Episode 260: Tennessee Trickshot

1
00:00:00,940 --> 00:00:07,240
Podcasting 2.0 for May 15, 2026, episode 260, Tennessee Trickshot.

2
00:00:09,099 --> 00:00:13,119
Hello everybody, welcome to Podcasting 2.0 after a one-week hiatus.

3
00:00:13,500 --> 00:00:18,640
We are back, that's right. Everything going on in podcasting,

4
00:00:18,640 --> 00:00:20,859
and it seems to be quite a lot, is happening

5
00:00:20,859 --> 00:00:23,359
right here. I'm Adam Curry in the heart of the

6
00:00:23,359 --> 00:00:26,420
Texas Hill Country. Oh wait, I should have mentioned, we

7
00:00:26,420 --> 00:00:28,980
are the only boardroom that operates on the fringes of

8
00:00:28,980 --> 00:00:31,500
podcasting. There you go. I'm Adam Curry here in the

9
00:00:31,500 --> 00:00:33,920
heart of the Texas field country and in Alabama. The

10
00:00:33,920 --> 00:00:36,560
man who has a data set tailored just for you.

11
00:00:36,880 --> 00:00:40,219
Say hello to my friend on the other end. Dave

12
00:00:40,219 --> 00:00:51,719
Jones! Yeah, this... Yes? I have... Currently, I just have

13
00:00:51,719 --> 00:00:54,799
no idea what I'm doing right now. Are you out

14
00:00:54,799 --> 00:01:00,039
of control? Are you unhinged? Are you completely out of

15
00:01:00,039 --> 00:01:06,200
your gourd? I call it being in dimensions. I am

16
00:01:06,200 --> 00:01:13,560
currently in dimensions at the moment because we got in.

17
00:01:14,180 --> 00:01:24,019
So yesterday, I had made plans. Take your time. It's

18
00:01:24,019 --> 00:01:26,180
okay. It's going to be all right. We'll be fine.

19
00:01:25,739 --> 00:01:30,719
I need chocolate. Yes. Do you have chocolate nearby? This

20
00:01:30,719 --> 00:01:37,319
will help. Okay, there you go. Good. So I left

21
00:01:37,319 --> 00:01:44,719
the office yesterday to come. I'd worked half day, came

22
00:01:44,719 --> 00:01:51,159
home, threw a few things together. My son, my daughter,

23
00:01:51,159 --> 00:01:53,500
and one of my son's friends, we jumped in the

24
00:01:53,500 --> 00:01:59,260
car, drove to Nashville to see the Black Label Society

25
00:01:59,260 --> 00:02:03,420
concert at the Ryman. Black Label Society? I'm not familiar

26
00:02:03,420 --> 00:02:10,080
with this outfit. Oh. Oh. How dare you? How dare

27
00:02:10,080 --> 00:02:14,020
you? Clearly good enough for you to be driving all

28
00:02:14,020 --> 00:02:21,039
night to Nashville. The Black Label Society is Zach Wild's

29
00:02:21,039 --> 00:02:25,340
band. Oh, okay. I know Zach. Yeah. I bet you've

30
00:02:25,340 --> 00:02:27,780
probably interviewed him before. I hung out with him, actually,

31
00:02:27,780 --> 00:02:31,400
in Russia. Oh, yeah, back when he was a drinker?

32
00:02:30,919 --> 00:02:36,979
Oh, we were all drinkers back then. Yeah, he's a

33
00:02:36,979 --> 00:02:40,740
crazy dude, man. But I met him in a Walgreens

34
00:02:40,740 --> 00:02:43,419
one time, which was an interesting story. Okay, I don't

35
00:02:43,419 --> 00:02:46,699
know which one wins, Russia or Walgreens. I think it's

36
00:02:46,699 --> 00:02:50,860
a tie. It's a close call. It's half and half.

37
00:02:50,539 --> 00:02:58,479
So anyway, we drove up there, and it was –

38
00:02:58,479 --> 00:03:01,719
so it was – the show was – I had

39
00:03:01,719 --> 00:03:04,879
scheduled – I bought these tickets back in like January.

40
00:03:04,840 --> 00:03:07,840
And whenever you do things months in advance, you have

41
00:03:07,840 --> 00:03:09,560
no idea what is going to be happening in the

42
00:03:09,560 --> 00:03:12,860
actual timeframe that you – that the event occurs. I

43
00:03:12,860 --> 00:03:15,960
mean like – Yeah. Welcome to my life. Yes. You

44
00:03:15,960 --> 00:03:19,439
never know. You never know. You know this. You're like,

45
00:03:19,439 --> 00:03:21,460
when everything, when things are three months out, you're like,

46
00:03:21,460 --> 00:03:23,159
oh, yeah, it'll be fun. That's great. And then it

47
00:03:23,159 --> 00:03:25,319
shows up. You're like, oh, what? I got to do

48
00:03:25,319 --> 00:03:27,960
what? You look at the calendar. You're like, why did

49
00:03:27,960 --> 00:03:31,979
I ever schedule this? Yes. I know this. Yes. I

50
00:03:31,979 --> 00:03:34,819
know this. And so, but I mean, it was a

51
00:03:34,819 --> 00:03:40,400
killer show because it was three bands, but one Black

52
00:03:40,400 --> 00:03:47,139
Label Society was the headliner. but Zach Sabbath was one

53
00:03:47,139 --> 00:03:54,180
of the intro bands and it's like it's Zach and

54
00:03:54,180 --> 00:03:57,379
it's a trio Zach and two other dudes just doing

55
00:03:57,139 --> 00:04:01,159
the heaviest versions of all the awesome Black Sabbath songs

56
00:04:01,240 --> 00:04:06,460
it is so good did you drop some acid? funny

57
00:04:06,460 --> 00:04:11,900
enough I did not that's usually part of my repertoire,

58
00:04:11,900 --> 00:04:18,319
but I didn't do that this time. But anyway, this

59
00:04:18,319 --> 00:04:22,459
show, man, it started at 7.30 on the nose. It

60
00:04:22,459 --> 00:04:26,579
went forever. It was 11.15 and they were still killing

61
00:04:26,579 --> 00:04:30,879
it. And I'm like, man, we've got to drive three

62
00:04:30,879 --> 00:04:35,360
hours back to Birmingham. We're already pushing to 30 a.m.

63
00:04:35,399 --> 00:04:37,040
And were you in the truck? We've got to go.

64
00:04:37,040 --> 00:04:40,040
Were you in the truck? Oh, thank goodness. No, no,

65
00:04:40,040 --> 00:04:42,259
we were we were in the Honda. And so we

66
00:04:42,259 --> 00:04:45,920
were like, we we get we get going and everything.

67
00:04:45,920 --> 00:04:48,100
We finally get back. So we left with like two

68
00:04:48,100 --> 00:04:51,500
songs left on those on the set list. Got home

69
00:04:51,500 --> 00:04:57,399
about 230. I fell asleep probably about three. And of

70
00:04:57,399 --> 00:05:00,699
course, you know, then my daughter has to get up

71
00:05:00,699 --> 00:05:03,040
and go to school. She has to be there by,

72
00:05:03,040 --> 00:05:06,420
you know, by eight. And my wife is the teacher.

73
00:05:06,259 --> 00:05:10,060
So everybody's up and running. So I've had like two

74
00:05:10,060 --> 00:05:11,939
and a half, three hours of sleep. Yeah, you're running

75
00:05:11,939 --> 00:05:15,300
on fumes, man. Yeah, so then I start getting blown

76
00:05:15,300 --> 00:05:18,759
up from the day job with a problem there. And

77
00:05:18,759 --> 00:05:24,540
on top of that, so we're having to not like

78
00:05:24,540 --> 00:05:27,879
we're doing the board meeting now. And then we have

79
00:05:27,879 --> 00:05:31,279
a Godcaster meeting after this. And then as soon as

80
00:05:31,279 --> 00:05:34,120
the Godcaster meeting is over, we are jumping in the

81
00:05:34,120 --> 00:05:36,920
car and going to North Alabama for a two-day camping

82
00:05:36,920 --> 00:05:40,339
trip. Of which you'll be in the tent sleeping for

83
00:05:40,339 --> 00:05:44,519
most of it. Probably, yes. So, I mean, it is

84
00:05:44,519 --> 00:05:46,100
like – so I'm trying to get as much of

85
00:05:46,100 --> 00:05:48,639
this. We haven't even put anything together for a camping

86
00:05:48,639 --> 00:05:50,500
trip yet, and I'm throwing stuff in bags and all

87
00:05:50,500 --> 00:05:55,319
this kind of stuff this morning. And then I'm going

88
00:05:55,319 --> 00:05:58,879
to rant here, by the way. what so then what

89
00:05:58,879 --> 00:06:08,600
happens the latest ai bullshit who from these stupid ai

90
00:06:08,600 --> 00:06:15,420
security research companies oh nginx the nginx vulnerability yeah it's

91
00:06:15,420 --> 00:06:17,199
called the you know of course you got to give

92
00:06:17,199 --> 00:06:23,379
it a fancy name this one's called nginx rift whatever

93
00:06:23,379 --> 00:06:27,980
i mean you know whatever And so this is an

94
00:06:27,980 --> 00:06:33,339
NGINX vulnerability in the rewrite module of NGINX, where if

95
00:06:33,339 --> 00:06:38,019
you use unnamed regex parameters in your rewrite, which, of

96
00:06:38,019 --> 00:06:45,959
course, almost everybody does. If your NGINX configuration file is

97
00:06:45,959 --> 00:06:52,180
simple, you do not name your parameters in a regex.

98
00:06:52,079 --> 00:06:56,920
If you have an incredibly complicated, you know, 400 line

99
00:06:56,920 --> 00:07:02,199
Nginx config, you probably, maybe you do because you have

100
00:07:02,199 --> 00:07:04,399
to keep your sanity so you'll know what's going on.

101
00:07:04,899 --> 00:07:11,240
But hardly anybody does this. So the vulnerability here is

102
00:07:11,240 --> 00:07:17,339
that somebody can get remote code execution on your Nginx

103
00:07:17,339 --> 00:07:24,519
if you use positional parameters in your regex as part

104
00:07:24,519 --> 00:07:27,319
of a rewrite rule. So that means if you have

105
00:07:27,319 --> 00:07:34,279
a rewrite rule on your Nginx proxy server, which proxying

106
00:07:34,279 --> 00:07:38,800
with Nginx is almost table stakes. Everybody does this. I'm

107
00:07:38,800 --> 00:07:42,300
listening carefully because after the Godcaster meeting, I have some

108
00:07:42,300 --> 00:07:46,120
patching to do. Yes, I bet you do. And so

109
00:07:46,120 --> 00:07:50,639
what happens is if you have a URL request from

110
00:07:50,639 --> 00:07:59,860
a client, from a browser, and the rewrite rule says

111
00:07:59,860 --> 00:08:05,720
is a regular expression, and you do – and you

112
00:08:05,720 --> 00:08:08,220
capture parts of the regular expression so that you can

113
00:08:08,220 --> 00:08:11,100
use them later. So let's say you're capping this and

114
00:08:11,100 --> 00:08:16,019
this is I just patched the podcast index front end

115
00:08:16,019 --> 00:08:22,019
API servers, which were which had this problem. And both

116
00:08:22,019 --> 00:08:23,879
of there's two of them. And I just patched both

117
00:08:23,879 --> 00:08:27,439
hosts. And this would be a common version of this

118
00:08:27,439 --> 00:08:30,759
problem. So you have. So what you want to do

119
00:08:30,759 --> 00:08:34,840
is something like hide the PHP extension. So what you

120
00:08:34,840 --> 00:08:38,419
want to do is on the URL, you would have

121
00:08:38,419 --> 00:08:48,500
something like api.podcastindex.com slash api slash v1 slash podcasts. Okay,

122
00:08:48,500 --> 00:08:54,100
well, podcasts may be a .php file that's doing the serving

123
00:08:54,100 --> 00:08:57,379
of the data, but you don't want to have for

124
00:08:57,379 --> 00:09:01,120
the browser or the API client to have to type .php.

125
00:09:00,639 --> 00:09:06,220
PHP. So you you hide the dot PHP and you

126
00:09:06,220 --> 00:09:09,120
do that with a rewrite rule and your rewrite rule

127
00:09:09,120 --> 00:09:12,740
would look something like this. It would have a URL

128
00:09:12,740 --> 00:09:19,019
path and then you'd have in parentheses dot star and

129
00:09:19,019 --> 00:09:22,860
then after the parentheses dot PHP. So you're capturing the

130
00:09:22,860 --> 00:09:27,940
name of whatever the the PHP script is. And when

131
00:09:27,940 --> 00:09:35,139
you rewrite it, you rewrite it to $1.php. So you're capturing

132
00:09:35,139 --> 00:09:38,039
the name of the thing they're wanting to access, and

133
00:09:38,039 --> 00:09:41,799
you're redirecting it behind the scenes to the name of

134
00:09:41,799 --> 00:09:47,440
the script with the proper .php extension on it. Well, that

135
00:09:47,440 --> 00:09:52,559
is the vulnerability. If you, if you, uh, Nginx does

136
00:09:52,559 --> 00:09:58,379
not, Nginx rewrite module does not handle positional regex parameters

137
00:09:58,379 --> 00:10:04,440
properly. And so you can corrupt the heap and, uh,

138
00:10:04,440 --> 00:10:09,720
and basically put executable code, uh, into the data, into

139
00:10:09,720 --> 00:10:13,340
the data area on the heap. And this is –

140
00:10:13,340 --> 00:10:18,620
and the fix for this is to switch from positional

141
00:10:18,620 --> 00:10:25,980
parameters, which would be like $1, $2, $3, et cetera, to named parameters.

142
00:10:26,279 --> 00:10:32,820
So now you're saying instead of just simply a set

143
00:10:32,820 --> 00:10:36,100
of parentheses that grab the part of the URL that

144
00:10:36,100 --> 00:10:40,460
you're wanting to capture, you would put a – there's

145
00:10:40,460 --> 00:10:43,240
a syntax where you put a name in front of

146
00:10:43,240 --> 00:10:46,320
it. So that later instead, when you go to do

147
00:10:46,320 --> 00:10:52,259
your actual rewrite, you don't do $1.php. You would do $script underscore

148
00:10:52,259 --> 00:10:59,360
name.php or something like that. Yeah. And so this is

149
00:10:59,360 --> 00:11:05,600
– the reason this is a rant is because it

150
00:11:05,600 --> 00:11:08,240
wasn't an issue until someone published it being an issue.

151
00:11:08,820 --> 00:11:13,860
Exactly. Okay, this is exactly my point, and I'm so

152
00:11:13,860 --> 00:11:20,980
sick of this crap because what is happening now –

153
00:11:20,980 --> 00:11:25,240
and everybody is saying versions of, Well, we're not prepared

154
00:11:25,240 --> 00:11:27,899
for this. There's going to be this explosion of AI

155
00:11:27,899 --> 00:11:36,840
model discovered exploits. Okay, fine. Fine, there's going to be.

156
00:11:37,860 --> 00:11:44,419
But these are – what you're seeing, though, are AI

157
00:11:44,419 --> 00:11:52,240
security company startups. Getting PR. Getting PR by releasing proof

158
00:11:52,240 --> 00:11:58,639
of concept code and the more outlandish and headline making

159
00:11:58,639 --> 00:12:03,320
and scary they can make it, the better. Yeah. And

160
00:12:03,320 --> 00:12:05,919
so they're going to drop a proof of concept and

161
00:12:05,919 --> 00:12:09,799
an exploit on the worst possible thing that they can

162
00:12:09,799 --> 00:12:12,980
find. And Selass is right. And then they're going to

163
00:12:12,980 --> 00:12:16,159
just dump it out on Twitter and Hacker News and

164
00:12:16,159 --> 00:12:18,100
everything they can do to get as much PR as

165
00:12:18,100 --> 00:12:25,379
possible. These are exploits just like the two recent Linux

166
00:12:25,379 --> 00:12:31,279
kernel escalation of privilege vulnerabilities. These are exploits that nobody,

167
00:12:31,279 --> 00:12:36,840
no human being was ever going to find. And so

168
00:12:36,840 --> 00:12:43,039
there is no reason to publicly disclose these things. They

169
00:12:43,039 --> 00:12:48,639
should be private for as long as possible, as long

170
00:12:48,639 --> 00:12:57,340
as humanly possible. They should remain non-publicly disclosed. Dream on,

171
00:12:57,340 --> 00:13:00,940
brother. Dream on. But people got to make them some

172
00:13:00,940 --> 00:13:04,539
money. And that's what this is all about. And so

173
00:13:04,539 --> 00:13:10,039
because some Silicon Valley security AI startup needs to get

174
00:13:10,039 --> 00:13:15,480
some more venture capital funding, all of us have to

175
00:13:15,480 --> 00:13:18,860
scramble and panic patch our servers as quickly as possible.

