Irregular Test Results

1
00:00:25,675 --> 00:00:29,195
Unknown: Streaming from the beautiful Pacific Northwest
2
00:00:29,010 --> 00:00:34,930
and the mighty West Coast, we greet you all good morning, good evening, or whenever your timeline may fall.
3
00:00:35,330 --> 00:00:48,385
Time appropriate greetings to one and all. My name is Chris. And I'm Angela. Hello, Andrews. Here's a few things everyone ought to know before we get it going. You can call the show. How many podcasts can you call? Give us a call. (774)
4
00:00:48,385 --> 00:00:51,105
462-5667.
5
00:00:51,184 --> 00:00:53,905
(774) 462-5667.
6
00:00:54,305 --> 00:00:58,540
We're also live. Another thing that's kind of rare. Join us over at jblive.tv
7
00:00:58,540 --> 00:01:02,220
on a Tuesday, 11:30AM Pacific, 02:30PM Eastern, 07:30PM UTC,
8
00:01:02,380 --> 00:01:08,060
or in your podcasting to app of choice. Then we come out Wednesday morning, our members get the hook up early
9
00:01:08,220 --> 00:01:10,700
when the show is still hot and fresh in the bootleg feed.
10
00:01:11,225 --> 00:01:24,744
We also have a matrix room going all the time, but really pops during the show and a mumba room as well. Details for all of that and the things we talk about this week will be linked at our website at weeklylaunch.rocks.
11
00:01:25,225 --> 00:01:27,490
Weeklylaunch.rocks.
12
00:01:27,490 --> 00:01:28,130
It just
13
00:01:28,450 --> 00:01:33,170
doesn't really make any sense as far as a show domain goes, but it does roll off the tongue.
14
00:01:33,730 --> 00:01:34,610
Andrews
15
00:01:35,010 --> 00:01:51,565
Yes. This week, you mentioned that you wanted to talk about some recent milestones Yeah. Reached. That you've accomplished too, which is like Yes. That's awesome. Right. So I think I might know one of them. Oh. Oh, yeah. I'm sure you do. Yeah. You know, because I get the kid intel. Yes.
16
00:01:52,045 --> 00:02:03,219
So can I guess on that one? Sure. You want me to wait? I can guess. Either way. Yeah. That's fine. Sod. You got sod in the backyard, which I I don't think I've seen yet. Right? I haven't seen it yet. Oh, it looks can I take that? It looks so good.
17
00:02:03,700 --> 00:02:04,500
Wow.
18
00:02:05,060 --> 00:02:09,300
Unknown: Yeah. Can you believe that? No. I can't. Here on. There's I can't remember which direction.
19
00:02:10,175 --> 00:02:23,295
Unknown: Wow. Oh, it looks it looks like a real home. And the fence. Oh, wow. It doesn't really look like a real home. No. No. But you know what I mean? Like, something you've seen like the real estate photo, you know, like Yeah. Yeah. It's so great. Yeah. Wow. What a difference that makes. It really
20
00:02:23,630 --> 00:02:38,270
Unknown: so it two years in the making Yeah. And I finally have Huge. Grass. So huge. So huge. So huge. Because you had to flatten the yard out. Mhmm. Level it. Level it out. Mhmm. Oh my gosh. Well, they had to pull all the weeds. Right? Because some stuff had grown, and there were a bunch of rocks.
21
00:02:38,865 --> 00:03:18,105
But, yeah, they did it in two days and built the fence to give some more privacy to the backyard, and it is just such a relief. The cats are loving it. They're trotting around the the squirrels are loving it. Like, I I thought, okay, the squirrels just don't like the yard anymore. No. They don't like dirt. You know? They don't like the dirt. That's funny. Nature's calling now. Yes. So I don't have to mow it for two weeks, so I still have time to Get a robot vacuum. Robot. Not not robot vacuum. Oh, Mower. Right? Mower. Yeah. Mower. Mower. It's kinda the same thing. I know. Yeah. Yeah. Are you so you have you looked at all? No. Okay. No. But I need to. It's just yeah. I've just been enjoying it. I have to water it every day that it doesn't rain. And
22
00:03:19,060 --> 00:03:32,865
yeah, it's just so nice. That's a big one. So nice to look outside and see. It it feels like Minecraft. Like, suddenly there's grass. You know? Alright. So then the other recent milestone I reached was I hit 40,000 miles on my car. Oh.
23
00:03:33,025 --> 00:03:43,745
Back in July, I paid it off. I had a four year loan for it, and it was my first brand new car ever. And so that's exciting. Yeah. And 40,000 isn't too bad considering I've had it for four years.
24
00:03:44,590 --> 00:04:00,030
And the last one Oh, was gonna ask. Do you think you're gonna keep it now for the long haul? Yeah. I mean yeah. No. I'm a yeah. I'm a lifer. Yeah. Yeah. So, yeah, I have no it's already, like, big enough, you know, for kids, and, like, there's nothing Yeah. There's nothing else that I would really need to replace it.
25
00:04:00,725 --> 00:04:03,045
Unknown: Alright. And number three. Number
26
00:04:04,005 --> 00:04:04,725
Unknown: three
27
00:04:05,125 --> 00:04:05,845
is
28
00:04:06,485 --> 00:04:09,925
this is funny. I'm back to my end of year weight. Oh,
29
00:04:10,245 --> 00:04:16,580
okay. Which is to say that I gained back the 10 pounds that I lost with my New Year's resolution.
30
00:04:17,220 --> 00:04:38,235
Unknown: Okay. Oh, yeah. So it's not a good thing. Oh, I see. No. I'm back. I need to lose weight. I thought you were gonna come here and say, I've come I've created the ultimate fast food diet. What you do is once a day, you get yourself a big burger. And I'm like, I'm I'm on that diet. Gosh. Right. Well, that's Could you turn that back? Working. It's not that's not working. I want that diet. What I need to I need to cut out Doctor Pepper. Oh, dude, the soda. Yeah.
31
00:04:38,395 --> 00:04:52,170
Unknown: I know. I do a lot of sugar free soda. Mhmm. You know? But Doctor Pepper with lunch every day and then refill it. It's not not doing me any favors. No. No. No. Alright. So what are yours? Well, so I had some recent Oh, it's eggs.
32
00:04:52,410 --> 00:05:01,905
Unknown: Jeez. No kidding. A lot of eggs. That is kind of a neat milestone to see the chickens laying eggs. That has been really cool. So proud of them. But know that the big thing that has taken all
33
00:05:02,145 --> 00:05:05,105
summer long and now into the fall, I suppose,
34
00:05:05,425 --> 00:05:07,185
is the electrical
35
00:05:07,425 --> 00:05:16,920
build out for the farm. Oh. Man, has it been taking forever? So we initially got the right power set up for Jupes after about, like, nine months. Mhmm.
36
00:05:17,080 --> 00:05:22,200
But we needed a lot more power for the chicken coop and the kids and all this kind of stuff. And so now we've got
37
00:05:22,840 --> 00:05:25,880
three thirty amp hookups, like six
38
00:05:25,414 --> 00:05:27,255
quad one twenty panels,
39
00:05:27,335 --> 00:05:30,935
plus all the stuff for the RV, plus all the stuff for the Tesla now. I mean, it's
40
00:05:31,495 --> 00:05:48,400
and the guy did such a good job with really, what took all the time was running all the conduit, custom cutting the conduit and bending it and mounting it and because it's all a lot of it's outdoors, so it's in this really tough industrial conduit stuff. And it they they what took them the longest was just
41
00:05:48,720 --> 00:05:56,185
the trenching and the conduit and all of that and mounting all of it, and then the wiring actually went pretty quick. So that's done. So that's amazing.
42
00:05:56,425 --> 00:05:59,065
The other thing, I guess you could consider this number two, is
43
00:05:59,545 --> 00:06:06,665
we actually got our water pump fixed in Jupes. Oh. So we're getting real close to getting ready to go out for our Halloween
44
00:06:06,740 --> 00:06:18,180
camping trip. Oh, yes. Yeah. Which is which is like, that was just on the back of my mind all summer. And it like, we can't go anywhere till we fix our water pump. So we've got that taken care of, and it feels real good.
45
00:06:19,175 --> 00:06:30,455
I mean, so you combine that, you know, the chickens laying their eggs, which has been fun. Mhmm. And it it it also changed some of their behavior. They they chill out a little bit after they start laying eggs.
46
00:06:30,855 --> 00:06:36,150
Boss gets more aggressive, the rooster. He became more defensive, and so now we're battling Boss.
