> 00:09:43.884
Interview a thought leader, I give it the transcript, and then I have it run through that transcript and identify all of the things that person talked about that could become blog posts.
67
00:09:44.384 --> 00:09:55.124
And then give me like three or four different angles for each idea. So if there's, like, five potential topics and then three ways to approach each, like, now I have, like, 15 things that I can think about.
68
00:09:55.744 --> 00:10:03.804
I find that to be pretty useful, and it kinda like helps me get to the part where I'm like, "Oh, I could take this, option number seven, and I can work with that."
69
00:10:03.824 --> 00:10:10.534
But from there, at least over the past, like I said, like, two weeks, I kinda just went back to doing things the old-fashioned way, and the results are...
70
00:10:11.764 --> 00:10:18.224
I mean, it's not like it's just me, I'm the only judge of these 'cause they haven't, like, hit the world yet, but they're so much better. So much better.
71
00:10:18.644 --> 00:10:29.144
And I don't find myself doing that thing that we all do when we're editing AI content, which is simultaneously trying to make it not sound like AI and try to make the piece better at the same time.
72
00:10:29.164 --> 00:10:33.823
It's really time-consuming, and it's, like, a very, like, tedious, annoying type of work.
73
00:10:33.884 --> 00:10:42.204
I think it depends on the approach that you're coming at this with because you guys all, you guys all have writing backgrounds, right?
74
00:10:42.324 --> 00:10:51.974
Like, very specifically, you started out as writers, and then you got into content marketing, right? Eric, you're like, kinda. [laughs] Because- More or less...
75
00:10:52.004 --> 00:10:58.823
and I say that because my background from content marketing is in editing. I started out as an- Yeah...
76
00:10:58.834 --> 00:11:09.564
editor, and then I had a mar- a marketing manager who was like, "I'm gonna teach you how to be a writer because I need you to write better than the people whose content you're writing."
77
00:11:09.664 --> 00:11:27.014
In this AI era, I feel like I have adapted well because my process has been, okay, how do I maintain the content creation process as best I can while guiding this agent to create content the way that I would guide a team as an editor?
78
00:11:27.244 --> 00:11:37.204
Instead of how do I go about automating this process to create content in the most creative, true to me way possible.
79
00:11:37.244 --> 00:11:57.914
Jimmy, it sounds like you're trying to maintain, and there has to be some kind of middle of how do I guide the AI in this editor creative way while also maintaining a human in the loop and ensuring that I can scale that out, and I, uh, to your point, I don't think that exists right now.
80
00:11:57.914 --> 00:11:59.254
[laughs] No. No.
81
00:11:59.324 --> 00:12:14.504
If you wanna, if you wanna do it good and, you know, do all of the things within a marketing and sales alignment process of does this content resonate, is it moving our customers down the funnel, I don't think you can do that unless you're gonna create sl- Yes.
82
00:12:14.564 --> 00:12:23.904
I also, one other thing I'll add is the, the work that I'm doing right now is thought leadership content for a Substack, like the Miro Substack, and it'll probably actually be live by the time this episode goes live.
83
00:12:23.944 --> 00:12:35.110
That has been a very, very helpful constraint because The whole point of Substack is subscribers, right? Like, nobody ever says, "Let's publish on Substack, see if we can get organic traffic." The...
84
00:12:35.140 --> 00:12:44.700
My goal in using AI was like, you know, come on, like, why not? Let's see if we can do this. But, like, my boss isn't telling me to use it. Scale isn't actually that important in this situation.
85
00:12:44.760 --> 00:12:54.010
When I sit down and I'm editing an AI-written piece, I'm thinking about, like, would someone subscribe to Substack after reading this? And the answer's just, like, no every single time.
86
00:12:54.040 --> 00:12:58.540
That's when I was like, "Okay, blank doc, start over. I can't work with this anymore. I gotta just do it the way I know how to do it."
87
00:12:58.760 --> 00:13:07.500
Found that I was able to sort of keep that idea of the, the future subscriber in my head, and that guided the writing. I couldn't find any way to articulate that to Claude.