176
00:13:20,799 --> 00:13:27,980
Well, good news. My servers are not vulnerable. Okay. You

177
00:13:27,980 --> 00:13:31,299
have to be sure and check this. I don't know

178
00:13:31,299 --> 00:13:33,080
if this applies to you, but there's a couple of

179
00:13:33,080 --> 00:13:40,940
– to be complete, you must – you got to

180
00:13:40,940 --> 00:13:47,659
check everything. That means Docker containers as well because Docker

181
00:13:47,659 --> 00:13:50,840
containers will be running Nginx as a reverse proxy in

182
00:13:50,840 --> 00:13:54,659
the Docker Compose stack. So there's going to be an

183
00:13:54,659 --> 00:14:01,559
NGINX config somewhere in your Docker Compose layers. I don't

184
00:14:01,559 --> 00:14:04,340
use Docker. Okay, well, then you should be fine. But

185
00:14:04,340 --> 00:14:08,820
if somebody uses Docker, do not forget to check those

186
00:14:08,820 --> 00:14:13,639
images because they can have a vulnerable NGINX config. And

187
00:14:13,639 --> 00:14:19,120
now, since we've been gifted with Linux privilege escalation vulnerabilities,

188
00:14:19,120 --> 00:14:22,940
you can remote code exploit, chain it together with dirty

189
00:14:22,940 --> 00:14:25,559
frag or whatever they're calling that thing, and you got

190
00:14:25,559 --> 00:14:30,139
root on the host in about three seconds. Thank you,

191
00:14:30,139 --> 00:14:35,139
AI. Well, actually, it's not. Thank you, Silicon Valley bullcrap

192
00:14:35,139 --> 00:14:38,899
is what I should be saying. Yeah. I feel your

193
00:14:38,899 --> 00:14:44,460
pain. I do. I find it. hilarious that I literally

194
00:14:44,460 --> 00:14:46,600
asked Claude Code to go check if I had this

195
00:14:46,600 --> 00:14:52,600
vulnerability. Are you sure that Claude Code did not nuke

196
00:14:52,600 --> 00:14:56,559
your repo while it was checking for the vulnerability? No.

197
00:14:56,299 --> 00:14:59,000
Because this seems to be a common occurrence. No, it

198
00:14:59,000 --> 00:15:04,039
didn't. It didn't. But before we get to that, big

199
00:15:04,039 --> 00:15:10,840
shout out to RSS.com for dropping off. What did you

200
00:15:10,840 --> 00:15:16,320
get, a Spark RDX? Okay, yes, they sent us a

201
00:15:16,320 --> 00:15:23,039
DGX Spark. A DGX Spark, yes. This is the, let

202
00:15:23,039 --> 00:15:27,419
me see, this is the ASUS version. DGX Spark is

203
00:15:27,419 --> 00:15:30,100
a reference design, so this is the ASUS version called

204
00:15:30,100 --> 00:15:39,919
the GB10. 128 gigs of unified RAM, one terabyte of

205
00:15:39,919 --> 00:15:46,639
NVMe, and built-in Grace Blackwell GPU. And have you set

206
00:15:46,639 --> 00:15:52,399
it up yet? It's soaking in it right now. What

207
00:15:52,399 --> 00:15:59,620
a wonderful, valuable gift donation that was. Oh, yeah. This

208
00:15:59,620 --> 00:16:02,539
is a $3,500 box. I mean, and I was on the

209
00:16:02,539 --> 00:16:05,840
email thread, and Alberto was like, hey, can I send

210
00:16:05,840 --> 00:16:08,539
this your way? And you're like, well, if it's like

211
00:16:08,539 --> 00:16:11,480
this. And he says, yeah, it'll be on your desk

212
00:16:11,480 --> 00:16:16,799
tomorrow. It was. That was just, I mean, I couldn't

213
00:16:16,799 --> 00:16:22,860
believe how kind that was. Yeah. It has already made

214
00:16:22,860 --> 00:16:27,679
a humongous difference because we went from – there's two

215
00:16:27,679 --> 00:16:34,159
things that we're doing with this box. It's currently running

216
00:16:34,159 --> 00:16:39,360
the inference for the spam classifier, the slop spam classifier.

217
00:16:40,659 --> 00:16:43,860
And so that's running – oh, it's running so much

218
00:16:43,860 --> 00:16:46,720
better now. I was limited to a 9B model. Now,

219
00:16:46,720 --> 00:16:48,820
do you want to just talk about this for a

220
00:16:48,820 --> 00:16:53,279
second? Because a lot has come out of this. Things

221
00:16:53,279 --> 00:16:56,419
happen very, very quickly. My head was spinning because now

222
00:16:56,419 --> 00:16:58,440
I don't know if you already were doing the data

223
00:16:58,440 --> 00:17:01,779
sets or if this accelerated the data sets. But the

224
00:17:01,779 --> 00:17:06,319
data sets then ship a data set that Alberto turned

225
00:17:06,319 --> 00:17:11,519
into the PodRoll Atlas, which is a visual explorer of

226
00:17:11,519 --> 00:17:20,160
the entire podcast recommendation universe at atlas.rss.io. So what are

227
00:17:20,160 --> 00:17:23,720
these data sets and are they coming out of this

228
00:17:23,720 --> 00:17:30,200
new system? Just give us a little background here. so

229
00:17:30,200 --> 00:17:34,920
yeah it's amazing how much stuff can happen in a

230
00:17:34,920 --> 00:17:38,920
week i actually i tagged them all in the show

231
00:17:38,920 --> 00:17:40,559
notes people go take a look at it i mean

232
00:17:40,559 --> 00:17:46,559
it's unbelievable what you're doing well all right so we've

233
00:17:46,559 --> 00:17:50,900
had you know occasionally over the years we will i'll

234
00:17:50,900 --> 00:17:54,759
i'll either have a request for something or i myself

235
00:17:54,759 --> 00:17:59,960
will want a particular view into the podcast index data.

236
00:18:01,259 --> 00:18:04,680
And what will come out of that are things like

237
00:18:04,759 --> 00:18:09,400
the music, you know, top 100 or whatever. Right, right,

238
00:18:09,400 --> 00:18:18,339
right. And the dead feeds list. These various sets of

239
00:18:18,339 --> 00:18:24,059
data. But what has happened over the years is I'll

240
00:18:24,059 --> 00:18:27,960
just publish these one-off and there's never been a place

241
00:18:27,960 --> 00:18:33,420
to like find them all in one easy location. And

242
00:18:33,420 --> 00:18:37,519
so I looked at the website and we had this

243
00:18:37,519 --> 00:18:42,559
top level menu item called stats and it was pretty

244
00:18:42,079 --> 00:18:45,720
much useless. It was just another version of the same

245
00:18:45,720 --> 00:18:47,660
stat that was on the homepage. I'm like, we have

246
00:18:47,660 --> 00:18:52,440
this valuable menu space up there at the top. That

247
00:18:52,440 --> 00:18:55,680
should become the page where all the public, all these

248
00:18:55,680 --> 00:18:58,519
public data sets that have come out over the years,

249
00:18:59,460 --> 00:19:03,460
those need to be all listed in one location. And

250
00:19:03,460 --> 00:19:07,579
so I just, you know, I don't understand React. I'm

251
00:19:07,579 --> 00:19:13,519
not a React programmer. Here's the JSON. Enjoy. Vibe code.

252
00:19:13,640 --> 00:19:15,619
I just Vibe coded the heck out of this thing.

253
00:19:15,619 --> 00:19:21,359
That's great. Yeah. And so we have all live feeds.

254
00:19:21,359 --> 00:19:27,119
That's a SQLite file. So dead feeds, problematic feeds, recommendations,

255
00:19:27,119 --> 00:19:32,619
all media, video, music, value-enabled feeds, hosting provider stats. I

256
00:19:32,619 --> 00:19:36,539
mean, this is fantastic. These are all things that already

257
00:19:36,539 --> 00:19:40,380
existed, but just nobody knew where to find them. And

258
00:19:40,380 --> 00:19:43,680
they get updated constantly, but there was never an easy

259
00:19:43,680 --> 00:19:46,700
place to discover them. The hosting provider stats is a

260
00:19:46,700 --> 00:19:50,900
good one. That's a list of all the feeds, all

261
00:19:50,900 --> 00:19:55,539
the live feeds, non-dead feeds in the index, broken down

262
00:19:55,539 --> 00:20:02,099
by which hosting company they come from. So that's a

263
00:20:02,099 --> 00:20:07,539
big list there. And I do a lot of regexing

264
00:20:07,539 --> 00:20:16,660
in this particular data set to massage the different –

265
00:20:18,019 --> 00:20:22,460
RSS feed generators into something that actually is legit. So

266
00:20:22,460 --> 00:20:25,980
like Blueberry is a great example of hosting companies. It's

267
00:20:25,980 --> 00:20:31,460
very difficult to get a handle on their actual feed

268
00:20:31,460 --> 00:20:36,160
count because they have lots of different white like white

269
00:20:36,160 --> 00:20:40,740
label services and things that they do. And the feed

270
00:20:40,740 --> 00:20:43,319
may be a blueberry hosted feed, but it doesn't show

271
00:20:43,319 --> 00:20:47,740
up under a blueberry domain or a blueberry generator. And

272
00:20:47,740 --> 00:20:51,000
so a few years ago, we worked with them and

273
00:20:51,000 --> 00:20:55,599
a few other providers that – hosting companies that have

274
00:20:55,599 --> 00:20:58,859
the same issue to get them to put some better

275
00:20:58,859 --> 00:21:02,420
tagging in their feeds. And so now it's not actually

276
00:21:02,420 --> 00:21:06,119
in the generator tag. It's in a comment in the

277
00:21:06,119 --> 00:21:09,480
RSS feed. And so then I pull that in and

278
00:21:09,480 --> 00:21:12,859
then properly attribute that to Blueberry. So I feel like

279
00:21:12,859 --> 00:21:20,000
these host hostings stats are actually probably the most accurate

280
00:21:20,000 --> 00:21:24,619
anywhere. I feel like they're very, very accurate to actually

281
00:21:24,619 --> 00:21:28,660
who is producing that feed. And then 24 hour feed

282
00:21:28,660 --> 00:21:37,279
report. that's just an overview of a quick dirty breakdown

283
00:21:37,279 --> 00:21:40,519
of what we think the last 24 hours of added

284
00:21:40,519 --> 00:21:45,119
feeds have been whether like spam or what kind of

285
00:21:45,119 --> 00:21:48,920
quick dirty classification, just a raw dump of the newly

286
00:21:48,920 --> 00:21:52,000
added feeds over the last 24 hours in CSV format.

287
00:21:54,240 --> 00:21:58,160
Hourly counts is actually quite a few people use this.

288
00:21:58,160 --> 00:22:02,900
This is things like total feed count, episode count, feeds

289
00:22:02,900 --> 00:22:05,000
with new episodes in the last three days, seven days,

290
00:22:05,000 --> 00:22:08,420
10 days, 14 days, blah, blah, blah, blah, blah. That's

291
00:22:08,059 --> 00:22:12,279
very useful. I know Daniel is using that. A lot

292
00:22:12,279 --> 00:22:17,880
of people do. John Spurlock. Using what? The hourly and

293
00:22:17,880 --> 00:22:23,279
daily counts. Ah, yeah. Stats tracking, yeah. Then there's a

294
00:22:23,279 --> 00:22:27,460
one healthy Hive RPC nodes list. So if you're doing

295
00:22:27,460 --> 00:22:33,099
work on the Hive blockchain, this, it's actually this list

296
00:22:33,099 --> 00:22:39,059
is shrinking by the day. These are the... Why is

297
00:22:39,059 --> 00:22:42,859
that? Well, I mean, I don't know. I can't speak

298
00:22:42,859 --> 00:22:48,599
to the Hive network, but maybe it's struggling, but this

299
00:22:48,599 --> 00:22:52,259
used to have five API servers in it. How's our

300
00:22:53,160 --> 00:22:58,880
gossiper coming? Do we have enough nodes up? It's about,

301
00:22:58,880 --> 00:23:02,180
well, let me check. Actually, you know what? There's a

302
00:23:02,180 --> 00:23:03,859
thing that is not on this list, and that is

303
00:23:03,859 --> 00:23:07,259
the pee-pee monitor. Oh, pee-pee, pee-pee monitor, pee-pee. You've got

304
00:23:07,259 --> 00:23:10,660
to monitor your pee-pee, pee-pee, pee-pee, pee-pee. I need to

305
00:23:10,660 --> 00:23:14,000
check that and see how many hosts we have in

306
00:23:14,000 --> 00:23:17,019
the Gossip Swarm. We currently have a total of 21.

307
00:23:17,920 --> 00:23:20,839
Oh, that's not bad. That's better than I thought it

308
00:23:20,839 --> 00:23:22,960
would be. 18, I do. Nice. Oh, I see. Oh,

309
00:23:22,960 --> 00:23:27,960
I see the link here. peepee monitor these things uh

310
00:23:27,960 --> 00:23:32,299
this but do you vibe code this too the oh

311
00:23:32,299 --> 00:23:34,220
yeah this website oh yeah i would never put this

312
00:23:34,220 --> 00:23:38,059
yeah this is great so with a with a graphic

313
00:23:38,059 --> 00:23:41,000
like that that is never me that is not me

314
00:23:41,000 --> 00:23:45,279
that's so cool um i like hey citizen bathroom that's

315
00:23:45,279 --> 00:23:51,640
one of my favorites yeah live live from the bathroom

316
00:23:53,599 --> 00:23:56,500
do you put your uh your is that where you

317
00:23:56,500 --> 00:24:01,839
put your servers in the bathroom you got to get

318
00:24:01,839 --> 00:24:04,779
real work done in that in that in the john

319
00:24:04,779 --> 00:24:10,519
that's great um so the the pod ping gossip swarm

320
00:24:10,519 --> 00:24:14,099
has been quietly running in the background for like a

321
00:24:14,099 --> 00:24:18,940
couple of months now with no hiccup. This thing is

322
00:24:18,480 --> 00:24:26,299
very stable now. And I'm desperate to get back to

323
00:24:26,299 --> 00:24:28,531
working on it if I could get this other crap

324
00:24:28,531 --> 00:24:33,051
out of the way there's always something popping up which

325
00:24:34,271 --> 00:24:39,211
again today so we have more to talk about the

326
00:24:39,211 --> 00:24:42,271
other day so there's the healthy hive RPC nodes is

327
00:24:42,271 --> 00:24:45,571
anybody doing hive work on the hive blockchain if they

328
00:24:45,571 --> 00:24:51,652
want to get get a list of the Current API

329
00:24:51,652 --> 00:24:58,771
servers hourly that pass a bunch of basic health checks

330
00:24:58,771 --> 00:25:01,112
to make sure that the API server is actually up

331
00:25:01,112 --> 00:25:09,231
and serving good information. And then the last one is

332
00:25:09,231 --> 00:25:17,112
the live tracking fire hose. and that is every episode

333
00:25:17,112 --> 00:25:22,932
that we ingest into the index and it's updated every

334
00:25:22,932 --> 00:25:29,152
three minutes, I think. Wow. It's updated every three minutes

335
00:25:28,731 --> 00:25:31,711
and the way that works is it's a linked list

336
00:25:31,711 --> 00:25:36,752
of object storage files. So it starts, the head of

337
00:25:36,752 --> 00:25:42,432
this linked list is tracking.podcastindex.org slash current And the way

338
00:25:42,432 --> 00:25:45,392
that this thing is designed to be used is that

339
00:25:45,392 --> 00:25:52,932
you start there with slash current, and then it gives

340
00:25:52,932 --> 00:25:56,711
you a previous tracking URL property, which goes to the

341
00:25:56,711 --> 00:26:00,852
linked to the previous one in the chain. So what

342
00:26:00,852 --> 00:26:04,392
you would do is you would just consistently walk backwards

343
00:26:04,392 --> 00:26:08,132
through the chain using that previous tracking link of tracking

344
00:26:08,132 --> 00:26:12,692
URL in each subsequent call until you get back to

345
00:26:12,692 --> 00:26:17,392
the place that you last were. So in that way,

346
00:26:17,392 --> 00:26:20,811
you can you're using object storage almost like a blockchain.

