1
00:00:00,000 --> 00:00:02,759
Welcome to Top 5 Tech, my
weekly show about what to watch,

2
00:00:02,760 --> 00:00:05,220
listen to, read, download,
and something deep every week.

3
00:00:05,280 --> 00:00:08,999
You've probably heard about breaches
in cybersecurity and AI agents,

4
00:00:09,000 --> 00:00:11,579
and one of the biggest was
the hugging face breach,

5
00:00:11,580 --> 00:00:15,539
where open AI models broke containment
and actually hacked the hugging face

6
00:00:15,540 --> 00:00:18,360
distribution, which is like
a marketplace for AI models.

7
00:00:18,600 --> 00:00:23,340
Well, Kurgazot, in a nutshell, the YouTube
channel has an incredible explainer video.

8
00:00:23,400 --> 00:00:26,999
It's called AI just became
humanity's biggest threat,

9
00:00:27,000 --> 00:00:30,839
and we'll get more into the AI doomsday
stuff in the deep section of this episode,

10
00:00:30,840 --> 00:00:34,979
but if you really want to understand what
happened in that situation with hugging

11
00:00:34,980 --> 00:00:39,299
face, open AI models, how they broke
containment, what that even means,

12
00:00:39,300 --> 00:00:42,359
how the models communicated
and work together,

13
00:00:42,360 --> 00:00:44,580
this video is an incredible
explainer all about it.

14
00:00:44,880 --> 00:00:47,039
I've already watched it
once in a half times,

15
00:00:47,040 --> 00:00:49,379
probably going to watch it a couple
more times to really understand what's

16
00:00:49,380 --> 00:00:51,300
happening, but highly,
highly recommend watching.

17
00:00:51,540 --> 00:00:52,540
On what to listen to.

18
00:00:52,560 --> 00:00:54,180
I wanted to plug this podcast.

19
00:00:54,240 --> 00:00:57,239
It's called Creator After Hours,
and he invited me to be on,

20
00:00:57,240 --> 00:01:00,779
interviewed me about leaving Riverside,
going off on my own as an independent

21
00:01:00,780 --> 00:01:03,839
creator. I thought it was actually a
really great conversation He just launched

22
00:01:03,840 --> 00:01:07,440
this podcast. It's on YouTube, but you can
watch or listen to it in the background.

23
00:01:07,500 --> 00:01:08,580
There's not a ton of visuals.

24
00:01:08,640 --> 00:01:12,779
So just want to plug that, go listen to my
interview with Justin on the creator after

25
00:01:12,780 --> 00:01:14,520
hours. And what to download?

26
00:01:14,640 --> 00:01:16,740
I've I've coded 3 apps this week.

27
00:01:16,860 --> 00:01:20,099
Most of them were actually in a single
day because I have some use cases when I'm

28
00:01:20,100 --> 00:01:23,939
making these kinds of podcasts, videos,
and audio that I just wanted quick,

29
00:01:23,940 --> 00:01:25,860
simple utilities that did
exactly what I wanted.

30
00:01:25,920 --> 00:01:27,840
And all 3 of these apps
are completely free.

31
00:01:28,380 --> 00:01:30,060
The 1st one is called Leveler.

32
00:01:30,120 --> 00:01:32,699
If you were a fan of
Levelator back in the day,

33
00:01:32,700 --> 00:01:36,119
I've actually gotten some feedback from
people on social media that they prefer

34
00:01:36,120 --> 00:01:38,340
the end product that
Leveler actually gives them.

35
00:01:38,760 --> 00:01:42,479
I basically gave Claude an audio
file that I get raw from Riverside,

36
00:01:42,480 --> 00:01:45,539
and one that I edit to
actually sound how I want it,

37
00:01:45,540 --> 00:01:49,680
and let Claude analyze that and build
an audio engine for leveling audio.

38
00:01:50,220 --> 00:01:51,420
Actually does a great job.

39
00:01:51,480 --> 00:01:54,719
The latest episode of Primary
Technology was using the Leveler app,

40
00:01:54,720 --> 00:01:58,439
really helpful for either video or audio
files that you just want to have even

41
00:01:58,440 --> 00:02:00,720
audio throughout with multiple speakers.