88
00:13:08.360 --> 00:13:15.620
Earlier, you had asked kind of about our own journeys with this experience. I think I've never really adopted AI, to be honest.
89
00:13:15.700 --> 00:13:25.760
I've tried it for a lot of things, and I don't know if it's just the type of work that I do or the way that my brain thinks, but every time I try to use it, I just get so, so, so frustrated.
90
00:13:25.860 --> 00:13:33.440
And it comes back to the point you made earlier about enjoyment. I'm not having fun. The work's not getting done. [laughs] It's not doing good. It's not speeding things up.
91
00:13:33.560 --> 00:13:37.940
I do think I am starting to hear now more people s- having this experience.
92
00:13:38.140 --> 00:13:46.930
At first, it was actually really helpful, and I had this hope that it would continue to expand the amount of things that it could be helpful with.
93
00:13:47.030 --> 00:14:00.500
And so, I continued to really experiment in the hopes that eventually I would figure out what the process was to get it to be able to do these other things. And I think more people now are reaching this inflection point.
94
00:14:00.540 --> 00:14:11.310
We've come to the same point. Other people were just more persistent to keep trying and now are, are also feeling like, "Actually, maybe this just is something that I should just do myself."
95
00:14:11.360 --> 00:14:23.440
There's other ways that these tools can be helpful, and maybe this just isn't one of them. But I really think a big piece of it is something that you just said. As I write these pieces, I can keep the audience in mind.
96
00:14:23.480 --> 00:14:28.800
What do they need? What do they wanna read? What are they gonna subscribe to? What's gonna connect with them?
97
00:14:29.020 --> 00:14:40.120
No matter what or how much information on the audience you give to an AI tool, it can't do those things in the way that a human can do those things.
98
00:14:40.200 --> 00:14:54.480
It can't know what somebody is going to want to read just based off of knowing some things about them. A human can because we can imagine ourselves as that other person. We have empathy. We have human experience.
99
00:14:54.520 --> 00:15:09.520
We have perspective-taking. We're not cognitively thinking about this, but when we write, we are doing this, and I think that is, A, what makes it so hard to explain to a system like this, and B, the system can't do it.
100
00:15:09.780 --> 00:15:14.840
Maybe someday it will be able to, but certainly, at this point, it can't. That's a piece of the editing, too.
101
00:15:14.860 --> 00:15:20.520
When you keep going over it and you're like, "Something's just not working," that's a big piece of the something that doesn't work.
102
00:15:20.580 --> 00:15:26.120
The audience has not been consistently thought about as the piece was written or generated.
103
00:15:26.840 --> 00:15:38.540
With the tools that exist and that keep coming from these tech evolutions, the tools that already exist keep finding ways to combat abusing the new technologies.
104
00:15:38.700 --> 00:15:51.940
Even if you wanted to build out the system, Jimmy, to create your content at scale, we've seen people experiment with that, and then we've seen Google [laughs] be like, "No, no, no, don't do that.
105
00:15:52.160 --> 00:16:04.840
We're not gonna send any traffic to that content because it's not good. It's not gonna resonate. It's not meeting what we want from it." Even if it was generated with AI, are you putting any thought into it?
106
00:16:04.880 --> 00:16:14.200
At that point, you are fighting more with the system and putting more, more thought into guiding it rather than writing it yourself. Yes, totally.
107
00:16:14.940 --> 00:16:21.860
A- anecdotally, I think there's, there's a trend that I've observed where people are sort of hitting this point of fatigue. Like, this is my experience.
108
00:16:21.920 --> 00:16:27.460
Like, I'm like, "Oh, okay, I could, like, spend a month building this amazing system, and then it will just spit out articles, and then I can do something else."
109
00:16:27.540 --> 00:16:35.660
It's, like, kind of exhausting to do it and keep getting results you don't like, and then, um, eventually you hit a point where you're like, "I just... This is, like, a waste of my time." [laughs] You know what I mean?