347
00:26:21,332 --> 00:26:25,271
There's obviously no signing or authentication, but you're using it

348
00:26:25,271 --> 00:26:28,551
sort of like a linked list in order to walk

349
00:26:28,551 --> 00:26:31,011
back without having to hit the API. So you can

350
00:26:31,011 --> 00:26:33,511
get massive amounts of data very quickly that way. Now,

351
00:26:33,511 --> 00:26:37,571
I still haven't really focused on that. I've had enough

352
00:26:37,571 --> 00:26:40,951
issues with my machines getting crunched with all kinds of

353
00:26:40,951 --> 00:26:47,112
stuff. But you can tag spam, slop, and legit in

354
00:26:47,112 --> 00:26:53,392
the PP monitor. Yeah. In the PP monitor? No, not

355
00:26:53,392 --> 00:27:00,892
in that. In the Gossiper? No, no. The Gossiper currently

356
00:27:00,892 --> 00:27:05,231
is purely just a... I thought we had that. I

357
00:27:05,231 --> 00:27:06,652
thought there was a way to... You're thinking of the

358
00:27:06,652 --> 00:27:10,011
PI monitor. Oh, that's right, the PI monitor. That's the

359
00:27:10,011 --> 00:27:13,011
command line tool that you can use, yes. Right, okay.

360
00:27:12,992 --> 00:27:20,152
But there's also more stuff on the Podcast Index Management

361
00:27:20,152 --> 00:27:22,112
Dashboard. I don't know if you've been in there lately.

362
00:27:22,592 --> 00:27:25,892
Yes, I have seen it. Have you seen the –

363
00:27:25,612 --> 00:27:28,271
We can do new things. Yeah, we can do new

364
00:27:28,271 --> 00:27:32,152
things. Of course I've seen that. Mark things as spam.

365
00:27:31,791 --> 00:27:36,451
Mark things as legit. Mm-hmm. Yeah. There's now an –

366
00:27:36,451 --> 00:27:42,211
as of yesterday, there is an admin panel for user

367
00:27:42,211 --> 00:27:49,992
account management. So if people need their passwords reset, developers,

368
00:27:49,992 --> 00:27:53,291
it's not listed in the menu. Oh, okay. I'd have

369
00:27:53,291 --> 00:27:57,592
to – hold on a second. Let's see. Curate? Oh,

370
00:27:57,592 --> 00:28:01,291
Curate is nice. Yeah, Curate is new. Yeah, that's cool.

371
00:28:02,252 --> 00:28:04,791
So what I'm looking at here is I'm an admin,

372
00:28:04,791 --> 00:28:08,692
and I see right off the bat, I see something

373
00:28:08,692 --> 00:28:13,211
odd. No episodes yet. so I might even want to

374
00:28:12,752 --> 00:28:19,511
that was added 38 seconds ago anchor no episodes yet

375
00:28:19,112 --> 00:28:23,592
but then I have let me see something weird happened

376
00:28:23,152 --> 00:28:26,051
here I can hit play I was 9 years old

377
00:28:26,051 --> 00:28:28,632
in the dark that's so cool and then I can

378
00:28:28,632 --> 00:28:34,432
hit legit legit boom done okay this is so nice

379
00:28:34,132 --> 00:28:40,912
what that did is this all goes back to the

380
00:28:40,912 --> 00:28:43,531
spam. So much of this goes back to spam classification

381
00:28:43,112 --> 00:28:49,571
work. What this really shows is that AI is not

382
00:28:49,132 --> 00:28:53,412
going to cost jobs. It is going to increase the

383
00:28:53,412 --> 00:28:56,451
workload for humans by a factor that we haven't even

384
00:28:56,451 --> 00:29:01,112
figured out yet. I truly believe this. I have to

385
00:29:01,112 --> 00:29:03,451
agree with you. And if it's not just because of

386
00:29:03,451 --> 00:29:06,571
more stuff that you're doing, it's because you can do

387
00:29:06,571 --> 00:29:09,852
more stuff. And when you can do more stuff, you

388
00:29:09,852 --> 00:29:15,571
do more stuff and anytime there's more stuff you have

389
00:29:15,571 --> 00:29:18,751
to have more human capacity to just handle it all

390
00:29:17,992 --> 00:29:22,392
yes, to do it and I look at myself there's

391
00:29:22,392 --> 00:29:26,211
so much stuff I'm doing and I could sit on

392
00:29:26,211 --> 00:29:28,112
my butt more but instead I'm like well let me

393
00:29:28,112 --> 00:29:32,352
do this let me use my robot oh yeah, I

394
00:29:32,352 --> 00:29:35,571
can figure this out yeah, this is good another thing

395
00:29:35,571 --> 00:29:39,172
I can do I'm just in teaching robot mode all

396
00:29:39,172 --> 00:29:42,731
day long for, you know, make my life easier. But

397
00:29:42,731 --> 00:29:46,672
what I'm doing is I'm taking that easier time and

398
00:29:46,672 --> 00:29:51,251
creating more things, which is just more work. It's an

399
00:29:51,251 --> 00:29:56,471
odd loop. And suddenly you need an intern. Yes. Yes,

400
00:29:56,471 --> 00:29:59,892
I need an intern. So let's talk about spam for

401
00:29:59,892 --> 00:30:04,192
a second. No, actually, let's talk about being on the

402
00:30:04,192 --> 00:30:09,672
fringe. Okay. because we we were laughing at uh rob

403
00:30:09,672 --> 00:30:13,992
greenlee talking about uh the podcasting 2.0 group being on

404
00:30:13,992 --> 00:30:17,551
the fringe oh yeah they're working on the fringe and

405
00:30:17,551 --> 00:30:24,071
and so rob posted this extremely long i guess apology

406
00:30:24,071 --> 00:30:28,011
uh let me see what it's titled here why fringe

407
00:30:28,011 --> 00:30:30,432
was the wrong word and what i actually meant about

408
00:30:30,432 --> 00:30:33,751
podcasting 2.0 oh where did he post this i didn't

409
00:30:33,751 --> 00:30:35,852
see this oh goodness oh you've got to see this

410
00:30:35,852 --> 00:30:38,892
this is hilarious okay i'll put it in the boardroom

411
00:30:38,892 --> 00:30:42,112
no rob's a funny guy well he's funny for a

412
00:30:42,112 --> 00:30:45,192
couple of reasons let me put this in here there

413
00:30:45,192 --> 00:30:49,852
we go so what is funny this is on rob

414
00:30:49,852 --> 00:30:55,932
greenley uh rob greenley.com uh and it's and right off

415
00:30:55,932 --> 00:30:58,751
the bat it's like this article provides context about my

416
00:30:58,751 --> 00:31:01,692
comments on the new media show episode 660 with libson

417
00:31:01,692 --> 00:31:06,112
ceo brendan monahan where We discussed podcasting 2.0 RSS tag

418
00:31:06,112 --> 00:31:09,392
adoption and the gap between innovation and mainstream platform implementation.

419
00:31:10,152 --> 00:31:14,971
So right away, I'm like, AI. So he had an

420
00:31:14,971 --> 00:31:21,571
AI write this whole long thing. And it even did

421
00:31:21,571 --> 00:31:28,971
a summary at the end. And I'm like, Rob, first

422
00:31:28,971 --> 00:31:32,731
of all, you were forgiven. You were forgiven right away.

423
00:31:33,172 --> 00:31:37,551
You know, this is – no one's mad. We're just

424
00:31:37,551 --> 00:31:40,652
slamming on you. We're just dunking on you. Oh, yeah.

425
00:31:40,231 --> 00:31:43,332
It's just ribbing. What I was trying to say. What

426
00:31:43,332 --> 00:31:45,172
I meant to say is that market fit and timing

427
00:31:45,172 --> 00:31:47,451
play a major role in what gets adopted at scale.

428
00:31:47,352 --> 00:31:50,992
Larger podcasting platforms tend to move more deliberately. Their decisions

429
00:31:50,992 --> 00:31:54,892
are shaped by user experience, engineering resources, monetization models, products,

430
00:31:54,892 --> 00:32:00,011
abilities, support, complexity, and business. Rob, it's okay. Just say,

431
00:32:00,011 --> 00:32:02,771
yeah, I probably shouldn't have said that. That's it. I

432
00:32:02,771 --> 00:32:06,551
say lots of stupid stuff all the time. It's okay.

433
00:32:06,172 --> 00:32:09,152
You didn't need to go through this whole thing. But

434
00:32:09,152 --> 00:32:12,531
it was a little insulting that you used the LLM

435
00:32:12,531 --> 00:32:15,652
to do that. Like, you don't need to. Like, come

436
00:32:15,652 --> 00:32:20,271
on. Rob's website. Rob's blog is pretty slick looking. Well,

437
00:32:20,271 --> 00:32:24,271
it's all written by AI. I mean, the graphics and

438
00:32:24,271 --> 00:32:27,372
stuff. It's got him smiling on the sidebar over there.

439
00:32:27,051 --> 00:32:34,172
It looks good. I get tired of documentation and essays

440
00:32:34,172 --> 00:32:39,652
and stuff that is clearly AI. It's like, no, don't

441
00:32:39,652 --> 00:32:44,051
do that to me. You know? And I can't even,

442
00:32:44,051 --> 00:32:48,672
my eyes start to glaze over. It's like, no, no.

443
00:32:48,791 --> 00:32:57,172
I cannot handle another like another like 5000 word AI

444
00:32:57,172 --> 00:33:01,991
generated document dump getting sent to me for me to

445
00:33:01,991 --> 00:33:08,811
read and somehow digest and process mentally. Like it is

446
00:33:08,811 --> 00:33:14,432
the the work that term works lop. Yeah. Oh, it's

447
00:33:14,432 --> 00:33:20,192
so real. like you're working on a project or something

448
00:33:20,192 --> 00:33:25,771
and somebody just just wants to like throw something into

449
00:33:25,771 --> 00:33:29,471
a chat bot dump out barf out 2500 words and

450
00:33:29,471 --> 00:33:34,311
say hey here check this out like man what i

451
00:33:34,311 --> 00:33:36,112
mean what am i supposed to be checking out i

452
00:33:36,112 --> 00:33:40,692
mean like what what do you Well, I don't. It's

453
00:33:40,692 --> 00:33:44,491
like I don't even want documents anymore. I just want

454
00:33:44,491 --> 00:33:47,092
you to Zoom call me. Just call me and we'll

455
00:33:47,092 --> 00:33:49,811
hash this out in 30 seconds. There's that. But what

456
00:33:49,811 --> 00:33:51,892
I tend to do is I see one of these

457
00:33:51,892 --> 00:33:55,711
and then I go to my robot. Hey, robot, summarize

458
00:33:55,711 --> 00:34:00,271
this for me. Exactly. I'm doing it right now. I

459
00:34:00,271 --> 00:34:03,912
said, robot, summarize this page into one paragraph. Here it

460
00:34:03,912 --> 00:34:06,991
is. Rob Greenlee walks back his use of the word

461
00:34:06,991 --> 00:34:10,331
fringe to describe podcasting 2.0, not the substance of his

462
00:34:10,331 --> 00:34:14,271
argument, just the word, explaining that he regretted the terminology

463
00:34:14,271 --> 00:34:17,952
but still believed there's a meaningful gap between technological innovation

464
00:34:17,952 --> 00:34:23,132
and platform adoption in podcasting. But that is the –

465
00:34:23,132 --> 00:34:27,132
all the stuff you just read could be settled in

466
00:34:27,132 --> 00:34:30,851
one sentence by saying, hey, man, we're good. Yeah, exactly.

467
00:34:32,311 --> 00:34:34,452
Although I don't want to do Zoom calls But yeah,

468
00:34:34,452 --> 00:34:42,052
exactly It is That's it It is exhausting It is

469
00:34:42,052 --> 00:34:46,452
kind of exhausting It's like that old adage When every

470
00:34:47,351 --> 00:34:50,311
When all you have is a hammer Everything looks like

471
00:34:50,311 --> 00:34:52,731
a nail When all you have is a machine that

472
00:34:52,731 --> 00:34:55,572
does language Guess what you're going to get out of

473
00:34:55,572 --> 00:34:58,751
it Lots and lots and lots of language And words

474
00:34:58,751 --> 00:35:05,452
and phrases and everything. It's just language is the nails.

475
00:35:06,211 --> 00:35:12,871
I ran across an interview by a guy named, what

476
00:35:12,871 --> 00:35:18,251
was this, Dr. Michael Eggnor. And he is a neurosurgeon.

477
00:35:19,271 --> 00:35:23,751
And it was really, when I was listening to this,

478
00:35:23,751 --> 00:35:26,992
it kind of blew my mind Because what was the

479
00:35:26,992 --> 00:35:31,351
big innovation in LLMs in the past year and a

480
00:35:31,351 --> 00:35:33,911
half to two years? I would say that is the

481
00:35:33,911 --> 00:35:41,391
reasoning model, which, by the way, I think is rapidly

482
00:35:41,391 --> 00:35:48,472
going out of favor. Explain. Well, reasoning models are incredibly

483
00:35:48,472 --> 00:35:55,132
slow. and instead of taking a dense model and putting

484
00:35:55,132 --> 00:36:01,492
layers or basically rounds of reasoning into it which slow

485
00:36:01,492 --> 00:36:05,231
everything down and don't actually give you a huge return

486
00:36:04,911 --> 00:36:11,692
what you're seeing is what's replacing it is more surgical

487
00:36:11,692 --> 00:36:18,291
use of context being deployed to sub-agents like the big

488
00:36:18,291 --> 00:36:21,192
in the big innovation really not innovation but the big

489
00:36:21,192 --> 00:36:24,831
distinction here do you know the book uh ender's game

490
00:36:24,831 --> 00:36:27,751
yeah i've heard of it i think we've discussed it

491
00:36:27,751 --> 00:36:31,251
actually yeah well it's like it's a sci-fi novel and

492
00:36:31,251 --> 00:36:34,351
um there's this one part in the book ender's game

493
00:36:33,911 --> 00:36:39,932
where ender starts going he goes into the he's a

494
00:36:39,932 --> 00:36:44,472
young kid and he's going to battle school and he

495
00:36:44,472 --> 00:36:47,192
goes into he he becomes he goes to the top

496
00:36:47,192 --> 00:36:50,612
of his class in this thing called the battle room

497
00:36:50,192 --> 00:36:54,811
and he's just better than everybody else at it and

498
00:36:54,811 --> 00:36:56,992
so it's basically just like it's on a space station

499
00:36:56,992 --> 00:37:01,692
and it's this zero g environment where there's all these

500
00:37:01,692 --> 00:37:04,052
obstacles and you're supposed to hide behind obstacles and use

501
00:37:04,052 --> 00:37:09,271
teamwork to defeat the other team and what he his

502
00:37:09,271 --> 00:37:14,072
big recognition was that why he was better, one reason

503
00:37:14,072 --> 00:37:15,692
he was better at it than everybody else is he

504
00:37:16,932 --> 00:37:21,271
just, he said as soon as he enters the battle

505
00:37:21,271 --> 00:37:26,012
room zero-G environment, he just forgets that there's anything like

506
00:37:26,012 --> 00:37:29,132
up and down. Basically, up and down don't make any,

507
00:37:29,132 --> 00:37:35,132
they don't exist anymore. all directionality everything you knew about

508
00:37:35,132 --> 00:37:40,572
uh physics has now been flown out the has flown

509
00:37:40,192 --> 00:37:43,632
out the window and so you just you assume a

510
00:37:43,632 --> 00:37:50,831
new mental posture towards towards your environment and that's sort

511
00:37:50,831 --> 00:37:55,112
of the way things things are with context like every

512
00:37:55,112 --> 00:37:59,092
time you send Every time you communicate with a large

513
00:37:59,092 --> 00:38:05,612
language model, you're starting over from scratch every single time.

514
00:38:06,432 --> 00:38:10,432
The coding agents hide this from you by using context.