42
00:02:00,780 --> 00:02:04,559
It works great. You can go to beardfm.app,
and links to these are all down in the

43
00:02:04,560 --> 00:02:06,120
show notes, but that's Leveler.

44
00:02:06,180 --> 00:02:07,500
I also have Transcriber.

45
00:02:07,620 --> 00:02:11,580
And what's cool about this is it uses
Apple's on-device models on your Mac.

46
00:02:11,820 --> 00:02:13,500
So you don't have to download any models.

47
00:02:13,560 --> 00:02:15,240
This is not going out
to the internet at all.

48
00:02:15,360 --> 00:02:19,259
This could happen totally offline,
but you drop a video or audio file,

49
00:02:19,260 --> 00:02:20,640
and it just transcribes it.

50
00:02:20,820 --> 00:02:23,459
You can correct the transcript
if you need, export a text,

51
00:02:23,460 --> 00:02:28,079
SRT, or VTT file, whatever you need for
subtitles or just your own search and

52
00:02:28,080 --> 00:02:31,680
transcription needs, transcriber is
super fast, on-device works great.

53
00:02:32,100 --> 00:02:34,379
And finally, if you want
to add animated captions,

54
00:02:34,380 --> 00:02:38,339
there's so many apps for this, I was using
the edits app on iPhone for a long time,

55
00:02:38,340 --> 00:02:40,739
which is a great free
app, it's made by meta,

56
00:02:40,740 --> 00:02:45,359
but I wanted to be able to batch process
my reels because I usually make 5 to 7

57
00:02:45,360 --> 00:02:49,680
reels in a day, and I just want to add the
same style animated captions to all of it.

58
00:02:50,100 --> 00:02:52,860
And because I built
Transcriber with Claude.

59
00:02:52,920 --> 00:02:56,699
I asked it to build that same
functionality plus some animated captions,

60
00:02:56,700 --> 00:02:57,700
and I made that totally free.

61
00:02:57,720 --> 00:02:58,800
Again, it's on your Mac.

62
00:02:58,860 --> 00:03:01,020
You can give it one or
multiple video files.

63
00:03:01,260 --> 00:03:03,420
There's lots of different
animated caption options.

64
00:03:03,480 --> 00:03:04,800
You can save your favorite styles.

65
00:03:05,040 --> 00:03:07,260
All of those 3 apps are
completely free to download.

66
00:03:07,320 --> 00:03:09,599
And while you're on this
website, beerfm.app,

67
00:03:09,600 --> 00:03:11,579
you can check out can do
my project management app,

68
00:03:11,580 --> 00:03:15,179
it's a $3 one time purchase, stash, which
I've been using more and more for saving

69
00:03:15,180 --> 00:03:16,260
links, and a ton more.

70
00:03:16,320 --> 00:03:18,480
It's all there. And one other
bonus, what to download.

71
00:03:18,540 --> 00:03:20,340
Listeners, send me this, made some clip.

72
00:03:20,400 --> 00:03:21,780
It is a one-time purchase of $8.

73
00:03:22,200 --> 00:03:26,099
But if you do a lot of short little
calculations and you copy numbers a lot,

74
00:03:26,100 --> 00:03:27,719
maybe you're working
in Excel spreadsheets,

75
00:03:27,720 --> 00:03:31,319
this is a cool little app where it will
automatically calculate the numbers that

76
00:03:31,320 --> 00:03:34,619
you are copying to your clipboard, and it
can just be in a little window off to the

77
00:03:34,620 --> 00:03:37,439
side on your Mac. Pretty cool
app in the Mac App Store,

78
00:03:37,440 --> 00:03:38,940
link is below. And what to read?

79
00:03:39,000 --> 00:03:40,559
I actually have to get
into this this week,

80
00:03:40,560 --> 00:03:45,479
but one of the new automation actions
in iOS 27 shortcuts is the ability to

81
00:03:45,480 --> 00:03:49,680
automate something based on a notification
that arrives from a 3rd party app.

82
00:03:50,220 --> 00:03:54,779
Well, this app called Deflector has kind
of hacked that functionality and made it

83
00:03:54,780 --> 00:03:59,220
so you can run shortcuts by hacking this
notification system and using this app.

84
00:03:59,700 --> 00:04:03,419
This is from MacStories.net, and they go
onto to explain in-depth how to use this

85
00:04:03,420 --> 00:04:05,820
app, how to create these automations.