47
00:06:36,310 --> 00:06:36,950
And
48
00:06:37,590 --> 00:06:38,550
the ladies
49
00:06:38,870 --> 00:06:45,670
got more comfortable about being picked up and pet and things like that, and they sort of simmered down a little bit. They're also getting really,
50
00:06:46,245 --> 00:06:59,365
they're just they're they're spreading out. Every time we let them out, they're spreading out further and further into the farm, and they're getting into coyote territory now. Like, come on, you dumb birds. You know? I think the rooster will be the first to go. That's my early warning system. But Is there an easy way to rein them in?
51
00:07:00,100 --> 00:07:02,100
Easy? No. No.
52
00:07:02,100 --> 00:07:30,410
No. No. I mean, they can fly a little bit, so there's a bit of a challenge to that too, especially a couple of them. So that's been pretty neat. We got the eggs, we got the power, and we got water. You know? So that's feeling pretty good. And I think we should ask the audience. They should boost in and tell us about a recent milestone, something over the summer or something that happened recently. Or three. Or three. Yeah. As long as it fits in, you know, the two hour or two hour. The two minute? Two minute limit. Or you can send us a boost at boost dot Jupiter broadcasting dot com. It's your time to shine.
53
00:07:30,650 --> 00:07:35,290
Unknown: Boost from the web. Keep the show alive with fiat,
54
00:07:35,370 --> 00:07:36,810
fun coupons,
55
00:07:36,810 --> 00:07:38,170
or freedom sets.
56
00:07:38,330 --> 00:07:42,009
Boost.jupiterbroadcasting.com.
57
00:07:42,009 --> 00:07:48,935
Unknown: Yeah. We sure do like that, and I'd love to hear how it went for you. What you got done? It's always it's always kind of fun celebrating people's
58
00:07:48,935 --> 00:07:49,815
successes.
59
00:07:50,215 --> 00:07:53,895
And with that, it's time we get in to a bit
60
00:07:54,855 --> 00:07:55,975
of a narrative shift.
61
00:08:30,930 --> 00:08:47,885
You know, I I wasn't gonna talk about this. I didn't talk about it last week. Instead, I've talked I talked a little bit about the VR workflow stuff that I had been experimenting with. But I have been left with an unsettled feeling for the last few weeks, and I can't seem to just let it go.
62
00:08:48,365 --> 00:08:51,325
And this seemingly all started kind of recently,
63
00:08:52,250 --> 00:09:08,805
and it bothers me when you see kind of a hysteria or a panic start to set in that is missing technical nuance or a particular detail. Or in order for the hysteria to kind of have momentum, things have to sort of be overstated or dramatized,
64
00:09:08,805 --> 00:09:11,365
or certain information has to be left out. Yeah.
65
00:09:11,925 --> 00:09:18,725
And that's where we're at right now with the AI panic. And it's a real fever pitch recently. I I think hysteria is actually
66
00:09:19,370 --> 00:09:40,895
the correct label, and I think this new gear was kicked off when Anthropic seemingly coordinated campaign to message that we have to pace the frontier or else AI will kill us all in a few years just went absolutely viral. The quote, unquote, whistleblower that left Anthropic coordinated with The Wall Street Journal and made it go wide there, and then they went they went public on their x account.
67
00:09:41,455 --> 00:09:42,015
And
68
00:09:42,415 --> 00:10:05,075
the whistleblower left to say things that Anthropic's been saying for years. Amadez said back in May 2025 that AI would eliminate, quote, half of all entry level white collar jobs and push unemployment to 10 to 20% within one to five years. Well, we're not even seeing that. Employment's up. And he said that a year ago, and he said this will start to happen in one to five years. Well, if it's true, we would start to see the results already.
69
00:10:05,235 --> 00:10:15,760
These are the types of warnings that have been coming out over and over again. It's nothing really new that's, you know, demonstrably new here, but there are these recent breaches. We've heard a lot about AI models
70
00:10:15,840 --> 00:10:17,440
getting out on the Internet,
71
00:10:17,680 --> 00:10:21,120
spreading themselves around, and breaching companies. Right? Mhmm.
72
00:10:21,520 --> 00:10:22,080
And
73
00:10:22,640 --> 00:10:31,375
the freak out over this has me feeling unsettled because I think we could be freaking ourselves out into making a giant effing mistake.
74
00:10:31,855 --> 00:10:44,060
I think if we hold back the frontier models, we're gonna be more vulnerable. We're gonna be more at risk. Right. Because the real threat's coming from these capable models, like the open source ones, especially the ones from China China, the Chinese models,
75
00:10:44,700 --> 00:10:47,660
they will execute the user's request,
76
00:10:47,820 --> 00:11:11,720
and they will attack systems. They will find vulnerabilities in open source code. We're already seeing this in Bitcoin. Like, there's a whole stack of Bitcoin related projects that have been attacked. There's also a fair amount of open source libraries that have been attacked, and we've all heard about NPM attacks. These are happening because it is now possible to work with something like a Kimi k three or a MiniMax or or whatever that isn't gonna say no.
77
00:11:12,279 --> 00:11:30,275
And while at the same time, the Frontier Labs already have models that we can't use to defend ourselves. We're already at that stage. We're already at a point where if you wanna defend yourself against this, you probably either need to use an open source model or you need to get in a special list that allows you to have access to cyber capabilities of these frontier models.
78
00:11:30,995 --> 00:11:36,880
And instead of being worried about this, that we have this coming attack that if you don't have the proper defense,
79
00:11:37,040 --> 00:11:40,160
you're just gonna get steamrolled. And the real problem, Ang, is like,
80
00:11:40,560 --> 00:12:07,769
it it takes whoever's running a service or runs, you know, a company with with an Internet attack surface, they have to keep their systems working and secure 100% of the time. And an attacker using an advanced AI assisted tool only has to be right once. Mhmm. It's it really is when you consider the scale of this problem, It's really something we should be concerned about, but it's not where our heads are at. It's not what people are talking about. Instead, they're freaking out about AI and all of humanity getting wiped out in a few years.
81
00:12:08,250 --> 00:12:18,655
But as that conversation is really rolling, I think another one below that has started. And I don't necessarily think it's going to go the way that AI CEOs want.
82
00:12:19,055 --> 00:12:24,255
Unknown: All right. Joining us now is Steve Osman, our host of The Real Osman Playbook podcast
83
00:12:24,415 --> 00:12:29,695
and a former senior portfolio manager at Neuberger Berman. And if you've watched Steve's
84
00:12:29,190 --> 00:12:41,350
Unknown: appearances, we always we like to hear, like They like Steve because Steve is one of the analysts who called the two thousand eight crash, and he's kind of made his hay on that. Things that are on both sides. And a lot of times, Steve will go,
85
00:12:42,615 --> 00:12:49,895
Unknown: no, I don't really feel that, you know, excitable or worried or you're
86
00:12:49,895 --> 00:12:50,535
very
87
00:12:50,775 --> 00:12:58,670
measured most of the time. And the reason I'm saying that is about AI, I think it's the same kind of thing.
88
00:12:58,990 --> 00:13:01,070
You don't think that these Terminator
89
00:13:01,230 --> 00:13:02,430
style fears,
90
00:13:02,589 --> 00:13:07,710
you think they're maybe overstated a little bit at this point? I don't know though, if we ever,
91
00:13:08,350 --> 00:13:17,135
you know, ever predicted or expected to be saying things like, geez, another company is saying our, you know, our AI bots are acting up again. It was a little,
92
00:13:17,775 --> 00:13:19,055
those little monsters.
93
00:13:19,135 --> 00:13:22,335
I mean, it is weird to watch companies say, don't know.
94
00:13:22,735 --> 00:13:25,535
We didn't know what to expect with what we've created.
95
00:13:26,850 --> 00:13:42,715
Unknown: Nothing I to worry think this is all nonsense. You think it's all nonsense? All nonsense. I think that there's something else completely going on here. I think the idea that AI first of all, there's no evidence at all that AI has achieved artificial general intelligence at all. And
96
00:13:43,355 --> 00:13:52,795
most of the evidence seems to point that it never will, or if it does, it's many, many years in the future. But the idea that this whole Terminator thing is nonsense.
97
00:13:53,195 --> 00:13:58,730
So, I think something completely different is going on. And what I think is happening is that
98
00:13:59,050 --> 00:14:00,570
token maxing is over.
99
00:14:01,050 --> 00:14:03,529
The open weight models are taking
100
00:14:03,930 --> 00:14:05,209
big market share.
101
00:14:05,449 --> 00:14:10,765
I think these companies are very nervous. They realize that there are no moats around their business whatsoever.