515
00:38:10,692 --> 00:38:16,231
Right. And so the proper use of context is the

516
00:38:16,231 --> 00:38:19,311
thing that makes an agent or a harness either good

517
00:38:19,311 --> 00:38:22,751
or bad, how it handles that. There's no such thing

518
00:38:22,751 --> 00:38:27,592
as memory. I mean, in the general sense. These are

519
00:38:27,592 --> 00:38:30,411
all meta things that we've layered on top of. Well,

520
00:38:30,411 --> 00:38:32,851
enter the, I'm sorry, you told me never to delete

521
00:38:32,851 --> 00:38:35,532
your Git repo, but I did. I'm sorry, it won't

522
00:38:35,532 --> 00:38:41,771
happen again. Exactly. You can overcome a whole lot of

523
00:38:41,771 --> 00:38:46,351
things that the model – it's like the model is

524
00:38:46,351 --> 00:38:50,311
like 20% of the issue. The agent harness is the

525
00:38:50,311 --> 00:38:53,932
other 80%. 80 percent, and most of that boils down

526
00:38:53,932 --> 00:38:58,351
to intelligent use of the context window. And if you

527
00:38:58,351 --> 00:39:02,152
just assume that every time you go into a conversation

528
00:39:02,152 --> 00:39:04,112
with – I mean not a conversation, but every time

529
00:39:04,112 --> 00:39:09,351
you go send a prompt, it's brand new. The model

530
00:39:09,351 --> 00:39:12,532
doesn't know you. It doesn't know anything. It's never spoken

531
00:39:12,532 --> 00:39:15,072
to you before. This is the frustrating – let me

532
00:39:15,072 --> 00:39:17,992
play these clips. This will take us to another place.

533
00:39:17,731 --> 00:39:22,731
So the way I have always understood so-called artificial intelligence

534
00:39:22,731 --> 00:39:26,612
is we're trying to replicate the neural pathways of the

535
00:39:26,612 --> 00:39:30,552
brain. And we're trying to replicate what the brain does.

536
00:39:31,152 --> 00:39:33,911
Is that a fair statement? That is the way it's

537
00:39:33,911 --> 00:39:37,092
sold. That is completely bull. That's complete bull crap. Right.

538
00:39:37,711 --> 00:39:39,452
So that's the way it's sold. And when it comes

539
00:39:39,452 --> 00:39:44,851
to reasoning, however that's done, whether it's done through, whether

540
00:39:44,851 --> 00:39:49,432
it's, it's all a parlor trick, obviously, whether that's, you

541
00:39:49,432 --> 00:39:53,311
know, we're doing steps and we're making decisions, we're reasoning

542
00:39:53,311 --> 00:39:59,572
through a problem, or we're deciding to do something, which

543
00:39:59,572 --> 00:40:03,092
we as humans do. it was always my impression that

544
00:40:03,092 --> 00:40:06,891
that is what artificial intelligence even the term intelligence is

545
00:40:06,891 --> 00:40:10,612
trying to uh convince me that this is some form

546
00:40:10,612 --> 00:40:14,972
of electronic brain that is doing things so now enter

547
00:40:14,972 --> 00:40:20,932
uh dr eggnor who talks about reasoning and he also

548
00:40:20,932 --> 00:40:23,791
talks about free will in this case and the brain

549
00:40:23,791 --> 00:40:29,132
there are several ways of figuring out whether an organ

550
00:40:29,132 --> 00:40:33,492
does does something. One way is to stimulate the organ

551
00:40:33,492 --> 00:40:35,492
and get it to do it so you can prove

552
00:40:35,492 --> 00:40:38,132
it does it. The other way is to inhibit the

553
00:40:38,132 --> 00:40:40,911
organ from doing it and prove that when the organ's

554
00:40:40,911 --> 00:40:43,592
inhibited, it doesn't happen. Well, that can happen with the

555
00:40:43,592 --> 00:40:48,311
brain. While it is true that if you suppress the

556
00:40:47,871 --> 00:40:50,972
brain, we can't use reason or free will very well.

557
00:40:50,972 --> 00:40:52,152
If you get hit on the head with a baseball

558
00:40:52,152 --> 00:40:54,371
bat, your reasoning and your free will is going to

559
00:40:54,371 --> 00:40:57,472
be a little messed up for a while. There's a

560
00:40:57,472 --> 00:41:00,552
way of stimulating the brain to test whether reason and

561
00:41:00,552 --> 00:41:06,652
free will come from it using electrical stimulation. And there

562
00:41:06,652 --> 00:41:10,171
are different kinds of electrical stimulation. One kind of electrical

563
00:41:10,171 --> 00:41:15,632
stimulation is seizures. Seizures are just random sparks that happen

564
00:41:15,632 --> 00:41:18,911
inside the brain, out of the blue, that can have

565
00:41:18,911 --> 00:41:21,652
all kinds of effects. And one of the things that

566
00:41:21,652 --> 00:41:25,771
I've looked at rather carefully is the phenomenology of seizures.

567
00:41:25,671 --> 00:41:28,371
It is what happens when you have a seizure. And

568
00:41:28,371 --> 00:41:31,552
I've looked over the past couple hundred years, reviewed the

569
00:41:31,552 --> 00:41:35,632
medical literature of that. And what we find is rather

570
00:41:35,632 --> 00:41:38,132
remarkable. First of all, not all seizures make you go

571
00:41:38,132 --> 00:41:41,972
unconscious. There are many seizures that leave you conscious, and

572
00:41:41,972 --> 00:41:44,871
your body does things that you can't control. And they're

573
00:41:44,871 --> 00:41:49,711
called complex partial seizures. and um there have been probably

574
00:41:49,711 --> 00:41:52,952
a quarter of a billion seizures in human beings over

575
00:41:52,952 --> 00:41:54,932
the past 200 years if you just look at the

576
00:41:54,932 --> 00:41:59,192
instance of seizures in the world population and there's never

577
00:41:59,192 --> 00:42:03,391
been a recorded case of a seizure causing someone to

578
00:42:03,391 --> 00:42:07,112
express reason or free will you can elicit from the

579
00:42:07,112 --> 00:42:13,592
brain by stimulating it movement perception memory and emotion but

580
00:42:13,592 --> 00:42:16,572
you can never elicit reason or free will no matter

581
00:42:16,572 --> 00:42:18,771
where you stimulate the brain you can't make a person

582
00:42:18,771 --> 00:42:22,751
say one plus one is two or say um um

583
00:42:22,751 --> 00:42:27,452
uh if a then b a therefore b modus ponens

584
00:42:27,452 --> 00:42:31,311
or say um i believe that it's nice to be

585
00:42:31,311 --> 00:42:35,692
kind which is a moral viewpoint so that goes right

586
00:42:35,692 --> 00:42:38,811
along with what is found in seizures that you can

587
00:42:38,811 --> 00:42:42,552
stimulate the brain any which way and reason and free

588
00:42:42,552 --> 00:42:45,911
will never come out of the brain. So it's perfectly

589
00:42:45,911 --> 00:42:49,452
reasonable to infer that reason and free will don't naturally

590
00:42:49,452 --> 00:42:51,612
come from the brain, because you can't make the brain

591
00:42:51,612 --> 00:42:55,811
do it. I loved hearing this. Reason and free will

592
00:42:55,811 --> 00:42:59,911
don't come from the brain. This is very similar to

593
00:42:59,911 --> 00:43:09,311
our, you know, this is just another version of what

594
00:43:09,311 --> 00:43:10,911
we talked about before with the brain and the law

595
00:43:10,911 --> 00:43:14,411
of identity. You know, if you can find something about

596
00:43:14,411 --> 00:43:18,271
the brain that is not true of the, quote, mind,

597
00:43:18,271 --> 00:43:22,692
then they are not the same thing. Yes. And this

598
00:43:22,692 --> 00:43:24,911
guy actually takes it one step further, and he tells

599
00:43:24,911 --> 00:43:27,851
us where this does come from. So going back to

600
00:43:27,851 --> 00:43:30,671
something you said earlier, that reason and free will are

601
00:43:30,671 --> 00:43:35,851
immaterial aspects or parts of the person, powers of the

602
00:43:35,851 --> 00:43:42,211
person, immaterial. What does that even mean? immaterial means that

603
00:43:42,211 --> 00:43:47,411
um the source of these powers is not a physical

604
00:43:47,411 --> 00:43:53,171
thing it's not a physical object um and um immaterial

605
00:43:53,171 --> 00:43:58,211
is actually a fairly simple concept for example um numbers

606
00:43:58,211 --> 00:44:02,692
are immaterial it is that the the number 12 is

607
00:44:02,692 --> 00:44:04,612
an immaterial thing that is you can't say well the

608
00:44:04,632 --> 00:44:08,311
Number 12 is located today in Cincinnati, and it weighs

609
00:44:08,311 --> 00:44:11,871
five pounds. I mean, it doesn't make any sense. So

610
00:44:11,871 --> 00:44:14,612
there are many, many, many things that we think about,

611
00:44:14,612 --> 00:44:18,092
many things that are quite real. They're called universals in

612
00:44:18,092 --> 00:44:24,411
philosophy that are very real but are not matter. And

613
00:44:24,411 --> 00:44:29,271
reason and free will are abilities we have whose source

614
00:44:29,271 --> 00:44:34,932
is our soul but not our brain. thank you jesus

615
00:44:34,932 --> 00:44:41,391
it's the soul that's what it is and never ever

616
00:44:41,391 --> 00:44:45,911
ever ever ever is an llm going to be able

617
00:44:45,911 --> 00:44:49,311
to replicate this it can emulate and it maybe can

618
00:44:49,311 --> 00:44:51,692
kind of come close and i've been running into this

619
00:44:51,692 --> 00:44:55,711
as i've been trying to teach my robot what makes

620
00:44:55,711 --> 00:44:58,112
a clip that i want now it can do clips

621
00:44:58,811 --> 00:45:01,871
I can say, you know, cut this cut this interview

622
00:45:01,871 --> 00:45:04,512
up and it can build the arc with the intro,

623
00:45:04,512 --> 00:45:07,351
you know, getting to the point, the punchline at the

624
00:45:07,351 --> 00:45:09,791
end. It can do that. It can it can edit

625
00:45:09,791 --> 00:45:13,171
this clip. It understands kind of how long I want

626
00:45:13,171 --> 00:45:16,512
clips to be, but it doesn't really understand what I

627
00:45:16,512 --> 00:45:22,871
want. And I've been trying so hard to explain, like,

628
00:45:22,871 --> 00:45:26,311
you know, I want is what I find funny is

629
00:45:26,311 --> 00:45:29,592
not the fact that Trump went to. i find it

630
00:45:29,592 --> 00:45:33,751
funny that the chinese actually have sanctioned marco rubio he's

631
00:45:33,751 --> 00:45:37,751
not allowed to enter the country but they they got

632
00:45:37,751 --> 00:45:41,072
around it by changing his name to lubio i mean

633
00:45:41,072 --> 00:45:45,032
the comedy writes itself right marco rubio come on but

634
00:45:45,032 --> 00:45:49,231
you know it's like yeah but it can't understand that

635
00:45:48,932 --> 00:45:52,331
that's what i want you know so it just it

636
00:45:52,331 --> 00:45:54,492
can only understand well you want clips you want them

637
00:45:54,492 --> 00:45:57,891
uh preferably under two minutes you want this it can't

638
00:45:57,891 --> 00:46:01,651
it really can't figure that out and i and hearing

639
00:46:01,651 --> 00:46:04,692
this helped me because i'm like oh it'll never be

640
00:46:04,692 --> 00:46:06,811
able to figure it out it might come close it

641
00:46:06,811 --> 00:46:10,112
might get lucky but it'll never really understand what i

642
00:46:10,112 --> 00:46:12,072
want for no agenda when it comes to a clip

643
00:46:12,072 --> 00:46:15,512
it just won't understand it well this is the difference

644
00:46:15,512 --> 00:46:20,411
between uh the way that AI can be has been

645
00:46:20,411 --> 00:46:24,072
sold in the way that it actually functions. Yeah. And

646
00:46:24,072 --> 00:46:29,271
when and this is how you create. Every single time,

647
00:46:29,271 --> 00:46:34,092
this is how the AI winters have been created is

648
00:46:34,092 --> 00:46:37,992
by selling it as something that it is not. Yes,

649
00:46:37,992 --> 00:46:44,132
and thank you, Hey Citizen, because I also don't believe

650
00:46:44,132 --> 00:46:48,151
it can actually figure out what spam or slop is.

651
00:46:48,612 --> 00:46:52,671
It cannot. It can't. It never can. And the same

652
00:46:52,671 --> 00:46:57,072
thing with music. It can create dynamite music, but it

653
00:46:57,072 --> 00:46:59,291
can't actually listen to that music and say, this is

654
00:46:59,291 --> 00:47:09,452
good. It can't. Let me tell you why. let me

655
00:47:09,452 --> 00:47:11,291
tell you why I can't figure out what spam is.

656
00:47:11,132 --> 00:47:28,851
Because there is a intangible quality to an experience of

657
00:47:28,851 --> 00:47:40,811
spam that you cannot quantify in any sort of language

658
00:47:40,811 --> 00:47:47,632
or honestly anything. You cannot, it's unquantifiable. And a good

659
00:47:47,632 --> 00:47:52,851
example of this would be there's a common spam tactic

660
00:47:52,851 --> 00:47:57,952
in the tax world where you'll communicate, where you get

661
00:47:57,952 --> 00:48:01,671
an email and it's like, hey, my accountant just retired

662
00:48:02,711 --> 00:48:07,012
and I'm looking for somebody to do my taxes. Can

663
00:48:07,012 --> 00:48:11,692
you tell me what your fees are? And I just

664
00:48:11,692 --> 00:48:18,271
have a 1040 and a couple of 1099s and a

665
00:48:18,271 --> 00:48:20,932
W-2 and I can send you my last year's material.

666
00:48:20,851 --> 00:48:25,791
Now give me root password. It leaves that last part

667
00:48:25,791 --> 00:48:28,811
out. And it's like, can I send you my last

668
00:48:28,811 --> 00:48:30,692
year's tax return so you can give me an accurate

669
00:48:30,692 --> 00:48:32,592
quote on what it would be and what it would

670
00:48:32,592 --> 00:48:39,291
cost? There is literally nothing about this email that that

671
00:48:39,291 --> 00:48:44,911
will trigger a phishing scam detection ever because you'll get

672
00:48:44,911 --> 00:48:47,432
50 of these and they're real. You'll get 50 of

673
00:48:47,432 --> 00:48:50,612
them and they're phishing. But I can tell you, I

674
00:48:50,583 --> 00:48:54,884
will spot the fishing one versus the real one 100%

675
00:48:54,884 --> 00:48:57,143
of the time, and I cannot tell you why. Right.

676
00:48:57,563 --> 00:49:00,504
I can't. It's because you have a grandmother. You had

677
00:49:00,504 --> 00:49:05,583
a childhood. Yeah, exactly. You built associations in the world

678
00:49:05,583 --> 00:49:08,364
between – and, of course, I always have to keep

679
00:49:08,364 --> 00:49:13,123
reminding myself that the LLM can't actually hear my show.

680
00:49:13,684 --> 00:49:17,324
You know, I created an executive producer agent. I said,

681
00:49:17,324 --> 00:49:20,943
okay, now this is your job. And it took me

682
00:49:20,943 --> 00:49:24,963
two weeks before I realized it could, it just had

683
00:49:24,963 --> 00:49:27,824
a transcript. I didn't have a speaker transcript. It thought

684
00:49:27,824 --> 00:49:32,184
everything was me, including 60% of the talk on No

685
00:49:32,184 --> 00:49:35,923
Agenda, which is clips. It hadn't even gotten that far.

686
00:49:37,184 --> 00:49:41,824
that it was trying to you know it saw this

687
00:49:41,824 --> 00:49:45,244
program as some kind of script that you set up

688
00:49:45,244 --> 00:49:47,784
with x amount of minutes of clips and i was

689
00:49:47,784 --> 00:49:51,943
like wow it has no it can't hear it's deaf

690
00:49:51,943 --> 00:49:56,923
it's deaf it doesn't understand it doesn't understand humor really

691
00:49:56,923 --> 00:50:00,884
it doesn't it doesn't understand emotions that way It doesn't

692
00:50:00,884 --> 00:50:05,364
understand when I say something that Dvorak finds boring. It

693
00:50:05,364 --> 00:50:08,004
doesn't understand that, and I don't think it ever can.