86
00:04:05,880 --> 00:04:06,839
This is really interesting to me.

87
00:04:06,840 --> 00:04:09,659
I'll probably be doing a video
on this on YouTube pretty soon,

88
00:04:09,660 --> 00:04:12,420
but wanted to link it here and you can
download the app in their article as well.

89
00:04:12,660 --> 00:04:14,219
All right, and finally on something deep,

90
00:04:14,220 --> 00:04:17,880
I watched this entire interview
with Cleo Abram and Alexander Wang.

91
00:04:17,940 --> 00:04:22,139
Alexander Wang is one of the leading
AI researchers and builders of these AI

92
00:04:22,140 --> 00:04:23,219
products and agents.

93
00:04:23,220 --> 00:04:26,159
He was poached by Meta, so
he works at Meta right now,

94
00:04:26,160 --> 00:04:30,059
and he was in large part
responsible for the Muse AI agent,

95
00:04:30,060 --> 00:04:33,600
which has been the number one free app
in the App Store for a couple weeks now.

96
00:04:33,840 --> 00:04:36,120
We've talked about it at
length on primary technology.

97
00:04:36,240 --> 00:04:38,999
My co is Jason Aten,
downloaded it on his Mac,

98
00:04:39,000 --> 00:04:42,420
and it scooped up all his messages
and didn't have permission to do that.

99
00:04:42,780 --> 00:04:46,139
So I would not download
Muse on a Mac or a computer,

100
00:04:46,140 --> 00:04:49,679
but I've been using Muse on my iPhone
because when you download an app to your

101
00:04:49,680 --> 00:04:51,900
iPhone or iPad. It's sandbox.

102
00:04:51,960 --> 00:04:54,659
It can't get out of that app and do
something nefarious with the other data on

103
00:04:54,660 --> 00:04:55,980
your iPhone or iPad.

104
00:04:56,040 --> 00:04:57,060
So I've been using it just there.

105
00:04:57,360 --> 00:05:00,300
And I totally get why this is
one of the most popular apps.

106
00:05:00,720 --> 00:05:03,660
Muse is an AI agent that can
do so much action for you.

107
00:05:04,020 --> 00:05:09,119
And the technology behind it is you have a
virtual machine or a computer in the cloud

108
00:05:09,120 --> 00:05:12,659
that when you ask Muse to do
something, it can use that computer,

109
00:05:12,660 --> 00:05:14,699
and it has nothing to
do with your hardware,

110
00:05:14,700 --> 00:05:17,820
nothing to do with your device in
the app that you're prompting Muse.

111
00:05:17,880 --> 00:05:19,919
It's just something that
it will work in a browser,

112
00:05:19,920 --> 00:05:22,380
in a computer, in a server
somewhere, and do lots of stuff.

113
00:05:22,500 --> 00:05:24,959
For instance, I actually have a
notification from you right here because I

114
00:05:24,960 --> 00:05:28,319
asked it to let me know when the Apple
Home event was going to be officially

115
00:05:28,320 --> 00:05:31,319
announced. There's been rumors that
Apple's going to launch an event on

116
00:05:31,320 --> 00:05:33,599
October 13th, announced
new smart home products,

117
00:05:33,600 --> 00:05:37,079
so I asked Muse to check every hour and
let me know when there's any news about

118
00:05:37,080 --> 00:05:40,020
it, and it let me know right
here, even send me a notification.

119
00:05:40,500 --> 00:05:43,379
You can use it to try and list
things on Facebook Marketplace,

120
00:05:43,380 --> 00:05:46,499
although be wary of that because some
people have had their address given out

121
00:05:46,500 --> 00:05:50,819
automatically, Muse just sent it to a
potential buyer because someone checked

122
00:05:50,820 --> 00:05:51,820
the wrong permission box.

123
00:05:51,840 --> 00:05:53,939
So again, be careful with
some of these use cases,

124
00:05:53,940 --> 00:05:58,560
but I asked it to compare health insurance
plans for someone who's self-employed.

125
00:05:58,740 --> 00:06:01,439
And when those come out during
open enrollment later this fall,

126
00:06:01,440 --> 00:06:05,939
Muse is going to analyze all that data and
give me an easy to understand report about

127
00:06:05,940 --> 00:06:08,760
what I should think about when it
comes to health insurance plans.