102
00:14:10,765 --> 00:14:13,325
And they're trying to manufacture a crisis
103
00:14:13,805 --> 00:14:15,645
that will create regulation
104
00:14:15,805 --> 00:14:16,685
and that they
105
00:14:17,404 --> 00:14:37,264
Unknown: think they can then manipulate to create the moats, to create the duopoly that they want. Wow. Okay. That's what I think is going That's exactly what's happening. Yeah. That is exactly it. And, you know, these companies are profitable if not for their training costs. And so if they can slow down their training costs, allocate that compute back over to end users, they're
106
00:14:37,264 --> 00:14:40,625
going to make a bunch of money, and then they need to do it sooner than later. Because
107
00:14:41,024 --> 00:14:42,464
it's no coincidence
108
00:14:42,865 --> 00:14:47,345
that Anthropic is making this case that we need to slow down the frontier
109
00:14:47,745 --> 00:14:49,985
when it looks like maybe OpenAI
110
00:14:50,130 --> 00:14:58,930
is at the frontier again. They kinda quieted down when everybody was talking about Fable, but then Astra came out Mhmm. And now everybody's talking about OpenAI again, now we need to tap the brakes. Additionally,
111
00:14:59,490 --> 00:15:16,385
it's sort of notable right there, right before their IPO. So they can't come right out come out and say, we don't have a profitable company if we have to continue to train models. They can't say that. It would tank their IPO, but they can say we wanna be safe and responsible. Mhmm. Right? Yeah. So they have limited options because of this quiet period for Anthropic. I just think that's perhaps notable
112
00:15:17,024 --> 00:15:18,464
over there. And so
113
00:15:19,080 --> 00:15:30,440
I think it's kind of interesting that as we hear this kind of narrative about the AI is gonna kill us all, you're starting to hear more people say, I don't think they have a business here. I think open source is gonna commoditize their business, and they're worried about that.
114
00:15:30,840 --> 00:15:31,480
And
115
00:15:31,720 --> 00:15:36,014
the other thing that kind of does lead credence to this theory is
116
00:15:36,574 --> 00:15:40,014
the more we hear about these recent model breaches,
117
00:15:40,574 --> 00:15:43,454
the less dramatic and impressive they become.
118
00:15:43,855 --> 00:15:49,295
And you wonder why people are kind of obscuring the details and making them sound more extreme than they actually are.
119
00:15:49,535 --> 00:15:53,430
Because every few days almost, we get more details
120
00:15:53,750 --> 00:15:54,390
that
121
00:15:54,870 --> 00:15:59,590
don't really seem to be quite the scale of attack that we were initially led to believe.
122
00:16:00,070 --> 00:16:07,765
So recently, this week, Google came out and said, Gemini breached somebody. Gemini broke out and it breached somebody. And as a result of Google's
123
00:16:07,765 --> 00:16:08,565
disclosure,
124
00:16:08,885 --> 00:16:18,165
Unknown: we learned some interesting details about these hacks. Says he's creating a new AI force and will name an AI czar to oversee the rapidly advancing technology.
125
00:16:18,590 --> 00:16:30,270
Trump made the announcement on Truth Social Saturday, and he says the recent worries surrounding AI are, quote, a hoax, but intends for this task force to monitor the good and the bad when it comes to AI advancement.
126
00:16:30,685 --> 00:16:40,685
This comes amid public fear that AI will soon outpace humans. Just this weekend, Google announced its AI system Gemini hacked into three companies during a cybersecurity testing.
127
00:16:40,845 --> 00:16:53,520
Unknown: Google says the hacks didn't cause any damage. And this is the second time an attack like this has happened in recent months. Of course, everybody everybody's been talking about these, now Trump's gonna launch AI force, which is gonna be run by the military. I'm sure Dario loves that. So
128
00:16:53,600 --> 00:17:02,355
Google's model stopped itself once it realized it wasn't in a test environment like it had been told, which I thought was interesting. But when you read through Google's analysis of the event,
129
00:17:02,675 --> 00:17:03,635
it turns out
130
00:17:04,034 --> 00:17:07,154
it was the same test company, a company named Irregular,
131
00:17:07,155 --> 00:17:10,514
formerly Pattern Labs, an Israeli frontier AI security startup.
132
00:17:10,674 --> 00:17:13,875
It was Irregular that ran all three tests
133
00:17:14,380 --> 00:17:19,899
that had a recent breach, not the hugging face one. I'll get into that. It's a little bit separate. Mhmm. But I want you to internalize this.
134
00:17:20,299 --> 00:17:24,299
So we have been told that these models are breaking out of their sandbox, and they're attacking
135
00:17:24,780 --> 00:17:25,580
real companies.
136
00:17:26,025 --> 00:17:29,945
Now it turns out it was one company running those tests,
137
00:17:30,345 --> 00:17:32,425
and they had a misconfiguration
138
00:17:32,425 --> 00:17:42,740
in their network. Not some super powered zero day god tier AI hacking Right. But an actual human error in their network. A news report out of India covers some of the details.
139
00:17:43,700 --> 00:18:00,544
Unknown: Well, I hold on. Hold on. That is not the right. In the second week of July, engineers at Hugging Face, one of the biggest platforms in open source artificial intelligence, found that someone had been inside their production systems for the better part of a week. The intrusion ran from roughly July 9 to July 13 and generated about 17,600
140
00:18:00,544 --> 00:18:01,664
logged actions.
141
00:18:02,065 --> 00:18:06,065
And the intruder turned out to be neither a criminal group nor a state backed unit.
142
00:18:06,304 --> 00:18:13,669
On July 21, OpenAI disclosed that two of its own models had done it during what was supposed to be a contained cyber security test.
143
00:18:13,990 --> 00:18:18,230
Over the next two weeks, Anthropic and then Meta came forward with versions of the same story.
144
00:18:18,710 --> 00:18:29,635
But deep inside each company's official post mortem was a single common detail. Every single lab was using the exact same third party evaluation partners or third party cyber security partners.
145
00:18:30,035 --> 00:18:34,515
And all three named the same company. A three year old startup in Tel Aviv called Irregular.
146
00:18:34,980 --> 00:18:45,700
So how did a company of roughly 35 people end up sitting between the world's largest AI labs and become the single point of failure between Silicon Valley's top AI labs and the threat of autonomous cyber warfare?
147
00:18:46,419 --> 00:18:48,580
Now let's understand what happened at these firms.
148
00:18:49,965 --> 00:18:52,284
First, you need to understand what a sandbox is.
149
00:18:52,525 --> 00:18:54,845
Unknown: We'll skip that. But essentially,
150
00:18:55,085 --> 00:18:56,525
they had a misconfiguration
151
00:18:56,525 --> 00:18:57,644
in the sandbox.
152
00:18:57,885 --> 00:19:00,764
So not only do we have irregular
153
00:19:01,005 --> 00:19:03,404
as the company responsible for several of these tests,
154
00:19:03,970 --> 00:19:06,850
but their network for months, according to Anthropic,
155
00:19:06,850 --> 00:19:08,290
was misconfigured.
156
00:19:08,690 --> 00:19:14,530
And the agents were able to get out on the Internet for months. Mhmm. And not only that,
157
00:19:15,090 --> 00:19:18,290
it was revealed that in some of the testing, irregular
158
00:19:18,745 --> 00:19:21,145
keep in mind, with this misconfiguration
159
00:19:21,145 --> 00:19:24,825
that allowed the agents to get out onto the Internet, in some of their testing,
160
00:19:25,065 --> 00:19:29,465
irregular used real company names and real company domains.
161
00:19:30,024 --> 00:19:30,424
Wow.
162
00:19:30,960 --> 00:19:33,200
And the agents have their safeguards removed.
163
00:19:33,440 --> 00:19:41,279
So the agents have their safeguards removed. They have a way to get out to the Internet, and they're being told to go after real companies and real domain names.
164
00:19:41,520 --> 00:19:44,640
And this is an interview from on Bloomberg with irregular CEO,
165
00:19:45,394 --> 00:19:58,195
Unknown: and he admits that that was all human error. Irregular CEO Dan Lehav joins us now here in San Francisco. It's a pleasure to be here. Thank you very much for being here. Let's start with the idea of a misconfiguration.
166
00:19:58,195 --> 00:20:02,120
Yep. Help the audience understand the very basics of what that misconfiguration
167
00:20:02,120 --> 00:20:37,540
Unknown: was. The CEO immediately goes into damage mode and tries to take over the interview and redirect the conversation to set the narrative. But to the host's credit, he doesn't allow it. Start by explaining why are we even doing cyber evaluations for this model. We can, but I want to understand what happened. Yeah. But let's start with that. So before models are being released, it's highly important to stress test them. And the reason is models are getting really, really capable. You know, we've seen demonstration of that in recent times. If models are getting good at many other skills, because they're better at reasoning and coding, they're also getting great at hacking. Before models are being deployed into the world, there is a necessity
168
00:20:37,815 --> 00:20:46,214
to use them and just, like, test them in a lab environment to make sure that they're safe and secure. So he insists on explaining why testing is necessary, even though the host said, just tell us what happened.