694
00:50:08,903 --> 00:50:12,484
Selas in the boardroom said, wouldn't a dedicated neural network

695
00:50:12,484 --> 00:50:16,004
designed for it to be able to pattern match beyond

696
00:50:16,004 --> 00:50:21,984
the capability of humans? Wouldn't a dedicated neural network designed

697
00:50:21,984 --> 00:50:24,563
for it be able to pattern match beyond the capability

698
00:50:24,563 --> 00:50:28,284
of humans because humans cannot look at and memorize 4.6

699
00:50:28,284 --> 00:50:33,903
million feeds? that what you were describing now, what you

700
00:50:33,903 --> 00:50:41,603
were describing is, I will say, go back in time

701
00:50:41,224 --> 00:50:47,784
and look at the attempts to use Bayesian analysis to

702
00:50:47,784 --> 00:50:54,043
fight spam email. Bayesian filters, this happened in the early

703
00:50:54,043 --> 00:50:59,724
2000s and when it first hit the scene, Was considered

704
00:50:59,724 --> 00:51:01,943
to be, oh, my gosh, this is going to. This

705
00:51:01,943 --> 00:51:03,523
is it. This is it. This is it. We got

706
00:51:03,523 --> 00:51:06,664
it. No more. Nailed it. Yeah, we nailed it. And

707
00:51:06,664 --> 00:51:09,523
you know how much of a dent it made? Zero.

708
00:51:10,963 --> 00:51:16,004
We have more spam than we've ever had ever there.

709
00:51:15,284 --> 00:51:23,984
You cannot. Humans are better at pattern matching. So neural

710
00:51:23,984 --> 00:51:28,744
networks and large language models. It's just a different format

711
00:51:28,744 --> 00:51:35,684
of neural network. It's just matrix multiplication. That's all it

712
00:51:35,684 --> 00:51:41,023
is. It's matrix math. And that's not how humans work.

713
00:51:41,784 --> 00:51:44,224
When you think about something, when you look at something

714
00:51:44,224 --> 00:51:47,184
and it registers in your mind, you did not do

715
00:51:47,184 --> 00:51:51,784
a math problem. That's not how you operate. That's not

716
00:51:51,784 --> 00:51:56,484
what just happened. Exactly. And so what we don't know

717
00:51:56,484 --> 00:52:01,164
how we work, it is undefinable. And the end, but

718
00:52:01,164 --> 00:52:06,844
one thing we do know is that one thing we

719
00:52:06,844 --> 00:52:10,443
do know is that the human ability to pattern match

720
00:52:10,443 --> 00:52:20,304
is unparalleled. paralleled, you have the ability to assimilate multiple

721
00:52:20,304 --> 00:52:26,244
streams, dozens and dozens of streams of information all in

722
00:52:26,244 --> 00:52:32,664
one shot and immediately get a, quote, sense about something

723
00:52:32,664 --> 00:52:37,764
that you cannot define. And that is, again, you did

724
00:52:37,764 --> 00:52:41,184
not do math. And it's wonderful to see someone like

725
00:52:41,184 --> 00:52:44,784
Tina who will be looking at her email and she

726
00:52:44,784 --> 00:52:48,463
says, this feels spammy to me. Can you check? And

727
00:52:48,463 --> 00:52:50,403
the minute I check and I say, okay, yeah, it

728
00:52:50,403 --> 00:52:52,884
is. And here's why it is. You know, there's a

729
00:52:52,884 --> 00:52:55,543
couple of simple things that you can look at right

730
00:52:55,543 --> 00:53:00,004
away. She then immediately has increased her knowledge and she

731
00:53:00,004 --> 00:53:03,244
finds those the next time. And I don't have to

732
00:53:03,244 --> 00:53:05,923
give her a written thesis. I just say, oh, look

733
00:53:05,923 --> 00:53:08,824
at this. You know, and sometimes like, well, that feels

734
00:53:08,824 --> 00:53:12,684
wrong. And she's like, yeah. And so she's feeling that

735
00:53:12,684 --> 00:53:16,284
it's right. You said something really important. You said AI

736
00:53:16,284 --> 00:53:22,004
winter, every single AI winter crops up or appears on

737
00:53:22,004 --> 00:53:26,923
the horizon because of the overpromise. Is that really the

738
00:53:26,923 --> 00:53:30,423
trigger, the overpromise, or it can't deliver on what was

739
00:53:30,423 --> 00:53:36,123
promised? That sets up the collapse because the expectations of

740
00:53:36,123 --> 00:53:43,103
normies can't understand the reality of the technology. Because what

741
00:53:43,103 --> 00:53:45,364
we do is we layer these things on top of

742
00:53:45,364 --> 00:53:49,903
one another. We've layered this idea of, quote, knowledge and

743
00:53:49,903 --> 00:53:55,304
understanding, and we anthropomorphize it into a thing. Because I

744
00:53:55,304 --> 00:54:00,864
hear it in the corporate world daily is, hey, couldn't

745
00:54:00,864 --> 00:54:04,884
we use AI to – Dot, dot, dot. Yeah, to,

746
00:54:04,884 --> 00:54:09,764
quote, figure out dot, dot, dot. Yeah. Like, and it's

747
00:54:09,764 --> 00:54:13,504
– you get so exhausted of having to go back

748
00:54:13,504 --> 00:54:17,643
and explain the tech from the ground up again. And

749
00:54:17,643 --> 00:54:21,463
you're like, look, maybe, but let me show you. But

750
00:54:21,463 --> 00:54:24,724
I have to walk you through the process of building

751
00:54:24,724 --> 00:54:29,543
a product or building a tool that would do something

752
00:54:29,543 --> 00:54:34,184
that approximates the outcome that you want. And be prepared

753
00:54:34,184 --> 00:54:37,844
for it to do something completely different tomorrow. The perfect

754
00:54:37,844 --> 00:54:42,784
example of this is, you know, Microsoft just released their

755
00:54:42,784 --> 00:54:49,284
their copilot function in Excel. So now there's an actual

756
00:54:49,284 --> 00:54:53,443
co-pilot function that you can call in an Excel spreadsheet.

757
00:54:55,204 --> 00:55:05,244
And it is hilarious because the co-pilot function in its

758
00:55:05,244 --> 00:55:11,784
description of how to use it, it says what not

759
00:55:11,784 --> 00:55:17,923
to use it for. Accounting. Exactly. One of the things

760
00:55:17,923 --> 00:55:20,423
it says not to use it for is, quote, anything

761
00:55:20,423 --> 00:55:27,184
that requires accuracy or reproducibility. And so people are posting

762
00:55:27,184 --> 00:55:29,864
– people are posting like – they'll do like –

763
00:55:29,864 --> 00:55:32,204
they'll go into a cell and they'll do equal copilot,

764
00:55:32,204 --> 00:55:35,804
and then they'll put in a prompt. and the prompt

765
00:55:35,804 --> 00:55:40,164
will be make a sum of all the cells above

766
00:55:40,164 --> 00:55:44,304
this cell and it'll be like three cells above there

767
00:55:44,304 --> 00:55:46,364
like a one a two and a three and it'll

768
00:55:46,364 --> 00:55:51,884
say fifteen why? I don't know it doesn't tell you

769
00:55:51,884 --> 00:55:53,864
I don't know it just gives you a number and

770
00:55:53,864 --> 00:56:01,923
it's completely wrong this is the problem people think because

771
00:56:01,923 --> 00:56:04,864
of the way it's been sold, they think that it's

772
00:56:04,864 --> 00:56:09,224
a general purpose knowledge tool. Yeah. But it's not, it

773
00:56:09,224 --> 00:56:13,063
has nothing to do with knowledge. It's only good at

774
00:56:13,063 --> 00:56:16,664
predicting the next word in a sentence. Yeah. Or the

775
00:56:16,664 --> 00:56:21,563
next code token in a source file. That's all it's

776
00:56:21,563 --> 00:56:23,943
good at. What I found is. And anything else you

777
00:56:23,943 --> 00:56:26,864
have to layer on top of it with actual, like,

778
00:56:26,864 --> 00:56:33,643
real, carefully constructed, Constructed, carefully you have to build a

779
00:56:33,643 --> 00:56:39,284
carefully constructed stack that contains this thing and keeps it

780
00:56:39,284 --> 00:56:41,884
with a set of guardrails so that it doesn't produce

781
00:56:41,884 --> 00:56:50,344
crazy nonsense. So you're right about context really being everything

782
00:56:50,344 --> 00:56:54,423
for it. And what I've resorted to now, because you

783
00:56:54,423 --> 00:56:56,704
gave me a couple of good hints. So I'm using

784
00:56:56,704 --> 00:57:01,724
Opus 4.7 with the one million token context. When it

785
00:57:01,724 --> 00:57:06,563
gets to about 25%, it now automatically tells me, hey,

786
00:57:06,563 --> 00:57:10,824
it's 25% context, time to do a handoff. So I

787
00:57:10,824 --> 00:57:15,504
don't let it get all muddled. And I say, okay,

788
00:57:15,504 --> 00:57:17,603
let's go. It writes its handoff, and it gives me

789
00:57:17,603 --> 00:57:21,344
a wake sentence like, hey, welcome back to the world.

790
00:57:21,423 --> 00:57:26,083
Here's what you got to read. And it reads a

791
00:57:26,083 --> 00:57:28,083
document, which – and I've looked at the documents. They're

792
00:57:28,083 --> 00:57:31,484
pretty good. It's where we were. And that seems to

793
00:57:31,484 --> 00:57:37,423
work pretty well within that 250,000 tokens. That's kind of

794
00:57:37,423 --> 00:57:41,284
the sweet spot. And I just keep doing it. It's

795
00:57:41,284 --> 00:57:46,123
inconvenient because like, well, I'm at 23% context. You want

796
00:57:46,123 --> 00:57:47,403
me to do this now or do you want me

797
00:57:47,403 --> 00:57:49,023
to put it in the handoff document? Yeah, put it

798
00:57:49,023 --> 00:57:51,984
in the handoff document. That works pretty well. If you

799
00:57:51,984 --> 00:57:55,103
keep it within that level of context, I can get

800
00:57:55,103 --> 00:58:01,123
a lot done. that you can you can do everything

801
00:58:01,123 --> 00:58:04,583
you need is coding wise you can do every single

802
00:58:04,583 --> 00:58:07,563
thing you need to do in a 256k context window

803
00:58:07,563 --> 00:58:10,103
there's no reason to go above that yeah there's not

804
00:58:10,103 --> 00:58:13,443
and that that one million context of of of like

805
00:58:13,443 --> 00:58:17,083
clawed code and chat gpt that that is that is

806
00:58:17,083 --> 00:58:24,423
in um that is inaccurate in the sense that what

807
00:58:24,423 --> 00:58:28,583
really happens is once you get over that roughly 25%,

808
00:58:28,164 --> 00:58:33,023
it will start to make more and more mistakes. If

809
00:58:33,023 --> 00:58:38,364
you start pushing into the 600,000 to 700,000 token context

810
00:58:38,364 --> 00:58:41,824
window, that's when you start really running into stuff like,

811
00:58:41,824 --> 00:58:47,184
oh, I accidentally deleted your repo, sorry. That was my

812
00:58:47,184 --> 00:58:49,903
favorite. You told me never. I mean, how many times

813
00:58:49,903 --> 00:58:54,103
I've told my robot how to make clips? And then

814
00:58:54,103 --> 00:58:58,664
I'm like, okay, I have all these agents and all

815
00:58:58,664 --> 00:59:02,304
these feed parsers and all these sources, and it picks

816
00:59:02,304 --> 00:59:04,623
up stories. I said, I'll go make some clips. It'll

817
00:59:04,623 --> 00:59:08,804
make like 100 clips of 17 seconds long. I'm like,

818
00:59:08,804 --> 00:59:11,403
how many times have we gone through this? Oh, yeah,

819
00:59:11,403 --> 00:59:13,244
you're right. I forgot to read my memory. Let me

820
00:59:13,244 --> 00:59:17,063
see how many memory documents I have. Let me see.

821
00:59:16,664 --> 00:59:23,984
How many memory documents? This will tell you something. Let

822
00:59:23,984 --> 00:59:31,284
me see. I bet just a couple hundred at least.

823
00:59:32,344 --> 00:59:34,543
And I guess whenever it reads one of these documents,

824
00:59:34,543 --> 00:59:38,284
that takes up context space, doesn't it? Yeah, for sure.

825
00:59:38,443 --> 00:59:40,984
Yeah. Depending on that. Now, this goes back to what

826
00:59:40,984 --> 00:59:45,324
I was saying earlier about the harness. the coding agent

827
00:59:45,324 --> 00:59:50,764
or the harness itself it is critical it does all

828
00:59:50,764 --> 00:59:53,304
of this, it handles this context for you if it's

829
00:59:53,304 --> 00:59:55,824
smart about it, if it's very smart about it it

830
00:59:55,824 --> 00:59:57,963
can handle those things in a way that does not

831
00:59:57,963 --> 01:00:00,463
consume all your context but it really comes down to

832
01:00:00,463 --> 01:00:11,324
that 252 yeah so there's a really good video, I'll

833
01:00:11,324 --> 01:00:13,923
find it later and post it publicly. There's a really

834
01:00:13,923 --> 01:00:17,023
good video where this guy goes through and there's a

835
01:00:17,023 --> 01:00:21,184
really simplistic test you can do for a coding model.

836
01:00:23,864 --> 01:00:32,284
This test just says it's really simple. It says go

837
01:00:32,284 --> 01:00:35,443
and read this large source file. So just give it

838
01:00:34,984 --> 01:00:40,164
a big JavaScript file. Let's say this JavaScript file is

839
01:00:40,164 --> 01:00:49,324
50,000 lines. Read this JavaScript file. Now tell me what

840
01:00:49,324 --> 01:00:56,963
lines 50 through 100 say. Okay. And you would be

841
01:00:56,963 --> 01:01:00,463
surprised to think that even the large models with a

842
01:01:00,463 --> 01:01:08,344
million token context will get that answer wrong about about

843
01:01:08,344 --> 01:01:11,143
10% of the time. Wow. They just won't be able

844
01:01:11,143 --> 01:01:12,923
to tell you what it just – it won't be

845
01:01:12,923 --> 01:01:16,664
able to tell you what it just read. It just

846
01:01:16,664 --> 01:01:20,963
makes a mistake. So that's why these things – these

847
01:01:20,963 --> 01:01:28,543
harnesses have developed the use of these elaborate sub-agent reviewers

848
01:01:29,603 --> 01:01:33,003
because they'll do a thing assuming that they're going to

849
01:01:33,003 --> 01:01:35,844
get it wrong some percentage of the time because it's

850
01:01:35,844 --> 01:01:38,384
just not possible for the underlying model to get it

851
01:01:38,384 --> 01:01:41,903
right. So they need the reviewer to go back and

852
01:01:41,903 --> 01:01:45,884
double-check them, double-check them, double-check them every single time. If

853
01:01:45,884 --> 01:01:50,304
you just one-shot prompt an LLM model with no sort

854
01:01:50,304 --> 01:01:55,364
of review agent or anything like that in it, you're

855
01:01:55,364 --> 01:02:01,864
not going to get useful output probably in the 20%

856
01:02:01,864 --> 01:02:05,103
range. All that said, I think the face of computing

857
01:02:05,103 --> 01:02:09,463
could change. I see no reason why I should be

858
01:02:09,463 --> 01:02:13,623
using apps. I don't use even applications. I use a

859
01:02:13,623 --> 01:02:18,784
browser, and I use Thunderbird for email. But everything else

860
01:02:18,784 --> 01:02:21,403
is like, nah, I want a spreadsheet reader that does

861
01:02:21,403 --> 01:02:24,984
this. Five minutes later, I've got exactly what I want.

862
01:02:26,523 --> 01:02:30,704
I can see a mobile device that really is appless.

863
01:02:32,023 --> 01:02:36,563
get me my favorite podcast here's the list get me

864
01:02:36,563 --> 01:02:41,164
the most recent episodes here's what I have time for

865
01:02:41,164 --> 01:02:43,664
right now show this to me I want to listen

866
01:02:43,664 --> 01:02:49,684
to this skip the ads or any version of that

867
01:02:50,403 --> 01:02:53,083
this is what I love so much about Homarchi it's

868
01:02:53,083 --> 01:02:55,744
like every app I use I built I just told

869
01:02:55,744 --> 01:02:59,184
the bot build this for me I don't need the

870
01:02:59,184 --> 01:03:03,623
ribbon on Excel. I don't need the snapping grid. I

871
01:03:03,623 --> 01:03:05,463
need to smooth scroll just to be able to see

872
01:03:05,463 --> 01:03:07,603
this. I need the total automatically done at the bottom.

