1
00:00:00,00 –> 00:00:04,410
Hey everybody, and welcome to the Friday wrap up on Streamline Solopreneur, a short

2
00:00:04,410 –> 00:00:08,580
episode where I cover three things, what’s on my mind this week, recommended reading

3
00:00:08,580 –> 00:00:13,40
and something fun. This is the show that helps overwhelmed coaches, consultants,

4
00:00:13,40 –> 00:00:18,350
and service providers build systems so their business doesn’t depend on them. Hopefully

5
00:00:18,350 –> 00:00:22,550
this week’s curation will get you thinking about your own systems before you close

6
00:00:22,550 –> 00:00:27,210
your laptop for the weekend. I’m your host, Joe Casabona, and here’s what’s on my

7
00:00:27,210 –> 00:00:30,530
mind for August 14th, 2026.

8
00:00:30,530 –> 00:00:37,740
Am I wrong about AI generated YouTube thumbnails? You likely know how I feel about

9
00:00:37,740 –> 00:00:41,190
using AI for creative work. I don’t like it.

10
00:00:41,190 –> 00:00:46,110
I fear too many business owners, including myself at times, suffer from ai, brain

11
00:00:46,110 –> 00:00:50,640
rot, from relying too much on it for certain tasks

12
00:00:50,640 –> 00:00:56,520
for a while now, one of those tasks was creating YouTube thumbnails. I worry they

13
00:00:56,520 –> 00:01:00,510
don’t look real. And some I’ve seen look nothing like me.

14
00:01:00,510 –> 00:01:06,660
I am so bullish on this that I’ve told my thumbnail designer to never use AI on them,

15
00:01:06,900 –> 00:01:12,210
especially when it comes to images of me or my brand assets like my hat.

16
00:01:12,210 –> 00:01:12,930
But maybe

17
00:01:12,930 –> 00:01:14,930
that doesn’t matter to get the click.

18
00:01:14,930 –> 00:01:21,320
I mean, it seems to me creating YouTube thumbnails is closer to a science than an

19
00:01:21,320 –> 00:01:27,210
art. There are countless tools that will evaluate a thumbnail and score it based

20
00:01:27,210 –> 00:01:28,70
on how

21
00:01:28,70 –> 00:01:29,840
clickable it is.

22
00:01:29,840 –> 00:01:35,330
If all we’re trying to do is appease an algorithm, then what’s wrong with using a

23
00:01:35,330 –> 00:01:37,160
computer program to do it?

24
00:01:37,160 –> 00:01:39,260
But the thing is, we’re not right.

25
00:01:39,260 –> 00:01:44,900
We’re trying to convince human beings to think our video is worth watching.

26
00:01:44,900 –> 00:01:50,450
And some AI thumbnails are simply God awful. I mean, I will link this in the show

27
00:01:50,450 –> 00:01:54,470
notes, but if you’re watching the video, look at this one.

28
00:01:54,470 –> 00:02:02,690
The most generous opinion I could muster of this is that it’s a convincing projection

29
00:02:03,20 –> 00:02:09,590
of what I might look like in 10 to 15 years, which is very similar to my dad,

30
00:02:09,590 –> 00:02:19,70
but the dead eyes absolutely kill me. And to me, this does not look like me so much

31
00:02:19,130 –> 00:02:22,460
that I don’t want to use it on my channel.

32
00:02:22,460 –> 00:02:23,420
But again,

33
00:02:23,420 –> 00:02:24,770
does it matter?

34
00:02:24,770 –> 00:02:28,970
I know it doesn’t look exactly like me, but it is close enough,

35
00:02:28,970 –> 00:02:34,370
and if it convinces people to click, then it’s done the job, right? I mean, if you

36
00:02:34,370 –> 00:02:39,680
look at Mr. Beast’s thumbnails, they look like a super weird version of him. But

37
00:02:40,70 –> 00:02:46,740
all of this thumbnail optimization and the type of thumbnails we create is because

38
00:02:46,950 –> 00:02:47,730
of him.

39
00:02:47,730 –> 00:02:50,80
I am still unsure.