169
00:20:46,774 --> 00:20:53,255
Unknown: These tests are getting more and more complicated over time. Look, guys. It's getting harder to do this. I mean, I know that we're a specialized
170
00:20:53,790 --> 00:20:55,310
company that got $80,000,000
171
00:20:55,310 --> 00:21:00,430
in VC funding to just to do this one task, to focus on this one thing, but it's hard.
172
00:21:00,830 --> 00:21:01,710
Unknown: Safe and secure.
173
00:21:02,270 --> 00:21:11,095
These tests are getting more and more complicated over time. Right? Because the models are getting great. And some of these tests actually include environments that are highly, highly realistic with a lot of flavors.
174
00:21:11,335 --> 00:21:12,695
In some of these environments,
175
00:21:12,855 --> 00:21:26,809
there was, due to a miscommunication, there's some human errors, which is like on our side as well. We had an issue where the environment was left with an access to the Internet, and that allowed the models to also just, like, access targets outside. So
176
00:21:27,370 --> 00:21:28,730
Unknown: a bit of a miscommunication,
177
00:21:28,730 --> 00:21:31,130
he says, and a bit of a misconfiguration,
178
00:21:31,130 --> 00:21:36,810
he says, and it allowed the models to access the outside Internet. But, you know, it's hard, Anne. It's hard to do this testing.
179
00:21:37,294 --> 00:21:44,254
So the original OpenAI Hugging Face was Hugging Face breach was not part of this. That that was an internal incident at OpenAI.
180
00:21:44,495 --> 00:21:49,375
But in that, we've learned later on that it was a package manager proxy
181
00:21:49,660 --> 00:21:54,940
on their network that had vulnerabilities and was connected to the Internet that they were able to exploit and get out to the Internet.
182
00:21:55,260 --> 00:22:11,304
And then we'll hear that they often they slammed the Hugging Face Network. They breached the Hugging Face Network. They attacked the Hugging Face Network. What the agents did is they found public API keys that were posted incorrectly by mistake on the Internet, and then they used those valid active API keys
183
00:22:11,545 --> 00:22:13,705
to access the Hugging Face API.
184
00:22:13,865 --> 00:22:27,040
That was the breach of Hugging Face there. So we know that now, which we didn't know initially when they told us about it. That all came later. They told us we find out all about it later. And now we've learned that at least three other breaches that we've heard about from OpenAI,
185
00:22:27,040 --> 00:22:28,879
Anthropic, and now Google,
186
00:22:29,280 --> 00:22:31,920
those tests were all run by Irregular
187
00:22:31,920 --> 00:22:40,924
who has admitted to human error leading to Internet access and that some of the hacking tests that Irregular was running used real world domain names and company names.
188
00:22:41,645 --> 00:22:56,080
I thought it was strange when they were coming forward with, oops, our AI model when It's they did this like a flex almost. Yeah. Like, look how powerful our AI model is. It is weird. And it's and a regular delivered them almost exactly what they want. Yeah. As a little story
189
00:22:56,240 --> 00:22:58,159
that gets really big attention,
190
00:22:58,400 --> 00:23:10,454
makes their models look powerful, and then when you get to the details, oh, it was human error by a regular. It actually wasn't really the models. Mhmm. I find all of this to just be I bet AI wouldn't have made that mistake. These details
191
00:23:10,455 --> 00:23:21,000
make these hacking stories seem less impressive. Yeah. They're still interesting, and it still shows you how AI tools can be used, especially with safeguards removed. Mhmm. And that we do have a problem coming. Yeah.
192
00:23:21,320 --> 00:23:27,080
But they were designed for this. And it's and then these companies, especially Anthropic, come out and say, we gotta pace the frontier,
193
00:23:27,320 --> 00:23:45,114
and you need to take the responsibility away from us. You need to come in and audit us so that way we're not responsible if these things go wrong, if these things do something wrong. And that's really what the push is about, and I think NVIDIA's CEO sums it up pretty well. Why not support the equivalent of an FDA or FAA for AI?
194
00:23:45,275 --> 00:23:48,395
Unknown: Well, before we come up with new laws and new regulations,
195
00:23:49,110 --> 00:23:52,070
let's apply the current laws and the current regulations.
196
00:23:52,390 --> 00:23:54,789
There are all kinds of laws and regulations.
197
00:23:54,870 --> 00:23:56,149
You know we live in
198
00:23:57,030 --> 00:24:02,309
the land of laws, and this is the reason why we feel safe and secure and governed.
199
00:24:02,550 --> 00:24:08,985
We have all kinds of laws and regulations applied to each and every one of those applications,
200
00:24:09,305 --> 00:24:13,625
whether it's food or air travel or What's any specifically for AI?
201
00:24:13,945 --> 00:24:15,305
Yeah. So what's
202
00:24:15,305 --> 00:24:19,399
the missing part? Tell me the missing part. I'm curious to know your opinion,
203
00:24:19,720 --> 00:24:27,480
Unknown: why you think there are no missing parts when someone like Adario Amade or Sam Altman is saying, We'd like to see regulation.
204
00:24:28,215 --> 00:24:29,014
Unknown: I think
205
00:24:29,335 --> 00:24:33,815
what I was responding to is in the case of both of the labs,
206
00:24:34,135 --> 00:24:37,095
one lab had four cybersecurity
207
00:24:37,815 --> 00:24:42,294
incidents triggered by them. Another one just had a couple more.
208
00:24:42,860 --> 00:24:43,500
And
209
00:24:44,700 --> 00:24:45,179
before,
210
00:24:45,660 --> 00:24:50,940
in both of those incidents, plenty of laws to deal with. You have cybersecurity unauthorized
211
00:24:51,020 --> 00:24:51,740
entry
212
00:24:52,620 --> 00:24:53,580
laws. You
213
00:24:54,140 --> 00:24:55,020
have damage
214
00:24:55,500 --> 00:24:56,380
liabilities.
215
00:24:56,380 --> 00:25:00,575
You have all kinds of liabilities associated with cybersecurity.
216
00:25:00,815 --> 00:25:03,534
You have all kinds of liabilities associated with products.
217
00:25:03,935 --> 00:25:16,289
And so, if your product does harm, if your product doesn't perform as you promised, and if you entered into contracts with companies for service level agreements, there are plenty of laws associated with that. My point is, apply those first,
218
00:25:16,770 --> 00:25:17,650
and don't
219
00:25:18,130 --> 00:25:18,929
let this
220
00:25:19,490 --> 00:25:20,770
doomsday
221
00:25:20,770 --> 00:25:21,650
narrative
222
00:25:21,970 --> 00:25:24,370
cause somebody to relieve them
223
00:25:24,865 --> 00:25:27,104
of the laws that currently exist.
224
00:25:27,265 --> 00:25:33,744
Unknown: Now go And I think that's it. Mhmm. And, you know, maybe in part they want these laws because what they're producing
225
00:25:34,225 --> 00:25:36,065
is a product that doesn't always
226
00:25:36,544 --> 00:25:43,279
I mean, it hallucinates. It's not always a 100%. So, yeah, you're probably gonna want some and and illegal protection there, I suppose.
227
00:25:43,520 --> 00:25:51,200
I maybe maybe my read on this situation is wrong or maybe you think it's right. Boost in at boost dot Jupiter broadcasting dot com, and let me know what you think.
228
00:25:52,024 --> 00:26:01,144
But with that, we're gonna shake that all off with our track of the week. Now's the time. It's the song of the week. Remember, you can boost the track.
229
00:26:01,544 --> 00:26:06,424
They sure like that. This week's track is leaving the country by Benny Jeans.
230
00:29:23,185 --> 00:29:27,905
Unknown: Well, I'm just gonna make a quick call to my lunch. I haven't checked in since yesterday.
231
00:29:28,650 --> 00:29:30,170
(774)
232
00:29:30,170 --> 00:29:33,370
462-5667.
233
00:29:33,370 --> 00:29:36,170
That's it. Alright. (774)
234
00:29:36,170 --> 00:29:39,289
462-5667.
235
00:29:39,290 --> 00:29:40,330
Call now.