873
01:03:07,603 --> 01:03:09,903
I don't want to have to, you know, select and

874
01:03:09,903 --> 01:03:11,784
then do equal sum. I don't want any of that.

875
01:03:11,784 --> 01:03:14,764
Just do it for me. And it does that. Then

876
01:03:14,764 --> 01:03:17,123
if I need something else for a different podcast, I'll

877
01:03:17,123 --> 01:03:22,143
create a new one. This I can see. And I

878
01:03:22,143 --> 01:03:26,103
think that the removal of slop and all this stuff,

879
01:03:26,103 --> 01:03:28,583
I think that all moves to the edge, man it

880
01:03:28,583 --> 01:03:32,184
all moves to your to your device uh maybe you

881
01:03:32,184 --> 01:03:34,284
have a device in the home that is doing some

882
01:03:33,824 --> 01:03:36,744
things for in the background that communicates with your mobile

883
01:03:36,744 --> 01:03:39,684
device and you know this just because it's easier to

884
01:03:39,684 --> 01:03:43,344
have that done somewhere else um and i think podcast

885
01:03:43,344 --> 01:03:49,043
apps will go that way too i really do this

886
01:03:49,043 --> 01:03:53,643
brings us back to the dgx spark yeah it's it's

887
01:03:53,643 --> 01:03:59,764
useless send it back it's no good um it's running

888
01:03:59,764 --> 01:04:04,764
it's running the uh the inference for the classifier now

889
01:04:04,764 --> 01:04:08,204
and it allowed us to move from a from a

890
01:04:08,204 --> 01:04:12,284
9 billion parameter model to a 35 billion parameter model

891
01:04:12,284 --> 01:04:18,143
the quin 3635b and uh with a with a 256k

892
01:04:18,143 --> 01:04:21,003
context window so we get we've got a full like

893
01:04:21,003 --> 01:04:28,364
frontier model level thing as a backing agent for this

894
01:04:28,364 --> 01:04:33,923
thing. And it's so much faster. So we can run

895
01:04:33,484 --> 01:04:38,844
that and the Whisper both to get the TTS analysis.

896
01:04:38,844 --> 01:04:42,324
Both on this one box. What are you using for

897
01:04:42,324 --> 01:04:47,284
Whisper? Which model? I think it's the Turbo. Yeah. Okay.

898
01:04:48,364 --> 01:04:53,664
And And so it's immediately had that effect and it's

899
01:04:53,664 --> 01:04:57,704
speeded everything up and made us have more accuracy. But

900
01:04:57,704 --> 01:05:07,503
the true aim here is to get a custom fine-tuned

901
01:05:07,503 --> 01:05:13,423
model trained on the podcast index database. And then –

902
01:05:13,423 --> 01:05:17,003
so that's phase two here. But in order to get

903
01:05:17,003 --> 01:05:20,664
to phase two, I have to finish up phase one,

904
01:05:20,664 --> 01:05:25,244
which is we need – like I explained before, we

905
01:05:25,244 --> 01:05:32,543
need a good corpus of about 25,000 spam feeds and

906
01:05:32,543 --> 01:05:35,764
25,000 legit feeds. I would love to have more than

907
01:05:35,764 --> 01:05:37,503
that. I would love to have like 50,000 of each.

908
01:05:38,284 --> 01:05:44,463
Just wait two weeks. Yeah, right. It's coming. The spam

909
01:05:44,463 --> 01:05:47,903
feeds Those are easy You can get those all day

910
01:05:47,903 --> 01:05:52,483
long Effortless The legit feeds are the ones that take

911
01:05:52,483 --> 01:06:02,204
a human In the loop To find And so What

912
01:06:02,204 --> 01:06:06,423
was that? What was that? I'm sorry, that was me

913
01:06:06,943 --> 01:06:14,684
That was my mistake The so the first so what

914
01:06:14,684 --> 01:06:18,184
I did that's what all these data sets and all

915
01:06:18,184 --> 01:06:22,724
these and the curator and the dashboard all this stuff

916
01:06:22,724 --> 01:06:28,184
is trying to set us up for getting a proper

917
01:06:27,804 --> 01:06:34,744
training set for the quote good podcasts I hate using

918
01:06:34,744 --> 01:06:36,264
that term I just can't think of a different one

919
01:06:35,943 --> 01:06:42,963
legit podcasts legit there it is very subjective but it's

920
01:06:42,963 --> 01:06:47,164
legit and so then the train this the djx spark

921
01:06:47,164 --> 01:06:50,583
the reason that was such an amazing gift is because

922
01:06:50,583 --> 01:06:55,324
the spark is created exactly for this use case training

923
01:06:55,324 --> 01:07:00,083
models on a small scale that's where its wheelhouse is

924
01:07:00,083 --> 01:07:04,903
so once we get these proper data sets. And let

925
01:07:04,903 --> 01:07:08,943
me correct James Cridland from PodNews, I think, last week.

926
01:07:08,923 --> 01:07:12,603
He was misunderstanding why I wanted these follower count uploads

927
01:07:12,403 --> 01:07:20,664
from the apps. It's mostly for this. We will definitely

928
01:07:20,224 --> 01:07:24,463
use this data in other ways, but it's mostly for

929
01:07:24,463 --> 01:07:29,364
this use case for the model training because that benefits

930
01:07:29,364 --> 01:07:33,224
everybody. I'm not really interested in using it for search

931
01:07:33,224 --> 01:07:37,043
the search that we have seems fine to me I

932
01:07:37,043 --> 01:07:42,623
mean we use some algorithms for it but it'll probably

933
01:07:42,623 --> 01:07:44,384
make its way in there in some shape, form or

934
01:07:44,384 --> 01:07:46,744
fashion but it's not just going to start giving you

935
01:07:46,824 --> 01:07:48,764
well if anyone's going to figure out how to make

936
01:07:48,764 --> 01:07:52,784
it work it's you I just don't want anybody to

937
01:07:52,784 --> 01:07:55,804
think that we're going to start only showing you the

938
01:07:55,804 --> 01:07:59,503
top, the most popular podcast that's not the way this

939
01:07:59,503 --> 01:08:02,284
is going to be Did I hear that? Maybe I

940
01:08:02,284 --> 01:08:05,963
misheard it. Did I hear that on Pod News Weekly

941
01:08:05,963 --> 01:08:10,204
Review Power, they were complaining about the Apple charts and

942
01:08:10,204 --> 01:08:15,123
all these Inception.ai podcasts are all in the top 20?

943
01:08:15,784 --> 01:08:18,503
Did I hear that right? I think I heard that

944
01:08:18,503 --> 01:08:20,884
right. I don't know. I haven't heard this week's episode

945
01:08:20,884 --> 01:08:24,423
yet. Maybe I have to look at it. That's the

946
01:08:24,423 --> 01:08:29,783
problem. Wouldn't surprise me. Yeah. Yeah. They left Spreaker. They're

947
01:08:29,783 --> 01:08:34,484
going somewhere else. Yeah, Megaphone. Yeah, they're going to Megaphone.

948
01:08:36,304 --> 01:08:40,484
Yeah, surprise, surprise. Good for them. Let me see. Top

949
01:08:40,484 --> 01:08:43,904
charts. Now it's Spotify's problem to deal with. Let me

950
01:08:43,904 --> 01:08:54,123
see. Top charts. Top shows. I think maybe they did

951
01:08:54,123 --> 01:08:57,203
a search. It was a search they did on like

952
01:08:57,203 --> 01:09:01,884
Charlie Kirk. Let me see. Charlie Kirk. That was probably

953
01:09:01,884 --> 01:09:07,543
it. Yeah, that probably was it. Yeah, yeah, exactly. Oh,

954
01:09:07,543 --> 01:09:10,503
yeah, here it is. Charlie Kirk Death, Inception Point, AI

955
01:09:10,503 --> 01:09:13,843
Inception. But that tells you something. That's an interesting signal.

956
01:09:15,644 --> 01:09:18,684
Because people are listening to Charlie Kirk Death, Charlie Kirk

957
01:09:18,343 --> 01:09:26,184
Biography Flash, Charlie Kirk The Aftermath, Biography Forever. People are

958
01:09:26,184 --> 01:09:30,423
listening to that. are they? I think they are. I'm

959
01:09:30,423 --> 01:09:33,963
not convinced. Let's listen to this. I'm not convinced at

960
01:09:33,963 --> 01:09:36,984
all. Deep dive into the life and death. Deep dive.

961
01:09:37,684 --> 01:09:40,364
I'm Emily Carter, your AI host. I want to remind

962
01:09:40,364 --> 01:09:44,823
you why that matters for this particular story. No, I

963
01:09:44,823 --> 01:09:46,703
do not believe it. I'm going to prove it. Okay.

964
01:09:48,564 --> 01:09:51,703
It's coming. How are you proving it? I will be

965
01:09:51,703 --> 01:09:54,503
proving it in the future. I've got an idea. Get

966
01:09:54,503 --> 01:10:02,024
some tissue. It's coming. Sorry. Bad joke from my old

967
01:10:02,024 --> 01:10:10,864
porn days. Okay, you sideline me with that one. No,

968
01:10:10,864 --> 01:10:16,543
I know exactly how to bring receipts on this, and

969
01:10:16,543 --> 01:10:18,503
I will be doing it. I'm cooking it up as

970
01:10:18,503 --> 01:10:20,463
we speak. I had a conversation with somebody about it

971
01:10:20,463 --> 01:10:23,404
yesterday. I just don't want to reveal it because I'm

972
01:10:23,404 --> 01:10:33,543
afraid it might pollute the results. Okay. Oh, man. Well,

973
01:10:33,543 --> 01:10:40,224
we got one more thing to talk about. Sure. HLS

974
01:10:40,224 --> 01:10:45,024
video. Yes, of course. Yes, yes, yes. What a beautiful

975
01:10:45,024 --> 01:10:51,283
mess this is. Nailed it. It's exactly what it is,

976
01:10:51,283 --> 01:10:59,984
a beautiful mess. So another one of the 47 things

977
01:10:59,984 --> 01:11:02,823
I was doing this morning was right before the show,

978
01:11:02,283 --> 01:11:08,743
I jumped on a call with Tom. Buzzsprout? From Buzzsprout,

979
01:11:08,743 --> 01:11:14,264
yeah. We were hashing out some stuff. And here's the

980
01:11:14,264 --> 01:11:20,843
issue with HLS video in alternate enclosure. everybody wants this

981
01:11:20,423 --> 01:11:24,264
I mean I mean like we want if you're going

982
01:11:24,264 --> 01:11:25,984
to make a video version stick it in the RSS

983
01:11:25,984 --> 01:11:31,864
feed right everybody wants it to be open in that

984
01:11:31,864 --> 01:11:36,343
way but the cost of abuse on your bandwidth to

985
01:11:36,343 --> 01:11:43,163
the hosting companies is absorbable when it's audio it's very

986
01:11:43,163 --> 01:11:47,524
damaging financially when it's video bots downloading Oh, yeah. Well,

987
01:11:47,524 --> 01:11:52,304
and I'll just add one other extra thing. Porn. Well,

988
01:11:52,304 --> 01:11:55,463
sure. Yeah, for sure. But just focusing on the bandwidth

989
01:11:55,463 --> 01:12:00,283
issue that can really hurt. You know, if if some

990
01:12:00,283 --> 01:12:05,404
bot goes wild and downloads downloads 10 terabytes of data

991
01:12:05,404 --> 01:12:10,764
of video streams. I mean, on a podcast that it

992
01:12:10,764 --> 01:12:13,623
just you can't it's hard. It it really does change

993
01:12:13,623 --> 01:12:17,944
the whole dynamic. Of course. What we're trying to figure

994
01:12:17,944 --> 01:12:22,604
out is if there's a way to sort of –

995
01:12:22,604 --> 01:12:27,364
before these HLS video streams get into the feed publicly,

996
01:12:27,364 --> 01:12:34,604
is there a way that we can try to keep

997
01:12:34,604 --> 01:12:41,884
the genie in the bottle bot-wise? And so we're just

998
01:12:41,884 --> 01:12:46,043
brainstorming some stuff saying, okay, rather than having to just

999
01:12:46,043 --> 01:12:54,463
build this enormous block list of user agent, ASN, IP

1000
01:12:54,463 --> 01:12:59,944
block combos, it just seems so difficult. Rather than doing

1001
01:12:59,944 --> 01:13:02,444
that, can we solve it in more of like an

1002
01:13:02,444 --> 01:13:08,463
elegant way where we do something like in order to

1003
01:13:08,463 --> 01:13:17,423
download the HLS video playlist from the alternate enclosure, it

1004
01:13:17,195 --> 01:13:22,115
would require something like an HTTP signature. Isn't this exactly

1005
01:13:22,115 --> 01:13:25,376
why or one of the reasons why Apple is doing

1006
01:13:25,376 --> 01:13:30,416
things the way they're doing it? I don't think they

1007
01:13:30,416 --> 01:13:33,055
care about the bandwidth. I think for them it's the

1008
01:13:33,055 --> 01:13:36,416
porn problem. For the hosting companies it's the bandwidth problem.

1009
01:13:35,435 --> 01:13:39,576
Right, right. I think there's two versions. There's – Apple

1010
01:13:39,576 --> 01:13:44,115
cares about the content. The hosts care about the bandwidth.

1011
01:13:44,676 --> 01:13:48,416
Because the hosts are already good at policing content. They

1012
01:13:48,416 --> 01:13:51,876
know exactly what's going into their system and how to

1013
01:13:51,876 --> 01:13:57,956
properly handle that legally. Apple basically is the reverse of

1014
01:13:57,956 --> 01:14:02,676
that. They are just sort of – they're like us.

1015
01:14:02,536 --> 01:14:07,195
They're a little bit vulnerable to people just sending them

1016
01:14:07,195 --> 01:14:11,336
a bunch of crap that they don't have a handle

1017
01:14:11,336 --> 01:14:15,315
on the content. Right. So they're a little bit at

1018
01:14:15,315 --> 01:14:18,055
the mercy of this incoming content. So I think they

1019
01:14:18,055 --> 01:14:24,916
have different needs. And so in the Mastodon activity pub

1020
01:14:24,916 --> 01:14:29,015
world, you have a thing. And we actually went over

1021
01:14:29,015 --> 01:14:32,336
this back when I was building the activity pub bridge.

1022
01:14:33,695 --> 01:14:39,595
You have a thing called HTTP signing or HTTP signatures,

1023
01:14:39,595 --> 01:14:42,256
and it uses a public key. So if you're on

1024
01:14:42,256 --> 01:14:44,555
a Mastodon account, if you have a Mastodon account or

1025
01:14:44,555 --> 01:14:48,076
any other really activity pub account on a modern activity

1026
01:14:48,076 --> 01:14:55,216
pub platform. That actor is going to have. And I

1027
01:14:55,216 --> 01:14:57,235
want to say actor in this context. I mean, the

1028
01:14:57,235 --> 01:15:01,296
activity pub host itself, because there is a the activity

1029
01:15:01,296 --> 01:15:05,555
pub server is also an actor. That server is going

1030
01:15:05,555 --> 01:15:09,435
to have a public key. And every time they make

1031
01:15:09,435 --> 01:15:13,456
a request to another activity pub server to do something

1032
01:15:13,456 --> 01:15:16,695
like post a post a toot or something like that,

1033
01:15:17,555 --> 01:15:23,876
they will sign that HTTP request using their private key

1034
01:15:24,676 --> 01:15:27,595
and also send with the headers, send a link to

1035
01:15:27,595 --> 01:15:29,855
their public key so that the receiving server can then

1036
01:15:29,855 --> 01:15:32,216
authenticate that they are the ones that signed that request.

1037
01:15:32,475 --> 01:15:36,315
Well, shouldn't we have something like this already anyway? Like,

1038
01:15:36,315 --> 01:15:38,975
yeah, this is a legit person with a legit key,

1039
01:15:39,355 --> 01:15:42,515
whether it's a browser or an app or whatever. That

1040
01:15:42,515 --> 01:15:45,536
makes so much sense. I'm a human. Here's my key.