110
00:16:35.740 --> 00:16:45.260
I personally have spent an enormous amount of time tinkering, and I'm just sort of tired of tinkering. I was having this LinkedIn conversation actually just yesterday with Tim Metz from Animals.
111
00:16:45.539 --> 00:16:53.800
I was venting my frustration that, like, Claude gives me something, but then I... if I wanna edit it, I have to tell Claude what I want. I can't just highlight the text and change it.
112
00:16:53.860 --> 00:17:03.370
That drives me insane because then you're just, like, constantly playing charades, where you're, like, dancing around and jumping and, and like, it like kinda gets what you mean but not exactly, and I'm like, "I could just do this myself in two seconds."
113
00:17:03.560 --> 00:17:09.220
And he was like, "Oh, well, you could just, like, vibe code up a text editor." And I was like, "Oh my God, absolutely not." [laughs] I will... I'm not... It...
114
00:17:09.280 --> 00:17:16.240
Like, a year ago, I'd have been like, "That's an amazing idea. I'm gonna spend three days doing that." [laughs] But I've just, like, hit a point now... Not that I think it's Tim.
115
00:17:16.340 --> 00:17:22.020
Tim actually probably could do it in five minutes. It would work 'cause he's, he's, like, amazing. I personally am just like, there's a million text editors.
116
00:17:22.060 --> 00:17:29.080
I'm not gonna rebuild Google Docs so that I can have this one little thing. I think a lot of it is a moving target of what's impressive.
117
00:17:29.440 --> 00:17:41.420
When any of us got our hands on ChatGPT two, three years ago, it was like, "Oh my God, a computer can write something that's 80% good enough?" That was absolutely mind-blowing at the time.
118
00:17:41.840 --> 00:17:49.800
Now, we're like, "Oh, I see it again. There's the ChatGPT essay." So we've- Yeah... like, as Rachel pointed out, as humans, we consume a lot of stuff. We internalize what's good.
119
00:17:49.860 --> 00:17:52.770
We internalize what's not, and like, the, the target has moved now.
120
00:17:52.960 --> 00:18:06.340
What felt, like, really close, "Oh, if the model gets 10% better, then I'll be able to make stuff with AI at this massive scale, and everything will be brilliant, and I'll be happy," the quality target has, like, shifted over time as a reader as well.
121
00:18:06.380 --> 00:18:11.300
We develop filters or heuristics for us to, like, notice what's AI and what's not, and we get sick of it.
122
00:18:11.440 --> 00:18:20.840
Part of me wonders sometimes, too, like, is this a case of Wayne Gretzky was a great hockey player and a terrible coach? Am I just a bad coach of AI? Is this a me problem?
123
00:18:20.900 --> 00:18:31.500
Like, is there someone out there who's actually, like, very good at abstracting their process in a way that they can give Claude this waterfall of steps to do? 'Cause I definitely don't think of my work in that way.
124
00:18:31.640 --> 00:18:42.980
I pull up a Google Doc and I'm like, "I'm gonna write a case study," and I hammer away at it over the course of two weeks, 15 minutes at a time, and it's rough, and it's terrible, and then eventually it's good, versus I'm asking Claude to be like, "Step one."
125
00:18:43.716 --> 00:18:52.476
Read the transcript. Step two, write the whole case study. And then, I mean, think about it. Step three. Like, that's not really how writing works, right? And I don't know, like, am I just explaining the task wrong?
126
00:18:53.286 --> 00:19:01.706
Have you... But have you ever had to handhold somebody and walk them through that process in guiding them in how to create something like that?
127
00:19:01.796 --> 00:19:15.016
To an extent, but I think, like, people can fill in the gaps a lot better than Claude is able to right now for whatever reason. Do marketers understand that you are s- literally talking to a large language model?
128
00:19:15.436 --> 00:19:27.576
The technicalities underneath how it actually works, I think, is really important to understand so that you know how to guide them, how you have to in order to get them to work the way you need to.
129
00:19:27.856 --> 00:19:30.516
That only comes with practice, unfortunately.