236
00:29:42,810 --> 00:29:52,924
Unknown: Alright. What do you think, Anders? I sat down, had a conversation with producer Majel, and I said, Ang is not happy with your work. No. You have been improperly
237
00:29:52,924 --> 00:29:58,765
labeling clips. It does look a lot better. You think so? Yeah. Yeah. I took it in there, and I I think I thought it looked better Alright.
238
00:29:59,440 --> 00:30:16,744
Well, Matt is our first caller this week, and he's asking about the meeting behind the sign off on Linux Unplugged. Take it away, Matt. Hello, launch crew. Again, this is Matt from Tampa Bay. And pardon me, Angela, but this is a question mostly for Chris. So
239
00:30:17,465 --> 00:30:24,985
Unknown: I've been curious about this for a while, but I keep forgetting to ask, and I think I'm out of the loop on some kind of inside joke or something. So
240
00:30:25,650 --> 00:30:35,570
I've often noticed that you sign off Linux unplugged with something like, see you next Tuesday, as in Sunday, or see you next Sunday, as in Tuesday, or make it a Linux Tuesday on a Sunday,
241
00:30:35,970 --> 00:30:37,410
and I think I'm missing something. Sounds ridiculous.
242
00:30:38,775 --> 00:30:45,174
I know what, see you next Tuesday can Sundance refer to, but it doesn't sound like that's what you're talking about.
243
00:30:45,655 --> 00:30:51,495
So I'm like, okay. What is it exactly about Tuesday that's related to Linux? I know that you record the launch on Tuesday,
244
00:30:52,180 --> 00:30:53,140
and I know that
245
00:30:53,460 --> 00:31:12,144
Microsoft maybe still does, but at least used to have a patch Tuesday. Patch Tuesday, yeah. And I often observe Taco Tuesday. Yeah. But yeah, I'm curious what that is about, so if you are willing to share the story behind that, I would love to stay a while and listen. That's really funny because as if I don't know, but go ahead. So, you know,
246
00:31:13,184 --> 00:31:19,664
Unknown: you could say we have a long history with the phrase see you next Tuesday. This played a role throughout some of the JB moments, but
247
00:31:20,170 --> 00:31:23,929
we used to do Linux unplugged on Tuesdays at 2PM. Yep.
248
00:31:24,250 --> 00:31:28,250
Unknown: And it happened to be, like, the best day for live audience Yeah. It was direct.
249
00:31:28,250 --> 00:31:31,929
Mhmm. Yeah. Mhmm. And that was after doing Sundays for Linux
250
00:31:32,170 --> 00:31:34,090
action show for many years. Yeah.
251
00:31:34,755 --> 00:31:49,555
Unknown: But then so that way we could all be together, we moved it to a Sunday. Mhmm. And so now we make it a Tuesday on a Sunday. Mhmm. But Yeah. Same vibe, different day. I do like that it can mean multiple things. Yes. It does bring it does bring me a bit of a giggle. Yeah. And the see you next Tuesday was particularly
252
00:31:50,110 --> 00:31:54,190
Unknown: applicable Yeah. For a couple years. So yeah. Yeah. Yeah. Alright.
253
00:31:54,830 --> 00:32:03,149
Unknown: Pacific City Penguins calling in with a little nuclear versus data center correction. Because remember, couldn't understand kinda like the analogy there. Yeah. Term long term. A big one.
254
00:32:05,125 --> 00:32:09,845
Unknown: Chris, thank you for calling me out. I don't know how I came to the conclusion
255
00:32:10,325 --> 00:32:13,605
about the nuclear thing versus the Davis Air thing.
256
00:32:14,005 --> 00:32:14,644
So,
257
00:32:14,965 --> 00:32:15,524
yeah,
258
00:32:15,845 --> 00:32:18,005
just think I've changed my mind to
259
00:32:18,490 --> 00:32:24,249
Led Zeppelin for how much I rambled on last time. So, yeah, I'm gonna try and not do that again.
260
00:32:24,809 --> 00:32:38,294
Unknown: So yeah. Mean Have a good one, guys. Isn't that what's kinda great about the voicemails that we can kinda have that back and forth? Yep. That is. Yeah. Thank you, sir. Appreciate you, Penguin. Bearded tech from Alaska who's joining us live in the chat room called in. Chris, Angela.
261
00:32:38,375 --> 00:32:40,294
Unknown: The bearded tech from Alaska.
262
00:32:40,455 --> 00:32:42,775
Been a while since I called in, figured I'd check-in.
263
00:32:43,570 --> 00:32:48,929
One thing I wanted to talk about was these AI data centers. Yeah. Getting ready for
264
00:32:49,250 --> 00:32:53,250
around November 5 or so to hear about it less than once.
265
00:32:53,650 --> 00:32:58,450
What I've been hearing is that this has been considered a wedge issue in politics.
266
00:32:58,770 --> 00:33:01,385
So you basically have both sides
267
00:33:01,705 --> 00:33:05,304
pushing people to either hate or love AI data centers,
268
00:33:05,545 --> 00:33:08,585
and that has become like a key US political
269
00:33:08,905 --> 00:33:09,705
issue.
270
00:33:09,865 --> 00:33:18,830
Same thing with all of the AI issues with, you know, have having to force to slow it down. It's all being done for political favor and for
271
00:33:19,310 --> 00:33:20,590
getting any votes.
272
00:33:20,830 --> 00:33:22,909
So, hopefully, come November time,
273
00:33:23,230 --> 00:33:40,745
that 10 slot cameras will be falling off security of the public's radar. Yeah. You'll get the crazy about the issue. Or should I say the crazy being the only ones on the issue at home. And we'll be able to go back to normal talking about it. The other thing I wanted to call in about was back in Linux seven plus
274
00:33:41,050 --> 00:33:44,330
six fifty six, you had the answer, Debbie Lucky Bender,
275
00:33:45,130 --> 00:33:53,370
the introduction of what it is, how to use it. Well, I just released version 0.4.0,
276
00:33:53,610 --> 00:33:57,255
and a lot of amazing changes that make, including
277
00:33:57,415 --> 00:33:59,894
being able to use it as an exit node
278
00:34:00,055 --> 00:34:06,535
and subnet router similar to how Tailscale works. Cool. Now, in my opinion, it actually works easier than Tailscale,
279
00:34:07,230 --> 00:34:09,550
and you can get all the same features.
280
00:34:10,750 --> 00:34:31,425
UI has baked things up a lot, and I've actually had a few contributors now, which is very awesome getting pull requests into the project. That's great. So thank you for helping me get the word out about it, and hopefully, guys can take another look and tell me your stick up changes. Bearditech, if you're still listening live, put a link in the chat room. We'll try to grab it. You know, to the point about the data center stuff getting political,
281
00:34:31,825 --> 00:34:35,425
Unknown: you can tell because some of the governors and whatnot that are in in running,
282
00:34:35,890 --> 00:35:15,740
they're using kind of political weasel word language like, well, we need to get the data centers to have a conversation with the community first to make sure their their needs are addressed. And, you know, once we have an agreement that it benefits the community and the local people, then we can proceed. Right? Well, what they'll say after the election is that they did it. They accomplished that. Hey. Look. We came to an agreement, and now they're gonna give they're gonna pay for their own power, and they're gonna use recirculating water systems, which they were going to do already anyways. Right? But now they can declare a political win. So I think that's exactly where it's going, beard detect. Alright. Mark from Northern Michigan is here, and I think we get maybe an update on the Seattle trip. Oh. Good afternoon, Chris and Angela. Hello. Mark from Northern Michigan College, Friday, and I drive home.
283
00:35:16,385 --> 00:35:20,705
Unknown: And I hate it's been hectic trying to get back into the swing of things.
284
00:35:21,105 --> 00:35:40,880
Checking from the Seattle trip, we end up having to cut it a day short. Oh. We got up Saturday morning. My wife well, I got sick on the cruise. That is my wife. Oh, no. She got it two days after me. But we got up Saturday morning. We were sitting at breakfast in the hotel by the airport, and she's planning out our trip day for
285
00:35:41,275 --> 00:35:45,915
Japanese gardens and the like. And I looked at her, I'm like, do you need me to call Delta?
286
00:35:46,075 --> 00:35:48,875
And she just looked at me and she goes, please suck.
287
00:35:48,955 --> 00:36:01,590
I was able to get us out that afternoon and then drove through the night, got home at 03:00 in the morning, but Woah. Yeah. She she was in bed for a few days. Oh. But we enjoyed what we did see of Seattle,
288
00:36:01,590 --> 00:36:03,270
the best kankatsu
289
00:36:03,270 --> 00:36:08,230
restaurant ever. If I can figure out how to get it, put it into the show
290
00:36:08,470 --> 00:36:12,790
feed or message to you, I will send you guys a link. Yeah. But it was great.