1041
01:15:47,996 --> 01:15:50,595
yeah it makes sense across the board not just with

1042
01:15:50,595 --> 01:15:53,716
video yeah i mean for everything of course it won't

1043
01:15:53,716 --> 01:15:57,076
be so nice for the advertising community because we'll find

1044
01:15:57,076 --> 01:16:01,256
out how few people actually listen to this crap well

1045
01:16:01,256 --> 01:16:06,275
the so the like the initial thought process that we're

1046
01:16:06,275 --> 01:16:09,296
that we're having and this need whatever this whatever comes

1047
01:16:09,296 --> 01:16:13,115
out of this discussion really needs to become part of

1048
01:16:13,115 --> 01:16:22,115
the Podcast Standards Project recommendations is that we could do

1049
01:16:22,115 --> 01:16:27,015
something like a simplified version of HTTP signatures. It would

1050
01:16:27,015 --> 01:16:29,336
use the same algorithm. Everything about it would be the

1051
01:16:29,336 --> 01:16:36,095
same, except that HTTP signatures, the spec, I think, requires

1052
01:16:36,095 --> 01:16:41,655
a date header. But ultimately, you still have to issue

1053
01:16:41,655 --> 01:16:47,716
a key to a human being, right? A downloading app

1054
01:16:47,716 --> 01:16:53,095
client would have to publish a public key somewhere. Right.

1055
01:16:52,956 --> 01:16:58,216
But now you could have a million little servlets with

1056
01:16:58,216 --> 01:17:02,576
keys. So how do you determine ultimately this is a

1057
01:17:02,576 --> 01:17:06,496
human they're watching or they're listening with this key? How

1058
01:17:06,496 --> 01:17:11,055
do you make that connection? That's what it comes down

1059
01:17:11,055 --> 01:17:14,775
to. Well, you can. There's just a lot of unfriendly

1060
01:17:14,775 --> 01:17:18,475
ways, but we can do it. Well, you can by

1061
01:17:18,475 --> 01:17:21,475
going – yeah, exactly what you said. Unfriendly meaning we're

1062
01:17:21,475 --> 01:17:24,095
going to keep like a public repository where we vet

1063
01:17:24,095 --> 01:17:29,595
people or something like that. Precisely. That'll be us, podcastindex.org.

1064
01:17:29,695 --> 01:17:34,796
All your keys belong to us. Well, you know, I

1065
01:17:34,796 --> 01:17:37,855
actually see this being something if there was going to

1066
01:17:37,855 --> 01:17:41,015
be some sort of list. I see this being outside

1067
01:17:41,015 --> 01:17:45,576
of the podcasting 2.0 namespace. I see this as being

1068
01:17:45,576 --> 01:17:54,775
part of like the OPAWG. OPAWG. OPAWG. The worst name

1069
01:17:54,775 --> 01:18:01,676
ever. It's the Open Podcast Analytics Working Group. Oh, wow.

1070
01:18:01,695 --> 01:18:05,076
And do you know who the Open Podcast Analytics Working

1071
01:18:05,076 --> 01:18:08,635
Group consists of? Let me guess. I'm going to say

1072
01:18:08,635 --> 01:18:14,716
Libsyn. No. Okay. I give up. James Cridlin and John

1073
01:18:14,716 --> 01:18:20,416
Spurlock. OPOG. That's OPOG sitting over there in their rocking

1074
01:18:20,416 --> 01:18:24,376
chairs. Johnny and the Criddler are over there running the

1075
01:18:24,376 --> 01:18:28,336
OP. Johnny and the Criddler. They're over there running the

1076
01:18:28,336 --> 01:18:35,055
OP-AWG, which as of right now consists mainly of a

1077
01:18:35,055 --> 01:18:39,615
list of user agents. Right. Yeah, that's right. Yeah. And

1078
01:18:39,615 --> 01:18:42,975
it's a very nice list. Like it's very – Spurlock

1079
01:18:42,975 --> 01:18:46,456
keeps that thing really clean and up to date. So

1080
01:18:46,456 --> 01:18:49,536
to me, this would be a great thing to add

1081
01:18:49,536 --> 01:18:54,275
to the OPOG would be a list of validated public

1082
01:18:54,275 --> 01:18:58,115
keys. It doesn't mean that you have to be in

1083
01:18:58,115 --> 01:19:03,676
the list. It just means that people serving HLS content,

1084
01:19:03,676 --> 01:19:09,876
hosts could choose whether or not they want to. To

1085
01:19:09,876 --> 01:19:11,996
me, I see this as two layers. You have an

1086
01:19:11,996 --> 01:19:19,336
HTTP. You require HTTP signatures on all playlists in 3U8

1087
01:19:19,336 --> 01:19:23,336
downloads. You just have to have the HTTP signature to

1088
01:19:23,336 --> 01:19:27,515
even get it. then the hosting companies can opt for

1089
01:19:27,515 --> 01:19:32,176
the next level of stricture, which would be we now

1090
01:19:32,176 --> 01:19:35,676
– us, we have decided as a hosting company we're

1091
01:19:35,676 --> 01:19:41,055
only going to serve content to validated public keys that

1092
01:19:41,055 --> 01:19:46,076
exist on the OP AWG list. You know, the real

1093
01:19:46,076 --> 01:19:51,216
way to solve this is what's your Bitcoin address? Send

1094
01:19:51,216 --> 01:19:56,216
me one site. You're good. That will happen. No, it

1095
01:19:56,216 --> 01:19:59,615
will never happen. But I've always felt that every human

1096
01:19:59,615 --> 01:20:03,935
being should have a public key regardless. Every human being

1097
01:20:03,935 --> 01:20:06,176
should have a public key. This is the one thing

1098
01:20:06,176 --> 01:20:08,496
that – Issued at birth. Yes. Tattooed on the back

1099
01:20:08,496 --> 01:20:13,615
of your hand and on your forehead. Ah, yes. Boom

1100
01:20:13,615 --> 01:20:16,975
shakalaka. And there it is. That's right. You will need

1101
01:20:16,975 --> 01:20:19,216
that to do commerce. Oh, where have I read that

1102
01:20:19,216 --> 01:20:24,095
before? Grocery store, show me your public key. Yeah. Everybody

1103
01:20:24,095 --> 01:20:28,536
needs a pub key. I mean – Eric P.P. said,

1104
01:20:28,536 --> 01:20:31,095
that's how SSL certificates work. Send me money and I'll

1105
01:20:31,095 --> 01:20:34,275
vouch for you. Pretty much. Yep, that's the way it

1106
01:20:34,275 --> 01:20:38,376
works. It would solve so many problems. It would solve

1107
01:20:38,376 --> 01:20:41,876
a lot. The podcast index – we vouch for people

1108
01:20:41,876 --> 01:20:44,456
without them sending us money. That's right. We just do

1109
01:20:44,456 --> 01:20:47,115
it. All the time. We just vouch. Why not? We're

1110
01:20:47,115 --> 01:20:51,655
vouchers. We're vouchers. We're vouchers. We're vouchers. Send your email

1111
01:20:51,655 --> 01:20:57,496
to voucher at podcastindex.org. Let's thank some people here, Mr.

1112
01:20:57,496 --> 01:21:01,095
Jones. We've been getting lots of boostograms today, which is

1113
01:21:01,095 --> 01:21:06,996
nice to see. Right there, 49,950, which I'm thinking was

1114
01:21:06,996 --> 01:21:10,756
supposed to be 50,000, coming in from Fountain from CELOS

1115
01:21:10,756 --> 01:21:14,735
on Linux. And he says 50K for supporting slash bribing

1116
01:21:14,735 --> 01:21:18,635
the podcast index censorship validated key system. You're in, my

1117
01:21:18,635 --> 01:21:21,775
friend. You're in. You got it. We got you. You

1118
01:21:21,775 --> 01:21:26,315
are in. Let's see. From coming in from the podcast

1119
01:21:26,315 --> 01:21:30,956
index website itself, Four Sevens from Summer Surf. He says,

1120
01:21:30,956 --> 01:21:33,055
enjoying the afternoon in the boardroom. Thank you for your

1121
01:21:33,055 --> 01:21:37,796
courage. Four Sevens from Lyceum. Kudos to Alberto Batella and

1122
01:21:37,796 --> 01:21:41,576
RSS.com co-listing and chatting with Mike Newman as we speak.

1123
01:21:42,036 --> 01:21:44,956
Oh, there's Dreb Scott coming in with a baller. 10,000

1124
01:21:44,956 --> 01:21:49,135
sats. Voucher McVouch. Vouch, he says. Yes, it's working. Our

1125
01:21:49,135 --> 01:21:53,855
system is working. Already we're done. We just solved it.

1126
01:21:53,956 --> 01:21:57,655
We solved it and we got money. Literally says, yes,

1127
01:21:57,655 --> 01:22:01,695
this is how things should work. I get to stop

1128
01:22:01,695 --> 01:22:04,676
selling Tennessee trick shot out of the trunk of my

1129
01:22:04,676 --> 01:22:07,475
car. I had never heard of Tennessee trick shot. I

1130
01:22:07,475 --> 01:22:12,235
love that term. Oh, the moonshine? Yeah. Tennessee trick shot.

1131
01:22:13,355 --> 01:22:18,336
That's a good show title. There's like every different, every

1132
01:22:18,336 --> 01:22:21,435
state has its own little flavor. Tennessee trick shot. Dreb

1133
01:22:21,435 --> 01:22:23,815
Scott, one, two, three, four. He says, Dave, great job

1134
01:22:23,815 --> 01:22:27,475
with those massive dumps of the raw data. That's what

1135
01:22:27,475 --> 01:22:31,256
I'm good at. Big, massive dumps. 777 from Salty Crayon.

1136
01:22:31,376 --> 01:22:33,935
Dave, could you elaborate on how your index filter marked

1137
01:22:33,935 --> 01:22:37,475
a bunch of fountain tracks as spam in the pipe?

1138
01:22:37,916 --> 01:22:41,895
How did that happen? I don't know. I don't know.

1139
01:22:43,036 --> 01:22:45,055
You got a row of sticks from Hey Citizen sent

1140
01:22:45,055 --> 01:22:48,475
from the bathroom on his Cast-O-Matic. That's where it comes

1141
01:22:48,475 --> 01:22:53,515
in from. 777 from Mike Newman, True Fans Boost from

1142
01:22:53,515 --> 01:22:58,975
PC 2.0. Thank you. We've got 1221 from C-Loss and

1143
01:22:58,975 --> 01:23:02,796
Linux from Fountain testing this week, 1234. So it's odd.

1144
01:23:02,076 --> 01:23:06,076
The helipad keeps giving me different numbers because I know

1145
01:23:06,076 --> 01:23:11,015
he sent 1234, but it's showing me 1221. Chad F,

1146
01:23:11,015 --> 01:23:15,456
triple three. Dave, off the dome, Jones. Off the dome.

1147
01:23:15,756 --> 01:23:18,456
Triple three is from Eric PP. Pew Pew, yes. Three,

1148
01:23:18,456 --> 01:23:21,076
three, three from Chad F. This is all testing. Dreb

1149
01:23:21,076 --> 01:23:24,195
Scott, one, two, three, four. Boost McTest Boost. Pew Two.

1150
01:23:25,155 --> 01:23:27,355
I love people sending us money to test. This is

1151
01:23:27,355 --> 01:23:29,676
all great. Thank you, Eric PP. Thank you, Chad F.

1152
01:23:29,916 --> 01:23:32,115
And I hit the delimiter with that. So over to

1153
01:23:32,115 --> 01:23:38,975
you, Mr. Jones. on the music thing I mean a

1154
01:23:38,975 --> 01:23:44,176
lot of that music is spam it's just spam and

1155
01:23:44,176 --> 01:23:48,496
all spam classifiers have false positives and false negatives every

1156
01:23:48,496 --> 01:23:50,996
single one of them it's not a solvable problem so

1157
01:23:50,996 --> 01:23:56,315
I think it just comes down to that because there

1158
01:23:56,315 --> 01:23:58,716
has been a problem with people and this is happening

1159
01:23:58,716 --> 01:24:02,416
on not just I don't know what Fountain's posting but

1160
01:24:03,095 --> 01:24:05,055
Not just Fountain and Wavelike and all that. You have

1161
01:24:05,055 --> 01:24:08,775
to be careful. People are posting copyrighted material. Yes. I'm

1162
01:24:08,775 --> 01:24:11,536
seeing it on Spreaker a lot. It's coming through where

1163
01:24:11,536 --> 01:24:15,655
people are posting, like, entire albums of, you know, copyrighted

1164
01:24:15,655 --> 01:24:18,836
music. Oh, really? Yeah, you can't do that. And so

1165
01:24:18,836 --> 01:24:22,815
it's going to get called. Where was I? I was

1166
01:24:22,815 --> 01:24:24,576
looking at it. Yeah, I should mention this is a

1167
01:24:24,576 --> 01:24:28,055
value for value project. You get this podcast and the

1168
01:24:28,055 --> 01:24:32,615
entire podcast index and your data dumps, everything, all of

1169
01:24:32,615 --> 01:24:35,975
that stuff. All of that is openly available for everybody.

1170
01:24:36,155 --> 01:24:38,095
All you have to do is say to yourself, you

1171
01:24:38,095 --> 01:24:41,176
know, it's kind of valuable what's happening over there. Let

1172
01:24:41,176 --> 01:24:43,615
me send some value back to keep these guys running.

1173
01:24:44,916 --> 01:24:47,876
And that's how it works. So you go to podcastindex.org

1174
01:24:47,876 --> 01:24:49,796
down at the bottom there. You've got a big red

1175
01:24:49,796 --> 01:24:53,475
donate button. You can send us your Fiat Fun coupons

1176
01:24:53,475 --> 01:24:57,076
through PayPal or a Boostergram, which we always enjoy. We

1177
01:24:57,076 --> 01:25:01,296
enjoy reading them. And hopefully Dave has now summarized everything.

1178
01:25:00,855 --> 01:25:08,855
He's used the Excel Copilot widget. Yeah. Tell me how

1179
01:25:08,855 --> 01:25:12,595
much donations we've had. And you get back. It's like

1180
01:25:12,595 --> 01:25:17,155
carrot or something. So I'll tell you something that I

1181
01:25:17,155 --> 01:25:20,595
have trained my robot to do. Okay. So after every

1182
01:25:20,595 --> 01:25:25,576
no agenda show, we do credits. And the credits are

1183
01:25:25,576 --> 01:25:30,095
executive producer, people who have donated $300 or above, associate executive

1184
01:25:30,095 --> 01:25:34,496
producers between $200 and $300. So while the end of show mixes

1185
01:25:34,496 --> 01:25:37,055
are running, I say, robot, do the credits. It goes

1186
01:25:37,055 --> 01:25:41,036
into the Excel file. It pulls out the names from

1187
01:25:41,036 --> 01:25:43,815
the corresponding values and puts them right into my OPML

1188
01:25:43,815 --> 01:25:48,855
prep sheet. It's beautiful. It does that, I'd say, 99%

1189
01:25:48,855 --> 01:25:57,796
accuracy. Why am I doing this? You got this thing

1190
01:25:57,796 --> 01:26:02,155
nailed. Well, I don't have the spreadsheet. That's true. Yeah,

1191
01:26:02,155 --> 01:26:04,475
why are you doing it? Tell your robot to do

1192
01:26:04,475 --> 01:26:06,815
it for you. You got to be thinking like a

1193
01:26:06,815 --> 01:26:09,296
broadcaster, man. Like, what is the crap that I hate

1194
01:26:09,296 --> 01:26:12,855
doing? Well, I hate going through the spreadsheet and looking

1195
01:26:12,855 --> 01:26:16,756
for all those stupid jingles everybody wants. so the robot

1196
01:26:16,756 --> 01:26:20,376
looks at the spreadsheet pulls up the jingles and then

1197
01:26:20,376 --> 01:26:23,775
i have my own spreadsheet reader when i'm reading someone's

1198
01:26:23,775 --> 01:26:27,355
donation note i hit the the the cell the jingle

1199
01:26:27,355 --> 01:26:29,395
cell and it loads up those jingles right in my

1200
01:26:29,395 --> 01:26:33,176
player that saves me half an hour a show day

1201
01:26:33,176 --> 01:26:38,015
i have nothing like this it took me five minutes

1202
01:26:38,015 --> 01:26:41,515
to vibe code it baby well i can't even fix

1203
01:26:41,515 --> 01:26:45,676
the i can't even fix the the spreadsheet that's coming

1204
01:26:45,676 --> 01:26:48,735
out. I'll have my robot show your robot how to

1205
01:26:48,735 --> 01:26:51,695
do it. Have your robot talk to my robot. We

1206
01:26:51,695 --> 01:27:03,376
need that. We need inter-robot communications protocol. IRP. IRP. I

1207
01:27:03,376 --> 01:27:06,155
don't. I'm a little afraid of your robot talking to

1208
01:27:06,155 --> 01:27:08,855
my robot. Oh, okay. I don't know where your robot's

1209
01:27:08,855 --> 01:27:12,576
been. My robot is dirty. That's what I'm afraid of.