130
00:19:30.596 --> 00:19:41.616
And yeah, Jimmy, I think a year ago, a lot of us would've been like, "Yeah, let me spend all weekend doing this," but we can only tinker so much and execute our jobs at the same time.
131
00:19:41.716 --> 00:19:52.906
That's a great point, because it's two jobs. Yeah. It's- Right? It's two jobs. Yeah. A quick follow-up point on the idea of coaching writers, like, this is my first job at Animals in, like, I guess it was 2016.
132
00:19:53.256 --> 00:20:00.896
There's, like, seven people there. My job was to coach up this team of, of really young writers. The, the coaching piece had almost nothing to do with writing.
133
00:20:01.076 --> 00:20:11.976
The conversation is not about, like, how do we write a good article about product analytics? The coaching conversation is, like, what are product analytics? Who is installing these events to be tracked? For what purpose?
134
00:20:12.096 --> 00:20:18.936
What's usage-based pricing? Anyway, you sort of, like, go through this whole thing, and then the writing flows so much more easily.
135
00:20:18.976 --> 00:20:22.196
And, like, I can't even imagine trying to do that with Cla- Like, I wouldn't even bother.
136
00:20:22.376 --> 00:20:30.636
That person now has this bank of knowledge, so then when they pick up the next customer, they, like, already have so much context around the B2B SaaS industry and how these tools work and blah, blah, blah.
137
00:20:30.736 --> 00:20:39.696
Of course, you get into the writing bit of it, but that part just kind of flowed naturally. For all the reasons that people become good writers, like, they're suffering on their own. The coach doesn't have to suffer.
138
00:20:39.756 --> 00:20:46.086
The writer always has to suffer, for the most part. And, like, that, uh, that worked well, and, like, it's not that it was easy, but it was...
139
00:20:46.136 --> 00:20:56.446
It felt worth doing, where it feels, at least, at least right now, it does, it does not feel worth doing that type of work with AI. It's like AI tools can't do the whole trial by fire, you know?
140
00:20:56.616 --> 00:21:05.826
But they tell you that you- [laughs] You, you could give it, like, an insane task, and two seconds later it'd be like, "Great, done. Here's, like, an app I built that looks amazing," and is, like, is complete nonsense.
141
00:21:06.356 --> 00:21:13.666
And it's just, like, very confusing. And it'll gaslight you into thinking it's the best thing in the world. [laughs] Yeah, yes. Kind of, like, always having to calibrate. I'm like, "Wait a minute. It says...
142
00:21:13.916 --> 00:21:20.896
Is this good? I don't think this is good," but it looks like it's good, but it's actually not. It's, it's really... You use a lot of mental energy on that.
143
00:21:20.976 --> 00:21:39.386
Jimmy, one of my favorite parts of the piece, you were saying that you spend a lot less time vetting ideas now because I, I think the argument is because it's so easy to build things with AI that there's less pressure to vet the idea, whereas, like, you used to have to spend thousands of dollars to write something or produce something.
144
00:21:39.856 --> 00:21:47.436
Now you don't have to, and so you, you kind of skip that filter step. Yes. Can you just- I hate that. [laughs]... talk about that? [laughs] Yeah, yeah.
145
00:21:47.536 --> 00:21:59.396
I mean, um, yeah, like, so at Reforge and also now at Miro, we sell to product managers. In the PM space, like, something that comes up a lot lately is AI allows you to explore quite a bit.
146
00:21:59.556 --> 00:22:08.556
Building a prototype used to be, like, insanely expensive. Like, you would never build a pro- You would never have a developer write code until you were so confident that you were building the right thing.
147
00:22:08.996 --> 00:22:18.836
But now you can use Lovable and just make, uh, 10 prototypes in f- in a couple minutes. But the, the cool part is, like, you can explore and then try to feel out, like, oh, could I imagine this in our product?
148
00:22:19.096 --> 00:22:27.525
I've kind of tried to, like, take that and apply it to writing, and it just does not work. It... I kind of imagined it would be, like, a one-to-one substitute, but it's just not.