40
00:02:50,410 –> 00:02:55,540
I am experimenting with them. I am using them on my channel. I’m usually ab testing

41
00:02:55,540 –> 00:02:58,660
them against what my thumbnail designer

42
00:02:58,660 –> 00:03:01,870
creates, and I’m not getting enough hits to have anything conclusive.

43
00:03:01,870 –> 00:03:02,320
But

44
00:03:02,320 –> 00:03:05,710
while I believe in strong opinions loosely held,

45
00:03:05,710 –> 00:03:09,250
I might be holding onto this one a little tighter.

46
00:03:09,250 –> 00:03:14,590
I’d love to know what you think about this. Let me know@streamlinedfeedback.com.

47
00:03:15,10 –> 00:03:20,590
You can write in or you can leave a voice note. If you leave a voice note, I might

48
00:03:20,590 –> 00:03:24,310
play it on the show unless you ask me to keep it anonymous.

49
00:03:24,310 –> 00:03:28,300
So that’s what’s on my mind for this week. Let’s move over to recommended reading,

50
00:03:28,300 –> 00:03:31,60
which is another thing about ai. It is

51
00:03:31,60 –> 00:03:36,250
called Super Intelligence is a dragon. I suspect you may have seen this. Basically

52
00:03:36,250 –> 00:03:42,760
everybody I know who’s doing curation this week has mentioned this. Over on platformer,

53
00:03:42,760 –> 00:03:49,960
Casey Newton has a great article about Mark Zuckerberg’s AI manifesto. And I, I do

54
00:03:49,960 –> 00:03:54,280
wanna give a shout out to my friend and listener, Liz, for sending this to me.

55
00:03:54,280 –> 00:03:55,660
The truth is,

56
00:03:55,660 –> 00:04:00,370
anything that references Game of Thrones is good in my book, but the analogy between

57
00:04:00,370 –> 00:04:06,370
dragons and AI or dragons and super intelligence that Newton makes is salient

58
00:04:06,370 –> 00:04:08,530
and very clever.

59
00:04:08,530 –> 00:04:12,820
I’m not gonna spoil any more of it. I strongly recommend you read this.

60
00:04:12,820 –> 00:04:19,120
He’s also uniquely positioned to make the points he does. As a trustworthy, independent

61
00:04:19,120 –> 00:04:22,840
journalist and someone close to the tech industry,

62
00:04:22,840 –> 00:04:28,780
I tend to default to what’s the ulterior motive of this billionaire. And it’s nice

63
00:04:28,780 –> 00:04:35,830
that Newton is informed enough to state it so plainly. So I will have that as a link

64
00:04:35,830 –> 00:04:40,960
in the show notes along with everything I’m talking about as usual. But I strongly

65
00:04:40,960 –> 00:04:44,890
recommend you read this. I’m a subscriber. I’m not a paid subscriber. I’m very close

66
00:04:44,890 –> 00:04:45,190
though.

67
00:04:45,190 –> 00:04:47,800
but I am subscribed to Platformer now and,

68
00:04:47,800 –> 00:04:51,70
I’ve admired Casey Newton’s work for a while.

69
00:04:51,70 –> 00:04:58,300
Now, moving to the exact opposite of ai, something fun is Apple TV’s Show Schmid

70
00:04:58,300 –> 00:05:05,770
Doon. I finished this far too short series mdo earlier this week, and I thought it

71
00:05:05,770 –> 00:05:08,500
was delightful. I’m also a couple of years late to the party.

72
00:05:08,500 –> 00:05:15,250
It’s two seasons, six episodes long. Each season is about the length of a musical,

73
00:05:15,250 –> 00:05:20,620
right? So each season is six episodes, 12 episodes total, and it’s fun and catchy

74
00:05:20,620 –> 00:05:23,410
and a testament to great talent.

75
00:05:23,410 –> 00:05:28,250
It was also a really good palette cleanser after House of the Dragon wrapped up.

76
00:05:28,250 –> 00:05:32,600
The worst part about the show is that it didn’t get renewed for season three, which

77
00:05:32,600 –> 00:05:38,450
I read. They had like 25 songs written, and they were gonna focus on musicals of

78
00:05:38,450 –> 00:05:44,00
the eighties and nineties, like Les Mis, which I love Les Mis. So I, I really would’ve

79
00:05:44,00 –> 00:05:51,710
loved this. If you like musicals, satire, and yelling, Hey, that person’s in this,

80
00:05:51,710 –> 00:05:56,30
this is a perfect show for you. It’s available on Apple tv.