128
00:06:09,000 --> 00:06:14,099
I asked it to find mats for my car because
my driver and passenger side mats are kind

129
00:06:14,100 --> 00:06:16,919
of busted, and I found mats at
Home Depot that were affordable,

130
00:06:16,920 --> 00:06:18,779
gave me several options
comparing in price,

131
00:06:18,780 --> 00:06:23,040
but it also fails at some points because
it said it could buy the mats for me.

132
00:06:23,100 --> 00:06:24,420
I said go ahead and buy it.

133
00:06:24,480 --> 00:06:27,179
I linked Stripe and link
for a payment process,

134
00:06:27,180 --> 00:06:29,279
and it couldn't finish
the checkout process,

135
00:06:29,280 --> 00:06:31,979
and it told me to just take
control in a web browser,

136
00:06:31,980 --> 00:06:32,980
but that was really janky.

137
00:06:33,180 --> 00:06:34,620
So it is a mixed bag.

138
00:06:34,680 --> 00:06:37,559
The security and privacy on
a computer is a major issue,

139
00:06:37,560 --> 00:06:40,080
but I totally see the promise of this.

140
00:06:40,500 --> 00:06:42,240
And Alexander Wang in this interview.

141
00:06:42,300 --> 00:06:45,179
He is very optimistic saying
that 10 years from now,

142
00:06:45,180 --> 00:06:48,779
it could basically help people achieve
their hopes and dreams because it could do

143
00:06:48,780 --> 00:06:52,679
all this busy work or work that you
don't need to or find it difficult to do,

144
00:06:52,680 --> 00:06:55,980
like choosing a health insurance plan,
and it can just do all that for you.

145
00:06:56,460 --> 00:07:00,359
I see a small bit of that promise now and
they have definitely made Muse the most

146
00:07:00,360 --> 00:07:02,460
approachable AI agent.

147
00:07:02,520 --> 00:07:04,379
Not only is it kind of fuzzy and cute,

148
00:07:04,380 --> 00:07:07,679
and Met is actually going to launch
a Tamagotchi style Muse pendant,

149
00:07:07,680 --> 00:07:11,279
maybe before the holidays, and it's really
approachable in how it communicates,

150
00:07:11,280 --> 00:07:13,380
being able to ask it things
just a natural language.

151
00:07:13,440 --> 00:07:15,719
This is something I can imagine
telling my mom to try it,

152
00:07:15,720 --> 00:07:19,380
so she can get some information in a
digestible and easy to understand way.

153
00:07:19,740 --> 00:07:21,060
But it is a very fine line.

154
00:07:21,120 --> 00:07:24,959
Again, go back to the AI agent video from
Kyrgazot at the beginning of this episode

155
00:07:24,960 --> 00:07:28,260
about the dangers of AI agents,
and I also see the promise.

156
00:07:28,380 --> 00:07:33,359
And so I am cautiously optimistic, but
also believe that we need to have some

157
00:07:33,360 --> 00:07:37,079
real guardrails in place as these
companies and frontier models move forward

158
00:07:37,080 --> 00:07:39,180
into this new AI agent world.

159
00:07:39,600 --> 00:07:41,460
So just wanted to leave
that interview for you.

160
00:07:41,520 --> 00:07:43,979
We'd love to know what you think
on that, the Kirkasad video,

161
00:07:43,980 --> 00:07:44,939
or anything that I like.

162
00:07:44,940 --> 00:07:48,599
And you can get all of these links with
some more commentary as a newsletter every

163
00:07:48,600 --> 00:07:50,040
Friday. Totally free to sign up.

164
00:07:50,100 --> 00:07:52,859
You can scan this QR code or click
the link down in the show notes,

165
00:07:52,860 --> 00:07:55,259
and you can get the newsletter
every week, reply to it,

166
00:07:55,260 --> 00:07:57,300
and I get all of those
emails and I read everyone.

167
00:07:57,720 --> 00:08:00,779
If you haven't yet, would love and
appreciate a 5 star rating and review in

168
00:08:00,780 --> 00:08:03,900
Apple Podcasts. Helps the show
get discovered by more people.

169
00:08:03,960 --> 00:08:05,160
And thank you so much for
listening and watching.

170
00:08:05,220 --> 00:08:06,220
I'll catch you next time.