149
00:22:27.536 --> 00:22:33.916
Because what happens is you just waste all this time chasing around half-baked ideas and seeing if you can
150
00:22:34.936 --> 00:22:50.496
write them into something good, and I'm finding that that just is just a total waste of time, and it's way better to spend that time vetting in the exact same way that we/I used to vet, where there is a real cost, there is real time involved.
151
00:22:50.556 --> 00:23:01.406
If you don't get it right, then there's a, a real consequence in terms of, like, time or money lost. One of the interesting things about writing is Rachel left a comment in the article, uh, in a, in a draft about how...
152
00:23:02.176 --> 00:23:07.686
You were saying, Rachel, like, the way that I vet ideas is by writing, and that's sort of how I find out is this any good.
153
00:23:07.696 --> 00:23:15.676
And I was thinking about that 'cause I, like, sort of do a similar thing, but it's, like, a super messy... I'm, like, copying and pasting stuff and putting links in a Google Doc and, like, trying to think it out.
154
00:23:15.856 --> 00:23:24.876
I wouldn't really call it writing. It's, it's kind of like my, like, research step before I do that. But it's- Scribbling is what I call it to myself. Yeah. Yes, scribbling. That's exactly what it is.
155
00:23:24.956 --> 00:23:31.236
And, like, sometimes you do that, or at least I do that, and I have a doc full of stuff, and I'm like, "Well, [laughs] it's just not gonna work."
156
00:23:31.456 --> 00:23:39.056
But I save it, and then I almost always end up coming back to it later, even if it's not that like, oh, like, light bulb went off. I can use it now.
157
00:23:39.086 --> 00:23:59.656
It's more like, oh, there was, like, a little thing I found, and that's, like, a super useful cognitive exercise where I'm, like, filling my brain and then, like, using my own brain [laughs] to try to make a judgment on whether this is worth pursuing or not, and whether you pursue it or not, you actually still get a ton of value from it, even if that value is not an article that you can publish.
158
00:23:59.776 --> 00:24:03.476
Yeah, I love that. I have that same experience with the Superpath newsletter all the time.
159
00:24:03.696 --> 00:24:11.376
I have a list of ideas, and I, I start writing it, and then if in, like, five minutes, like, the light bulb hasn't clicked, this idea was nothing.
160
00:24:11.556 --> 00:24:21.076
If it didn't immediately turn into, like, 1,000 words in five minutes, then it wasn't a good enough idea. Yeah, yeah. I, I will say I've had more... not fun.
161
00:24:21.196 --> 00:24:32.195
I've had more sort of satisfaction from work over the last, even just the last couple weeks, than I've had over the last year because I just gave up [laughs] on trying to do it a new way and just going back to doing it.
162
00:24:32.576 --> 00:24:38.256
I keep calling it the old way, but just kind of like the way that I know how and the way also that I, that I trust will get me a good result.
163
00:24:38.596 --> 00:24:48.306
That shift of focus from can I make this faster and almost as good to I don't care if this is a little bit slower because I know I'm gonna get the outcome I want, has been...
164
00:24:48.516 --> 00:24:58.664
I just feel like I'm not, like, trapped in this AI world, and I feel like I have- Some say again I feel like there's a lot of people who will be listening to this that are like, "Duh."
165
00:24:58.864 --> 00:25:02.924
Like, "I've, I have felt this the whole time. Like, I appreciate- Yeah...
166
00:25:02.954 --> 00:25:13.324
the art of writing, but the problem I have is the execs on my team say AI exists, and therefore, I need to churn out three times the amount of content that I did in the past."
167
00:25:13.844 --> 00:25:24.484
I'm curious, Jimmy, as you think about transitioning your process, and you're saying, "I'm comfortable pulling back a little bit," do you foresee that being a challenge, how you communicate that?
168
00:25:24.544 --> 00:25:32.584
Like, "Hey, you know, I've realized this about my process, and I'm gonna deliver you fewer things, but they're gonna be better, and you'll see the value in that."
169
00:25:32.804 --> 00:25:44.644
I think it's an easier case to make if you're launching a Substack, for example. I think it's a harder case to make if you are tasked with creating content meant for SEO or AEO.