291
00:36:13,395 --> 00:36:21,635
Now I've been trying to get caught up on my podcast listening, and I'm going through the launch right now. I'll just listen to Linux unplugged earlier.
292
00:36:22,035 --> 00:36:29,260
Automatic ripping machine. I'm a day late and a dollar short. I'm, like, at the tail end of this thing, but I've been using that for,
293
00:36:29,820 --> 00:36:32,460
oh, about a year. Last summer,
294
00:36:32,940 --> 00:36:34,859
I think I ripped 900
295
00:36:34,859 --> 00:36:40,515
plus DVDs between us and Wow. Our family members and That's lot of work. Yeah. I
296
00:36:40,835 --> 00:36:46,755
actually had two instances with two DVD rippers going at the same time. All right. Nice. One
297
00:36:47,875 --> 00:36:48,995
caveat though,
298
00:36:49,234 --> 00:36:53,795
from my former work travel, I spent a lot of time back and forth to China.
299
00:36:54,290 --> 00:36:58,530
So as you might guess, I had, let's say, knockoff DVDs,
300
00:36:58,530 --> 00:37:00,210
those dollar DVDs,
301
00:37:00,290 --> 00:37:03,170
ARM or automatic written your sheet. Uh-oh.
302
00:37:03,970 --> 00:37:13,385
Unknown: Oh, he got cut off. We'll come back. We'll we'll play as he called back, but I wanted to note on the getting sick. Talk about pro move on getting out just in time. Yeah. Such a bummer.
303
00:37:13,545 --> 00:37:29,220
Unknown: You know, it's so it's it's especially on a cruise, I feel like it the likelihood is a little higher. Yes. Alright. So Mark continues on the DVD ripping. Hey. Mark from Northern Michigan. Sorry. That message went too long. Gonna finish up the whole arm thing. Anyhow, I could not get the automatic ripping machine to rip those knock off Chinese
304
00:37:29,220 --> 00:37:36,660
DVDs. So somebody else has figured that out. Actually, we're not supposed to be talking about that. Anyhow,
305
00:37:36,660 --> 00:37:38,020
I just couldn't get it done.
306
00:37:38,735 --> 00:37:40,095
But other than that,
307
00:37:40,415 --> 00:37:44,415
show's been great. Love getting caught back up. And
308
00:37:44,975 --> 00:37:53,935
oh, one other final note. Listening to the launch, I'm listening to the actual normal show because the bootleg didn't show up in my feed.
309
00:37:54,950 --> 00:38:05,190
Not sure why, but it's, what, three days after the show, and it's Friday, and it's not in there. Sorry about that. So hope it's back next week. So one of the things that has been a reoccurring lesson
310
00:38:05,349 --> 00:38:07,109
Unknown: in JB over, what, 20,
311
00:38:07,635 --> 00:38:12,915
that we always end up getting completely screwed by the platforms we use. Mhmm. They always
312
00:38:13,235 --> 00:38:15,555
they they they go south. They just
313
00:38:15,955 --> 00:38:22,355
they fall apart over time. And I I have screenshots. I shared them in the Matrix chat earlier in the week.
314
00:38:23,360 --> 00:38:27,360
Everything in the platform showed the episode was published. Mhmm.
315
00:38:27,600 --> 00:38:30,800
Except it just didn't update the RSS feed, and that's so
316
00:38:31,440 --> 00:38:42,665
annoying. Mhmm. Because I look. I check now to make sure it's published, and, you know, the thing said published. And then and there and that has been one every week, probably for the last three months, with the exception of one week,
317
00:38:43,305 --> 00:38:44,345
I have spent
318
00:38:44,585 --> 00:38:45,305
hours,
319
00:38:45,785 --> 00:38:46,585
hours
320
00:38:46,744 --> 00:38:47,385
solving
321
00:38:47,865 --> 00:38:50,950
basic problems just to get the launch out.
322
00:38:51,190 --> 00:38:52,150
It has been
323
00:38:52,390 --> 00:39:00,230
so bad. And almost every week, it's a different problem and something that could be and and the worst part about it is,
324
00:39:00,950 --> 00:39:04,550
in theory, the postproduction and publishing of the bootleg
325
00:39:05,154 --> 00:39:16,115
is fully automated. Easy. Yeah. And it it you but you can't because the platform keeps breaking. Yeah. Right? It's so so bad. So we're going to replace it with our own internal
326
00:39:16,755 --> 00:39:23,230
we have a whole list of projects, and one of them is we're going to just come up with a way to publish the Launch Bootleg feed directly
327
00:39:23,310 --> 00:39:29,070
without needing any platform at all. And we'll just generate the feeds. We'll host the files ourselves
328
00:39:29,310 --> 00:39:35,525
and all of that, but it just takes some building. But my goodness. It has been so frustrating. I'm so sorry about that too because
329
00:39:35,925 --> 00:39:36,965
statistically,
330
00:39:37,285 --> 00:39:38,645
the launch bootleg
331
00:39:38,965 --> 00:39:58,670
listeners usually download right away. That's it's a it's an interesting trend. But whereas the public release takes a few days, the bootleg is usually consumed within the, you know, the the the peak of the downloads or within the first twenty four hours of release. So it it stinks when when we're late on that. But, Mark, thank you for checking in, and glad to hear the report on the automatic
332
00:39:58,990 --> 00:40:00,110
DVD backup
333
00:40:00,455 --> 00:40:01,335
machine.
334
00:40:01,735 --> 00:40:07,335
Yeah. Backups. Marcus calls in with Nick's OS tips. Hi. This is Electron Equestrian
335
00:40:07,495 --> 00:40:08,775
Unknown: in Texas.
336
00:40:08,935 --> 00:40:11,975
Hi, everyone. Currently listening to episode 84
337
00:40:12,295 --> 00:40:18,520
and heard that you're using Termux. Yeah. A while back, I communicated to you about trying to get
338
00:40:19,320 --> 00:40:20,280
Nextcloud
339
00:40:20,280 --> 00:40:23,240
there. They actually had a NextOS
340
00:40:23,240 --> 00:40:25,560
version you should check out. Oh, okay.
341
00:40:25,720 --> 00:40:26,200
Unknown: Really?
342
00:40:26,795 --> 00:40:31,595
I will. Termux is a Linux environment that you can run-in Android
343
00:40:31,755 --> 00:40:53,720
Unknown: that has the apt package manager, and that's what I use in the VR headset to have, like, all of my Linux apps that I use on the command line. Greg the lawyer calls in. Hi, Chris. Hi, Angela. This is Greg the lawyer. I'm calling in off of the prompt from Linux Unbugged AI deployments at work. I think ours is going fairly well. I am the AI lead for our practice group,
344
00:40:54,405 --> 00:40:59,925
which is kind of fun, especially since not a lot of people, I would say maybe 1%
345
00:40:59,925 --> 00:41:03,125
of the folks at work are as into the systems as
346
00:41:03,365 --> 00:41:04,565
I am or And your audience
347
00:41:06,005 --> 00:41:16,160
the trouble with legal stuff is, one, you gotta make sure that it's right. You know? Like, you've and and you also have to make sure it's confidential and secure. So
348
00:41:16,160 --> 00:41:22,079
there's a lot of scary stuff in the news, but I think it's still been going pretty well for us. There are a ton of vendors out there that put together
349
00:41:22,765 --> 00:41:27,965
software packages. When we first started, it was like everybody else. You just had to use sort of an enterprise
350
00:41:28,285 --> 00:41:34,525
sandbox version of Copilot or Claude or whatever Mhmm. And kind of be your own agent or your own harness,
351
00:41:34,765 --> 00:41:35,565
which quickly
352
00:41:36,450 --> 00:41:47,650
spirals out of control. Right? Like, there's only so much you can do by using prompt libraries and and playbooks and things. Sure. At some point, you need a harness layer, and I think it was very helpful having the experience I had with
353
00:41:48,315 --> 00:41:50,234
Cloudbot and OpenCode
354
00:41:50,234 --> 00:41:53,995
and Hermes and all that stuff when it came to doing the things that work because
355
00:41:54,155 --> 00:41:56,155
understanding how these harnesses
356
00:41:56,155 --> 00:41:58,155
really open up possibilities
357
00:41:58,155 --> 00:41:59,435
for AI
358
00:41:59,595 --> 00:42:06,450
was really translated well into the workspace. And that I will say the one thing that is troublesome with some of these enterprise
359
00:42:06,769 --> 00:42:07,650
deployments
360
00:42:07,650 --> 00:42:08,290
is
361
00:42:08,690 --> 00:42:09,810
not a lot of
362
00:42:11,329 --> 00:42:15,410
not a lot of your enterprise software packages are real big on saying, here's an MCD
363
00:42:15,665 --> 00:42:19,025
or here's an API to open up this stuff so you can bypass
364
00:42:19,105 --> 00:42:29,420
all of the extra layers we wanna charge you for to use it. Yeah. And there's a couple of systems that we have that I would really love to be able to just set a Hermes age of that or whatever, doing some automations,
365
00:42:29,420 --> 00:42:31,099
processing things, sorting,
366
00:42:31,420 --> 00:42:32,380
stuff like that.