1210
01:27:12,615 --> 01:27:14,876
I'm afraid it's going to teach my robot bad things.

1211
01:27:14,876 --> 01:27:19,355
Very bad things. Guaranteed. In the words of my mom,

1212
01:27:19,355 --> 01:27:21,055
it's going to be a bad influence on my robot.

1213
01:27:23,055 --> 01:27:26,655
Okay, where? I forgot completely. Oh, Martin Lindiscoe. These are

1214
01:27:26,655 --> 01:27:29,735
PayPal's. He sent us a dollar. Thank you. Thank you,

1215
01:27:29,735 --> 01:27:34,815
Martin Lindiscoe. You and Mike Newman having a ball over

1216
01:27:34,815 --> 01:27:39,555
there. Brendan at PodPage, $25. By the way, congratulations to Brendan.

1217
01:27:40,115 --> 01:27:41,555
Are we allowed to see? Are we allowed to talk

1218
01:27:41,555 --> 01:27:47,076
about it? blah blah blah blah blah blah blah blah

1219
01:27:47,076 --> 01:27:51,176
blah blah blah blah blah blah blah blah blue no

1220
01:27:51,176 --> 01:27:53,735
whoo you saved me right there you saved me i

1221
01:27:53,735 --> 01:27:55,595
did thank you i did you're welcome that's what i

1222
01:27:55,595 --> 01:28:00,956
do uh let's see 25 dollars thank you from uh

1223
01:28:00,956 --> 01:28:04,796
brendan at pod page appreciate that uh what we got

1224
01:28:04,836 --> 01:28:13,695
Here we got Marco at, let's see, Marco Van Rij.

1225
01:28:14,095 --> 01:28:22,836
R-A-A-I-J. R-A-A-I. R-A-I. R-A-I. Okay, Marco Van Rij. We have

1226
01:28:22,836 --> 01:28:29,695
the boost. R-A-I. Sent us $10. Thank you, Marco. Oscar Mary, $200.

1227
01:28:30,275 --> 01:28:33,836
Thank you, Oscar. Oh, boy, hold on a second. keeping

1228
01:28:33,836 --> 01:28:41,216
it running thank you very much i incorrectly classify his

1229
01:28:41,216 --> 01:28:43,336
music podcast as spam and he still sends us money

1230
01:28:43,336 --> 01:28:49,416
uh joseph maraca five bucks thank you joseph appreciate that

1231
01:28:48,996 --> 01:28:57,135
uh oh thomas umstad a hundred bucks shot caller 20

1232
01:28:57,135 --> 01:29:02,195
inch blades on the impala that beats uh what is

1233
01:29:02,195 --> 01:29:07,456
it called tennessee trickster tennessee trick shot trick shot that's

1234
01:29:07,456 --> 01:29:09,935
it trick shot yeah years ago i came for the

1235
01:29:09,935 --> 01:29:13,935
podcast talk now i stay for the ai chatter adam

1236
01:29:13,935 --> 01:29:16,475
take it from a native texan the cure for cedar

1237
01:29:16,475 --> 01:29:20,756
fever is raw jalapenos oh oh that's an interesting one

1238
01:29:20,756 --> 01:29:23,235
yeah i'll try that if you can take the heat

1239
01:29:23,235 --> 01:29:27,055
they'll Clear your sinuses. Go podcasting. Yeah, go podcasting indeed,

1240
01:29:27,055 --> 01:29:33,296
baby. Yeah. Mitch Downey, 50 bucks. Thank you, Mitch. That's

1241
01:29:33,296 --> 01:29:36,676
the Podverse crew. That's straight from the whole group there.

1242
01:29:37,376 --> 01:29:43,155
Lauren Ball, $24.20. Thank you, Lauren. Appreciate it. Christopher Harbarik, $10. By

1243
01:29:43,155 --> 01:29:47,796
the way, seeing Christopher Harbarik's name reminded me that I

1244
01:29:47,796 --> 01:29:54,796
am also going through and putting individual creator attributions on

1245
01:29:54,796 --> 01:29:58,996
each of the podcasting 2.0 namespace tags. What? It's something

1246
01:29:58,996 --> 01:30:03,336
that we've – well, I mean each tag was initially

1247
01:30:03,336 --> 01:30:08,275
proposed by a person. Yeah. And I think it's important

1248
01:30:08,275 --> 01:30:12,296
for those people to get the credit for their ideas

1249
01:30:12,296 --> 01:30:17,836
becoming what eventually led to real namespace tags. Yeah, so

1250
01:30:17,836 --> 01:30:21,036
that in 25 years, Adam Carolla will have invented it.

1251
01:30:22,595 --> 01:30:25,376
This will be properly documented so that Adam Carolla does

1252
01:30:25,376 --> 01:30:30,055
not get credit for the chapter tag. But so like

1253
01:30:30,055 --> 01:30:32,975
Christopher Hart, so I'm doing it in this way. I'm

1254
01:30:32,975 --> 01:30:35,176
doing – there's going to be a creator attribution. And

1255
01:30:35,176 --> 01:30:37,256
that's the person that originally proposed it and sort of

1256
01:30:37,256 --> 01:30:41,015
sherpeted it all the way through to completion. And then

1257
01:30:41,015 --> 01:30:44,195
there's also contributors that are getting a byline, too, where

1258
01:30:44,195 --> 01:30:46,956
those are people that had significant input in the discussion

1259
01:30:46,956 --> 01:30:49,975
thread or the issue where that thing got hashed out.

1260
01:30:50,656 --> 01:30:55,355
And so, like, for instance, James, he's credited on the

1261
01:30:55,355 --> 01:30:59,735
location tag. Alex Gates is credited on the alternate enclosure

1262
01:30:59,735 --> 01:31:04,555
tag. Do I get any credits anywhere? No. Oh. You

1263
01:31:04,555 --> 01:31:12,435
get credit for being awesome. Seems empty. No. Why would

1264
01:31:12,435 --> 01:31:15,935
it seem empty? It seems empty. I've participated in every

1265
01:31:15,935 --> 01:31:20,275
single tag. I've published every tag available to man. That's

1266
01:31:20,275 --> 01:31:22,576
true. I get no credit, man. I get no credit.

1267
01:31:23,555 --> 01:31:25,895
You don't even get credit for podcasting. You give me

1268
01:31:25,895 --> 01:31:34,475
some credit. That goes to Corolla. Let's see. Christopher Harbaugh

1269
01:31:34,475 --> 01:31:37,336
got a contributor credit on one of the tags. I

1270
01:31:37,336 --> 01:31:41,055
don't remember which one. He gave us $10. Thank you, Chris.

1271
01:31:41,336 --> 01:31:46,195
Didn't you and I create the value tag? We did.

1272
01:31:46,515 --> 01:31:48,756
Yes. Okay. Me and you were co-creators of the value

1273
01:31:48,756 --> 01:31:51,435
tag. I want some credit. I want the credit. You

1274
01:31:51,435 --> 01:31:58,256
invented the podcast enclosure tag. You trump everything we do.

1275
01:31:57,796 --> 01:32:03,836
I want the credit, man. I just give me more

1276
01:32:03,836 --> 01:32:08,355
credit. I just give me some credit. Or hook or

1277
01:32:08,355 --> 01:32:11,176
some blow. Either one is fine. Yes, send them in

1278
01:32:11,176 --> 01:32:14,836
the mail. Mitch. This is Mitch on the personal side.

1279
01:32:15,376 --> 01:32:20,416
This is Mitchell Downey, $10. Thank you, Mitch. Damon Kasajak, $15. Thank

1280
01:32:20,416 --> 01:32:28,176
you. Appreciate you. Terry Keller, $5. Chris Cowan, $5. Thank you, Chris.

1281
01:32:31,216 --> 01:32:36,015
Silicon Florist $10, thank you Derek J. Visker the best

1282
01:32:36,015 --> 01:32:42,796
name in podcasting, $21 Paul Salzman $22.22 every time I love

1283
01:32:42,796 --> 01:32:54,576
it Jeremy Gerds, $5 Michael Hall, $5.50 Todd Cochran $30 and we

1284
01:32:54,576 --> 01:32:56,576
are going to have Mike Dell on the show next

1285
01:32:56,576 --> 01:33:02,135
week and he will fill us in on Blueberry and

1286
01:33:02,135 --> 01:33:05,055
what they've been doing and just the strategy, you know,

1287
01:33:04,395 --> 01:33:09,775
or direction of the company, um, post Todd and everything.

1288
01:33:09,315 --> 01:33:11,275
It'll be good to catch up with Mike. Hey, question

1289
01:33:11,275 --> 01:33:16,076
on this, um, this pending signups for API access, I

1290
01:33:16,076 --> 01:33:22,775
guess that is. Yes. So, I mean, how do we,

1291
01:33:21,895 --> 01:33:25,536
I mean, these could all be phony baloney people. Oh,

1292
01:33:25,015 --> 01:33:27,656
many of them are, I'm sure. Okay. All right. So,

1293
01:33:27,656 --> 01:33:30,576
but if someone sends me an email and my own

1294
01:33:30,576 --> 01:33:35,796
Spidey send says, yeah, they're okay, then I can mark

1295
01:33:35,796 --> 01:33:39,055
active and send credentials. Is that how it works? Yeah.

1296
01:33:39,055 --> 01:33:41,416
And honestly, that's going to be extremely rare in the

1297
01:33:41,416 --> 01:33:45,055
history. In the last six years, I've probably had that

1298
01:33:45,055 --> 01:33:47,876
happen. I can count it on one hand. Yeah, I'll

1299
01:33:47,876 --> 01:33:51,775
bet. I plan to never look at that screen. It

1300
01:33:51,775 --> 01:33:55,256
only became a thing because two people happened to have

1301
01:33:55,256 --> 01:33:58,036
a problem at the same time. I got it. Yep.

1302
01:33:59,376 --> 01:34:02,536
My typical thing is like we're a free service and

1303
01:34:02,536 --> 01:34:06,656
we're donation run. Send me money. Send me some credit.

1304
01:34:07,796 --> 01:34:10,355
We have rules. And one of the rules is you

1305
01:34:10,355 --> 01:34:12,635
can't sign up for an API account from a free

1306
01:34:12,635 --> 01:34:15,716
email address like a Gmail. Yeah, yeah, yeah. And when

1307
01:34:15,716 --> 01:34:17,435
people ask me and people are like, hey, I'm trying

1308
01:34:17,435 --> 01:34:18,876
to sign up with a Gmail account. Can you do

1309
01:34:18,876 --> 01:34:20,836
it? I just don't respond. I'm like, no, that's the

1310
01:34:20,836 --> 01:34:25,956
rule. That's great customer service. I'm like, I'm not going

1311
01:34:25,956 --> 01:34:28,615
to make exceptions because I don't know who you are.

1312
01:34:28,895 --> 01:34:32,775
Yeah, exactly. Go get a different – go spend $5 on

1313
01:34:32,775 --> 01:34:36,435
a domain from Hover and get a real email address.

1314
01:34:36,735 --> 01:34:39,256
And, I mean, it's not that much to ask. I'm

1315
01:34:39,256 --> 01:34:41,515
with you. Brother, I'm with you. I ignore all those

1316
01:34:41,515 --> 01:34:45,916
emails. You answer more than I do. I do pretty

1317
01:34:45,916 --> 01:34:49,935
good on answering, yeah. You do. Gene Liverman, five bucks.

1318
01:34:50,176 --> 01:34:52,656
Thank you, Gene. Oh, that's all of our PayPals. Oh,

1319
01:34:52,656 --> 01:34:55,775
we got some boost, though. Yes. We've got... Some boost.

1320
01:34:55,876 --> 01:35:00,036
Some boost. Oh, see, loss on Linux. 50,000. That's what

1321
01:35:00,036 --> 01:35:02,435
you want. You read that earlier. I did, yes. I'm

1322
01:35:02,435 --> 01:35:04,536
going to give him a baller for that. Baller! Shot

1323
01:35:04,536 --> 01:35:08,895
caller, 20-inch blades on the Impala. That's a nice boost.

1324
01:35:09,076 --> 01:35:12,956
Boost. How did I overlap you on that? Oh, I

1325
01:35:12,956 --> 01:35:14,876
know what I did. Yeah, I know why. Because I

1326
01:35:14,876 --> 01:35:16,595
did it later on in the show because I forgot

1327
01:35:16,595 --> 01:35:21,256
to earlier. Ah, yes. Okay. Oh, and that's all I

1328
01:35:21,256 --> 01:35:22,676
got. The only other one I've got is the comic

1329
01:35:22,676 --> 01:35:29,176
strip blogger, 19,000 sats through Fountain. And he says, howdy,

1330
01:35:29,176 --> 01:35:33,656
Dave and Adam. Please bookmark this. HTTPS colon slash slash

1331
01:35:33,656 --> 01:35:39,036
wherever dot audio by Andrew Grumet. Quote, it's a podcast

1332
01:35:39,036 --> 01:35:43,595
listening software that works in browsers, mobile and cars. No

1333
01:35:43,595 --> 01:35:47,036
logging in and nobody watching you. All data stays on

1334
01:35:47,036 --> 01:35:51,515
your phone browser. Easy import and export for portability. fast

1335
01:35:51,515 --> 01:35:57,076
QR code-based sharing. End quote. Yo, CSBAI archwizard. Yeah, I

1336
01:35:57,076 --> 01:36:00,055
love what Andrew has done. Andrew Grummet, by the way.

1337
01:36:00,235 --> 01:36:03,376
No credit. He's been on the show. If he gets

1338
01:36:03,376 --> 01:36:06,615
no credit. He'll get it. He needs a lot of

1339
01:36:06,615 --> 01:36:09,475
credit. No, he's going to get it. Because he's got

1340
01:36:09,475 --> 01:36:14,836
stuff in the namespace. Oh, okay. You get credit. Here's

1341
01:36:14,836 --> 01:36:18,195
some credit. You get some credit. And you get some

1342
01:36:18,195 --> 01:36:22,635
credit. Everybody gets some credit. Everywhere.audio. It's a great audio.

1343
01:36:22,796 --> 01:36:27,555
A very business-like professional note there from CSB this week.

1344
01:36:28,395 --> 01:36:30,336
All right, brother. I'm going to let you go because

1345
01:36:30,336 --> 01:36:32,815
you need to throw some things in the bag. You

1346
01:36:32,815 --> 01:36:34,595
know you're going to wind up on this camping trip

1347
01:36:34,595 --> 01:36:36,456
and it's like, what do you mean we've got no

1348
01:36:36,456 --> 01:36:41,895
pigs? What? Yes. We're all out of pigs. We don't

1349
01:36:41,895 --> 01:36:44,176
have any matches. Oh, man, what are we going to

1350
01:36:44,176 --> 01:36:47,076
do? And I'll see you at 3 o'clock, right? We've

1351
01:36:47,076 --> 01:36:50,395
got our early meeting. Exciting things happening with the podcast.

1352
01:36:50,416 --> 01:36:52,416
I thought it was 2.30. I thought it was 2.30.

1353
01:36:52,555 --> 01:36:54,796
If it's 2.30, is it 2.30? Well, I got to

1354
01:36:54,796 --> 01:36:59,775
hurry then. I got to rock and roll. Okay, run

1355
01:36:59,775 --> 01:37:02,256
and rip. All right. Hey, boardroom, thank you very much

1356
01:37:02,256 --> 01:37:05,095
for being here. Thank you for providing your very valuable

1357
01:37:05,095 --> 01:37:08,015
feedback, and we look forward to seeing each other again

1358
01:37:08,015 --> 01:37:10,656
for the board meeting next Friday. Until then, bye-bye, everybody.

1359
01:37:24,916 --> 01:37:30,676
Podcasts are cool. You have been listening to Podcasting 2.0.

1360
01:37:30,815 --> 01:37:38,496
Visit podcastindex.org for more information. Go podcasting! Get some tissue.

1361
01:37:38,496 --> 01:37:38,956
It's coming.