170
00:25:44.904 --> 00:25:53.104
Also, though, content marketers have been dealing with short-sighted management forever. And so it's not like a... That's like a n- not really a new problem.
171
00:25:53.404 --> 00:26:05.664
If you're in a situation where someone is saying like, "No, you, you have to double your output," or you're a freelancer and the client is saying like, "We're gonna pay you half as much as we used to," like, that's toxic, in my opinion.
172
00:26:05.704 --> 00:26:12.764
Like, truly toxic. I'm imagining that those same types of things happened five years ago. The context was just a little bit different.
173
00:26:13.084 --> 00:26:26.264
It's relatively easy to show version A that you wrote and version B that Claude wrote, and even without explaining it, I think the difference would be, like, pretty obvious to most people that one is going to be better than the other.
174
00:26:26.324 --> 00:26:35.244
I mean, it depends on, like, the skill of the writer and a million kind of subjective things too, but I think it's pretty easy to beat AI, and I think it's pretty easy to prove that you can beat AI also.
175
00:26:35.344 --> 00:26:45.544
You know, maybe the conversation then is like, "Okay, I can use AI to, like, make everything else faster and easier, but, like, I'm going to protect this one thing because if we don't, then it's like we don't even respect our readers."
176
00:26:45.664 --> 00:26:51.504
I think there's, like, a hard line you can draw there while still leaning on AI to do, like, all kinds of other useful, productive things.
177
00:26:51.544 --> 00:27:01.384
I also think your own personal capacity is something you have to proactively communicate at all times. You can sort of just pull AI usage into that. You know, like, "I tried AI for this.
178
00:27:01.564 --> 00:27:08.764
I really thought we could, like, scale this up to this m- massive scale that I promised. It's not working as I expected, therefore, I have to pull back channel A."
179
00:27:08.944 --> 00:27:16.884
Uh, or like, "We decided to write all of our LinkedIn posts now with AI. They're clearly not working. I, I think we should take that back, sacrifice this other thing we're doing."
180
00:27:17.164 --> 00:27:25.694
I, I don't think any manager is like, "I want you to do these 15 channels, and I want you to use AI for all of them." What they want is business results, right? Mm-hmm. They, they think that that's a means to an end.
181
00:27:25.904 --> 00:27:28.424
If you can communicate that it's not, like, they'll get there.
182
00:27:28.524 --> 00:27:36.524
I just try to sort of keep an open dialogue with, with our CEO at Dock on, like, "Look at this cool thing I did with AI," but then you're like, "Ooh, look at how badly this failed."
183
00:27:36.564 --> 00:27:44.604
And I, I think, like, trickling it- [laughs]... into, like, day-to-day conversations is working better than, like, making grand statements about, like, when I'm going to use AI or not.
184
00:27:44.764 --> 00:27:54.684
I, I don't know if that's, like, a standard relationship people have with it. Maybe this is a hot take, but a good leader is going to respect you asking, "Why are we using this?
185
00:27:54.904 --> 00:28:09.524
What is the goal here, and what is, what is your goal? Why are you having me use this?" I think a lot of teams now are really trying to figure out, how do we actually get a return out of these tools that we're using?
186
00:28:09.604 --> 00:28:21.804
And this is kind of how you figure that out. Don't create content just to create content. Are you really gonna post that just to post that? Are you gonna use AI just to use AI to get that up? Come on.
187
00:28:22.084 --> 00:28:36.964
[laughs] At some point, this is all gonna change because OpenAI and Antropic will go public, they'll raise their prices, and then every company's gonna be like, "We have to rein in our token spend because this is insanely expensive, and the company is not any more profitable than it used to be."
188
00:28:37.694 --> 00:28:47.424
[laughs] And then we'll all have this, like, much easier case. And they, they're already doing that. Yeah, totally. Like- Yeah... like, I have zero sense how much money I have spent on Miro's behalf this summer.
189
00:28:47.584 --> 00:28:54.664
Not a clue. And like- You don't wanna, and you don't wanna know. [laughs] I don't really, I don't really, I don't really care, but I bet it's a lot. [laughs] You know?