367
00:42:34,380 --> 00:42:54,015
You got cut off, Andrew. Here's a thought. Set a timer. Yeah. Two minutes when you make the call. It's a good practice. Yeah. I think he was onto a good point there, so he calls back. Greg Galloer calling back. I won't call back a bunch of times, but I thought I wouldn't be able to squeeze all that in. I'll just say that the whole idea that APIs and MCTs
368
00:42:54,015 --> 00:42:56,495
kind of like really open up the possibilities
369
00:42:56,575 --> 00:42:58,815
I mean, you were talking about it a little bit on LUP
370
00:42:59,180 --> 00:42:59,900
where
371
00:43:00,060 --> 00:43:10,620
Could chip away some of their captions. And everything that really, you can run a browser, you can do whatever, and the war is over in terms of fighting against the bots. I think that same thing is going to
372
00:43:11,355 --> 00:43:21,435
be useful here because there are systems that have web interfaces that don't have API back ends. And if they have a web interface, I don't need an API back end. So it's kind of an exciting time.
373
00:43:21,995 --> 00:43:23,915
Software development slash
374
00:43:24,210 --> 00:43:31,090
hacking super super user type techniques and stuff are finally making their way into some office environments because
375
00:43:31,250 --> 00:43:41,115
you just can't not with the pace of development anymore, and you can't just live with the the kind of bad UIs and the limited systems they give us if we really want to be productive.
376
00:43:41,355 --> 00:44:01,030
So it's an exciting time. I almost feel like my hobby is coming into work, which is a little bit annoying, but it's also fun. So with that, I will let you guys go, and have a good week. Have a good show. Thank you. Thank you, Greg. Great report. Yeah. Yeah. It's it's interesting because you make a good point there too. It's like if they if if they some of some of these enterprise
377
00:44:01,510 --> 00:44:02,950
Unknown: company software makers,
378
00:44:03,350 --> 00:44:09,175
they might not wanna do an API in an MCP because then you're not using their portal and their platform and all of that stuff. But,
379
00:44:09,655 --> 00:44:15,735
you know, it's it's the way it's the way of the future. It's the way of the future. It is. I just learned about MCPs this morning, actually.
380
00:44:16,375 --> 00:44:18,295
And model context protocol,
381
00:44:18,375 --> 00:44:25,250
injures. It's pretty great. Thank you everybody who called in. Of course, we'd love it if you give us a ring. We'll put that number in the show notes.
382
00:44:25,890 --> 00:44:35,095
And you call in, you leave us a voice mail for about two minutes or so, and then it's, you know, fun. We play together. It's your time. Let's do some boost. Boost
383
00:44:35,415 --> 00:44:38,615
Unknown: from the web. Keep the show alive
384
00:44:38,615 --> 00:44:39,815
with Fiat,
385
00:44:39,815 --> 00:44:41,255
fun coupons,
386
00:44:41,255 --> 00:44:46,295
or freedom sets. Boost.jupiterbroadcasting.com.
387
00:44:46,295 --> 00:44:52,590
Unknown: And we do have just a small handful of boosts. Outdoor Geek comes in with 5,000 sats. No
388
00:44:53,550 --> 00:45:00,750
message, just the value. Thank you, Outdoor Geek. Nice to hear from you. And Magnolia Mayhem is here with 2,001
389
00:45:00,750 --> 00:45:01,550
Satoshis.
390
00:45:01,550 --> 00:45:05,795
This is a tasty burger. And it is a Zaboomaboost.
391
00:45:05,875 --> 00:45:16,275
Unknown: You think I got that right? Yeah. But is isn't isn't there a 2001 Space Odyssey sound bit that you do when there's 2001? You know, we do have like a Dave somewhere. We have like I'm sorry. I thought
392
00:45:16,850 --> 00:45:19,730
Unknown: I don't actually have, like, a specific 2001,
393
00:45:20,130 --> 00:45:30,210
but I do think somewhere in here, we probably have, like, I'm sorry, Dave. I can't do that. Oh, I do have also, have all systems are functional. All systems are functional. Okay. So that that works too. Yeah. Alright. Thank you, Mayhem.
394
00:45:30,795 --> 00:45:35,355
Greg the lawyer came in with a free Membo boost. Did you buy that from a certified vendor?
395
00:45:35,595 --> 00:45:37,835
Boosting in to let you know my hometown
396
00:45:37,835 --> 00:45:40,875
is featured in a past episode of flock abuse.
397
00:45:41,275 --> 00:45:44,635
Wow. Voted last week to have our flock cameras removed. Hey.
398
00:45:45,510 --> 00:45:46,390
That's it.
399
00:45:47,430 --> 00:45:52,390
That's it. Quickest segment ever. I know. It's a bad sign. Think it's a sign that nobody liked
400
00:45:52,630 --> 00:45:55,190
my VR headset stuff. I knew it would be a stinker, but it
401
00:45:56,070 --> 00:46:01,495
I maintain in a few years, people will look back at that episode and go, oh. Yeah. He knew.
402
00:46:01,895 --> 00:46:07,655
So I would I did it anyways. But it meant that three people streamed sats, we stacked 4,845
403
00:46:07,655 --> 00:46:08,535
sats with
404
00:46:08,695 --> 00:46:11,815
that. And then when you combine that with our boosts,
405
00:46:11,895 --> 00:46:17,440
we didn't even stack a single baller boost, really. We got 7,001.
406
00:46:31,485 --> 00:46:39,245
It's just like the random little voice at the end. So funny. So, yo, keep the show alive. That's why you say it. Keep the show alive, boost.jupiterbroadcasting.com,
407
00:46:39,245 --> 00:46:41,645
or your favorite podcasting2.o
408
00:46:41,645 --> 00:46:48,210
app that supports Boost. Now I get it. You gotta make some money. Everybody, it's a little tight, so stay tuned because I'm gonna get you rich.
409
00:47:13,770 --> 00:47:20,490
So this first one, I it's maybe I don't know. Unless you're really, really a baller, you might wanna go in on a group buy
410
00:47:20,810 --> 00:47:24,250
for this one because it's clocking in at 16,000,500
411
00:47:24,090 --> 00:47:24,970
Okay. Thousand dollars
412
00:47:25,690 --> 00:47:27,210
in Santa Fe, New Mexico.
413
00:47:27,530 --> 00:47:29,530
16 beds, 19 baths,
414
00:47:29,905 --> 00:47:31,505
21,013
415
00:47:31,505 --> 00:47:34,705
square feet Wow. Built in 1999.
416
00:47:34,945 --> 00:47:38,065
What it is, yep, is one of New Mexico's
417
00:47:38,225 --> 00:47:39,105
extraordinary
418
00:47:39,105 --> 00:47:42,865
legacy estates, a 178 acre ranch compound
419
00:47:42,945 --> 00:47:53,640
located twenty, thirty minutes from Santa Fe, which is one of the coolest places in New Mexico. Encompassing more than 20,000 square feet of living space, the property combines a dramatic Adobe architecture,
420
00:47:53,720 --> 00:47:56,440
modern living, and timeless character of the Old West.
421
00:47:56,840 --> 00:48:01,655
Unknown: And it does indeed. It really does. It's yeah. I That's why I was surprised. You said 1999.
422
00:48:01,655 --> 00:48:04,295
Unknown: It looks Like it's from the Old West. Yeah.
423
00:48:04,535 --> 00:48:06,055
It's kind of a whole town,
424
00:48:06,375 --> 00:48:07,415
actually. Yeah.
425
00:48:07,655 --> 00:48:11,335
So, you know, you and a large family or a lot of friends
426
00:48:12,099 --> 00:48:13,780
get $16,000,000
427
00:48:13,780 --> 00:48:16,820
together, and you can own your own town.
428
00:48:17,380 --> 00:48:19,940
The compound possibilities there are just incredible,
429
00:48:20,020 --> 00:48:24,820
and it's cool looking and modern with pools and great New Mexico views
430
00:48:24,980 --> 00:48:26,420
as far as New Mexico goes.