81
00:05:56,30 –> 00:06:00,980
Okay, so something to help you with your systems. This is formerly called Automation

82
00:06:00,980 –> 00:06:04,760
of the Week, but I’m deciding to expand the scope of it.

83
00:06:04,760 –> 00:06:11,60
this week’s something to help your systems is call debriefing. I have a Claude skill

84
00:06:11,120 –> 00:06:17,930
called Daily Call Debrief, and it does a few things. It runs daily and it gathers

85
00:06:17,930 –> 00:06:24,140
any calls I had recorded in Google Meet or Fathom. It summarizes the calls and adds

86
00:06:24,140 –> 00:06:26,390
those summaries to obsidian,

87
00:06:26,390 –> 00:06:32,600
it adds my action items to Todoist, which surfaces them in my task manager. Something

88
00:06:32,600 –> 00:06:36,200
I check every day. And something I added this week

89
00:06:36,200 –> 00:06:40,640
is it adds coaching calls and discovery calls the transcripts,

90
00:06:40,640 –> 00:06:46,700
it, it takes them and adds them to my voice of customer folder so I can pull from

91
00:06:46,700 –> 00:06:51,260
them. When I’m working on marketing copy or positioning or anything like that.

92
00:06:51,260 –> 00:06:53,120
AI call recorders

93
00:06:53,120 –> 00:06:56,270
is one of the great use cases

94
00:06:56,270 –> 00:06:58,460
for AI in my opinion

95
00:06:58,460 –> 00:07:04,10
because they are recording the call. They can crunch data or allows you to crunch

96
00:07:04,10 –> 00:07:08,300
data and they do a bunch of stuff with that data. You can do a bunch of stuff with

97
00:07:08,330 –> 00:07:08,960
that data

98
00:07:08,960 –> 00:07:15,230
and since you were on the call and present, presumably you have already consumed

99
00:07:15,410 –> 00:07:21,500
the primary source and you’re not just relying on the summary for important details.

100
00:07:21,500 –> 00:07:26,540
One thing I like to have AI do as well, something I got from,

101
00:07:26,540 –> 00:07:33,140
Christian who was a guest on this podcast back when I had guests, is have it come

102
00:07:33,140 –> 00:07:37,910
up with questions based on the calls. And here’s the exact prompt Christian gives

103
00:07:37,910 –> 00:07:41,30
listeners. I will put this in the show notes in the description as well.

104
00:07:41,30 –> 00:07:46,40
It goes like this. I want you to pose to me three extremely precise, very hard hitting

105
00:07:46,40 –> 00:07:51,470
questions, all of which are independent and have different coverage that if sufficiently

106
00:07:51,470 –> 00:07:56,960
answered, would yield some kind of transformative, genuinely novel, impactful insight.

107
00:07:56,960 –> 00:08:00,840
The thing that I added was, please make reference to the primary source material

108
00:08:00,840 –> 00:08:03,30
where appropriate in the context of the question.

109
00:08:03,30 –> 00:08:04,50
’cause I don’t trust ai.

110
00:08:04,50 –> 00:08:08,160
and I, I actually ran this this morning as I was putting the show together and it

111
00:08:08,160 –> 00:08:13,650
asked me a question that was so accurate, it physically hurt me. So,

112
00:08:13,650 –> 00:08:15,30
big fan of this.

113
00:08:15,300 –> 00:08:16,300
if you

114
00:08:16,300 –> 00:08:21,820
are thinking about your systems and you don’t know where to start,

115
00:08:21,820 –> 00:08:29,50
I have a really simple eight question quiz. It should take 60 seconds to help you

116
00:08:29,80 –> 00:08:29,830
figure out

117
00:08:29,830 –> 00:08:31,960
where to start and exactly how to start.

118
00:08:31,960 –> 00:08:40,120
You can find that over at streamlined fm slash quiz. But that’s it for this episode

119
00:08:40,330 –> 00:08:46,00
of the Friday wrap up. Thanks so much for listening, and until next time, I hope

120
00:08:46,00 –> 00:08:48,400
you find some space in your weekend.