190
00:28:54.824 --> 00:29:06.804
I mean- Like, companies are now having to look at, like, okay great, that list pull that just happened was actually so much more money than it could've been had we set parameters on which models our teams are using.
191
00:29:06.944 --> 00:29:10.184
But I, I don't wanna get too technical here. [laughs] Yeah, yeah.
192
00:29:10.224 --> 00:29:18.124
My, my brother's a software engineer at Apple, and he was saying that his boss told them, "You are not allowed to use Claude Code to change the color of anything.
193
00:29:18.424 --> 00:29:27.984
You have to go into the code and change the hex code because we are burning a ton of tokens with these tiny little requests that would probably just be easier to go and do manually anyways."
194
00:29:28.044 --> 00:29:31.084
Retraining them a little bit to then, like, use less of it. It's weird.
195
00:29:31.244 --> 00:29:41.204
I think it's ironic because AI was kind of initially, and I think still is, question mark, branded as this tool that, like, you're not gonna have to do any tedious, menial work anymore.
196
00:29:41.244 --> 00:29:50.164
You're just gonna be able to work on these amazing, big, blue sky projects. So it's ironic to tell your senior devs- That would be the tedious work... go in there and change all the hex codes manually.
197
00:29:50.174 --> 00:29:56.754
Like, you can't get any more tedious than that. [laughs] These gi- these giant areas where, like, it just makes everything so much easier, and then there's these- Mm-hmm...
198
00:29:56.754 --> 00:30:05.184
other giant areas where it just makes everything so much harder. Mm-hmm. Like, so much harder. In this, in this way that I feel like, at least for me, like, I'm only just now appreciating.
199
00:30:05.244 --> 00:30:13.004
As my, like, excitement wanes a bit, I'm like, "Oh, I actually hate doing all these little tedious things that I do as a result of AI usage."
200
00:30:13.144 --> 00:30:22.164
I guess I've been there for a bit now, but I'm kinda only just, like, kind of, uh, like waking up. There's always going to be tedium and suffering in the work that we do.
201
00:30:22.264 --> 00:30:28.664
We can try to escape that, and we can try to figure out how to do it, but, like, that's just kinda how work is.
202
00:30:28.853 --> 00:30:36.314
It's kind of about figuring out where we wanna apply that suffering and also, like, where that suffering counts for something.
203
00:30:36.684 --> 00:30:47.004
Most of us feel happy to do hard work that does something, but if it's just, like, making all these hex codes, you know? Like, that's not purposeful suffering.
204
00:30:47.444 --> 00:30:55.554
Your big idea in this piece is, is like, direct your suffering to a place where it will matter, where it will be meaningful and purposeful. Love that. I love that.
205
00:30:55.604 --> 00:31:04.324
I actually feel like that's, like, a perfect place to end. If you wanna come and suffer with us- [laughs]... join Superpath Pro. Join Superpath Pro to suffer purposefully with us.
206
00:31:04.444 --> 00:31:09.164
I would say the Reading List channel has been really good lately. Ronnie Fiegens, today or yesterday, dropped an article.
207
00:31:09.244 --> 00:31:19.544
If you wanna, like, really get fired up about AI and writing, go to the Reading List channel and read what Ronnie posted about the Claude watermarking thing. It's wild.
208
00:31:19.654 --> 00:31:22.044
We could do a whole episode on Claude watermarking alone.
209
00:31:22.084 --> 00:31:32.524
But anyways, we're all, like, dealing with the same, more or less the same things in our day-to-day work, and it's an awesome place, um, to see what other people are doing and also see what they're frustrated with and vent about it and then, uh, get support.
210
00:31:32.584 --> 00:31:42.104
That's like a, kind of a microcosm of what's happening in the community. Jimmy, thanks for joining us. Rachel, Ruth, Eric, thanks for being here. Talk to y'all soon. Always a great time. Bye, guys. Bye. Thanks, guys.
211
00:31:42.114 --> 00:31:49.204
See you. [outro music]