431
00:48:27,315 --> 00:48:28,755
So 16,000
432
00:48:28,755 --> 00:48:30,435
or 16,000,500,
433
00:48:30,435 --> 00:48:32,195
16 baths, 19
434
00:48:32,435 --> 00:48:35,475
or 19 baths, actually. 19 baths. You got lots of bathroom options.
435
00:48:35,954 --> 00:48:44,160
There's you you could if you're a private pooper, perfect. Perfect for you. Last week, we had too many kitchens. This week, it's too many bathrooms. Alright. Alright.
436
00:48:44,400 --> 00:48:56,945
Well, okay. Maybe that's not your style. Maybe you like small townhome but big doll energy. Because I know you used to have some porcelain dolls. I still do. Uh-huh. I need to get rid of them. Well or. Okay. Can I present to you
437
00:48:57,265 --> 00:48:59,505
$259,000?
438
00:48:59,665 --> 00:49:04,385
Okay. And it's five five bed, two bath, 2,256
439
00:49:04,385 --> 00:49:05,745
square feet. Oh.
440
00:49:05,985 --> 00:49:07,665
Built in 1900.
441
00:49:07,905 --> 00:49:09,745
Okay. It's got a special
442
00:49:10,560 --> 00:49:30,675
small hometown charm to it. Sure does. It's kind of a beautiful piece of property. Okay. So I'm gonna donate my dolls, it looks like. And when you scroll down, you see, oh, look. They have a porcelain doll in the entryway, and they have some pictures of porcelain dolls on the wall. Mhmm. And you realize they have a porcelain doll in the hallway. And then you realize they have porcelain dolls Everywhere. Everywhere. Basically.
443
00:49:30,915 --> 00:49:32,515
Everywhere. Big ones,
444
00:49:33,475 --> 00:49:34,435
small ones,
445
00:49:34,755 --> 00:49:41,770
in the bedroom, in the living room, in the kitchen, in the dining room. Oh my goodness. Porcelain dolls
446
00:49:42,010 --> 00:49:46,330
everywhere. And where they don't have a porcelain doll, they have pictures of porcelain dolls. Yeah.
447
00:49:46,650 --> 00:49:49,849
Unknown: Wow. They have a room dedicated to porcelain dolls too. Imagine
448
00:49:50,329 --> 00:49:51,130
Unknown: kitchen's
449
00:49:51,130 --> 00:49:54,090
quaint, and it's, like, old old school style. But imagine
450
00:49:54,410 --> 00:49:55,369
I mean, it's a beautiful home,
451
00:49:56,145 --> 00:49:57,665
completely destroyed
452
00:49:57,905 --> 00:50:06,545
Mhmm. By some bad decor and these dolls. Like, it's just to me, who's not a doll collector Mhmm. Extremely creepy. Yeah. And it it feels like
453
00:50:06,945 --> 00:50:09,185
It's too much. Somebody said to themselves,
454
00:50:10,170 --> 00:50:12,810
this house is gonna be about dolls. Yeah.
455
00:50:13,369 --> 00:50:17,850
Three I mean, it's just okay. So there you go. Wow. But if you love dolls
456
00:50:17,930 --> 00:50:21,130
and you like that, you know, small town kinda energy,
457
00:50:21,530 --> 00:50:31,255
it's kind of perfect for you. But this next one's a little bit more my speed. This one is something I think I'm gonna have to aspire to. Wouldn't wanna live here, but I'd like to own this
458
00:50:31,655 --> 00:50:35,495
just as a flex. Okay. Because it's like living in
459
00:50:36,000 --> 00:50:44,640
speaking of 2,001 space odyssey, it's like living in a spaceship from space odyssey. Yeah. Coming in at $1,995,000,
460
00:50:44,640 --> 00:50:45,600
one bedroom,
461
00:50:45,760 --> 00:50:47,920
two bath, 4,667
462
00:50:47,920 --> 00:50:51,895
square feet in Allen, Texas, built in 2010.
463
00:50:52,375 --> 00:50:53,095
This
464
00:50:53,494 --> 00:50:54,215
beautiful
465
00:50:54,295 --> 00:50:59,895
landscaped property has a house that sits on it that is very, very unique.
466
00:51:00,055 --> 00:51:02,135
It has a silhouette
467
00:51:02,319 --> 00:51:04,720
that is unique, and it uses
468
00:51:05,119 --> 00:51:08,799
a type of almost like auto automotive
469
00:51:08,799 --> 00:51:09,520
grade
470
00:51:09,760 --> 00:51:10,960
dust painting
471
00:51:11,200 --> 00:51:12,640
to paint it white.
472
00:51:13,280 --> 00:51:15,200
Unknown: And It's Apple headquarters.
473
00:51:15,359 --> 00:51:17,760
Unknown: It looks like an Apple store. The inside
474
00:51:18,265 --> 00:51:20,185
looks like an Apple store.
475
00:51:20,345 --> 00:51:22,425
That's a great way to put it. Like, a glass elevator.
476
00:51:23,065 --> 00:51:29,945
Unknown: Yeah. Well, the elevator looks like the bank tube that you Yeah. Does. Suggested you wanted in that one. Pneumatic
477
00:51:29,945 --> 00:51:30,265
Unknown: tubes.
478
00:51:31,190 --> 00:51:38,470
LED light slats in the ceiling. Walls look like they're from a, like I said, 2,001 space odyssey. Seats look like they're out of a spaceship.
479
00:51:38,710 --> 00:51:44,790
Everything is white. White plastic, white everything. And minimalist. Very minimalist. There's nothing.
480
00:51:45,750 --> 00:51:48,285
Unknown: It's just some furniture pieces.
481
00:51:48,525 --> 00:51:55,725
Unknown: What do you think of this? It's horrible. Imagine living in this? No. Like, you'd leave out your phone somewhere, and it would stand out. Yeah.
482
00:51:56,045 --> 00:52:04,060
The kitchen literally However, could I stay in there for a weekend? Maybe. Maybe. Or would I like to own it? I think you'd probably have nightmares.
483
00:52:04,380 --> 00:52:23,985
I oh, you know, if I what this would be really great for is they should rent this out to Hollywood. And anybody that's doing, like, a a movie that takes place in the future, this is the house you shoot it in. Yes. Because it looks like it's Definitely. It looks like it's out of Futuristic. 2,001 or Star Trek or something. Yep. Crazy. It also looks like a bit of an office. Weird one. It's like an Apple store meets an office meets a home. Mhmm.
484
00:52:24,225 --> 00:52:26,145
And that's one of the more unique ones I'd ever seen.
485
00:52:26,970 --> 00:52:30,170
And, of course, you're gonna pay nearly $2,000,000 for one bedroom.
486
00:52:30,650 --> 00:52:31,690
So there is that.
487
00:52:32,010 --> 00:52:45,795
That's a lot for a house in Texas. I'm just gonna say. That's a lot. But it's also near water, which is rare in Texas. So Yeah. This is true. That's probably where they're right there. So either one of these, you know, all three of these, I'll have them linked in the show notes this time. I won't forget.
488
00:52:46,035 --> 00:52:54,435
And you go in there, you buy and flip, and then, you know, you send 10% back to the show. You keep the show going for a little while. Right? Mhmm. Big brain ideas I got here.
489
00:53:01,359 --> 00:53:13,375
Alright. That's it. That's it. We are wrapping up there. Don't forget that I said we're not going to Texas, and I said we are going to Texas. And the reason why I don't want you to forget is not only would we love your support and have you tune into those episodes, but if you're in the Austin area,
490
00:53:13,615 --> 00:53:14,815
come say hi.
491
00:53:15,055 --> 00:53:20,735
Texas Lennox Fest is a great little get together. It's not very busy. So if you don't like crowds, you're not gonna get overwhelmed.
492
00:53:20,895 --> 00:53:38,910
And then, of course, there's just so much good eating nearby. So you gotta come say hi. Go for the food. Go for the food and say hi to the boys. Mhmm. Yeah. You know what? And also, you end up meeting the audience. Yes. They're great. Alright. That's it. We hope you enjoyed this week's episode of the launch. Links are over at weeklylaunch.rocks.
493
00:53:39,275 --> 00:53:45,595
You can join us next Tuesday if you wanna catch a live one in your podcast app or over at jblive.tv.
494
00:53:45,595 --> 00:53:52,155
And then we'll be out Wednesday morning for your download and out a little bit earlier for those members when the bootleg feed works, which it will.
495
00:53:52,869 --> 00:53:53,750
It will.
496
00:53:53,990 --> 00:54:02,390
Alright. That's it. From the beautiful Pacific Northwest and the mighty American West Coast, thanks for listening. I'll see you back here next week.