What do you call an endpoint security product that works perfectly but makes users miserable? A failure? The old approach to endpoint security is to lock down employee devices and rollout changes through forced restarts, but it just doesn't work. It is miserable because they got a mountain of support tickets. Employees start using personal devices just to get their work done, and executives up out the very first time. It makes
them late for a meeting. You can't have a successful security implementation unless you work with end users, and that's where collide comes in. Their user first device trust solution notifies users as soon as it detects an issue on their device, and teaches them how to solve it without needing help from it. That way, untrusted devices are blocked from authenticating, but users don't stay blocked. Collide is designed for companies with Okta, and it works on Mac OS, windows,
Linux and mobile devices. So if you have Okta and you're looking for a device trust solution that respects your team, visit callide.com/unsupervised learning to watch a demo and see how it works. That's callide.com collide.com/unsupervised learning. Welcome to Unsupervised Learning, a security, AI and meaning focused podcast that looks at how best to thrive as humans in a post AI world. It combines original ideas, analysis, and mental models to bring not just the news, but why it matters and how
to respond. All right. Welcome to unsupervised learning. This is Daniel Missler. I have been traveling like a crazy person, but have got to get this podcast out. Don't like skipping weeks, so let's jump in. So it has been a crazy week. Started for the last episode. Bit overexposed, been doing lots of travel, interacting with humans and stuff like that. Got some minor con flu or whatever it is, but I didn't actually get it. I didn't get it bad.
It's just slightly congested. Can't tell if it's allergies or whatever, but everything's fine. Got a new pattern here. Analyze presentation. Say a new fabric pattern that looks at the transcript of a presentation and tells you how much the presenter is like trying to brag versus how much they're trying to entertain versus how much they're actually trying to inform. So it tells you essentially how self versus other centered that they are. I spoke at equilibrium conference last week.
It was fantastic. Got to meet Chris Hughes for the first time in person and appreciate the invite from Rohit. It was a great event, just absolutely loved it and got to attend a live UFC event for the first time. It was so great. It was a UFC 300. It was supposed to be like the best UFC ever. You know how they always say that, but it turned out it actually was. It was unbelievable. Especially the major fight
between Cage and Holloway. It was beyond all expectations. It was just so great going to be speaking at Hardly Strictly Security soon, speaking at the Security Frontiers conference that are actually already happened, and also this one already happened. This was yesterday and this was super fun. So thanks to Jane Chief for the invite there and got to meet and hang out with song for quite a while.
That was super fun to talk with her. She's super bright and awesome and yeah, gearing up for RSA and just finish reading A Sense of Style by Steven Pinker for the third time. Highly recommend you put that in your rotation for every few years. It's my favorite writing
and style guide, so we talked about this one. Analyze presentation didn't really do any other essays or videos this week just because all the travel and stuff, but they are accumulating and I'm going to have a lot more content coming out soon, and it'll be both an essay form and YouTube. I'm actually transitioning really heavy towards YouTube, so go subscribe to the channel if you're not subscribed yet. It is. Unsupervised learning is the name of the channel
on YouTube. All right. Security. Microsoft says Chinese hackers are using AI to inflame social tensions in the US. And every time I hear about these, I can't wait to build my propaganda tracker, because if somebody starts a campaign, I want to see where it spreads, where it becomes viral. And we have it's like Dan Kaminsky used to say, we have the technology. I don't know if that's from
Bionic Man. Think it might be from Bionic Man. But anyway, yeah, we have the technology to do this type of stuff, especially once we start incorporating AI agents. So I'm really looking forward to this. Yeah. And so the thing about it is we need a combination of government and private for this. Right. Because we need the thing to be trusted. But we also need it to be dependable and actually incentivized to work. And of course it needs funding. But you got to come up with some kind of cool
name prop track or something. Propaganda tracking. Okay, over 92,000 D-Link NAS devices exposed nasty to IP has three high severity vulnerabilities us just blacklisted for additional Chinese companies for procuring AI chips for China's AI modernization efforts. And US and China are racing to dominate with AI driven drone swarms. Once again, you gotta check out this book by Daniel Suarez. Kill decision. Really entertaining and also very prescient at the
same time. Google's contract reveals a partnership with Israel's Defense Ministry, and this caused actually like a walkout or a protest at Google. And I think it was 23 people. They all got fired, which was pretty interesting. Technology. Great New York Times piece on automating finance jobs using AI, and
got a cool quote here. Some of Wall Street's major banks are asking the same question as they test AI tools that can largely replace their armies of analysts by performing in seconds, the work that now takes hours or a whole weekend. That is a big jump. Seconds versus hours or a whole weekend. Like like we've been saying, it's like at this point, if you don't see the impact of AI and jobs, it's because you don't read enough or you don't want to see. I mean, this
is willful ignorance at this point. It isn't like the.com boom. This is what everyone's saying. oh.com crypto. We got a lot of people saying things like this and it's like, no, this is tech that's instantly useful. And when Wall Street is talking about this in very concrete terms of like, yeah, we're getting rid of lots of people because seconds versus hours or a whole weekend. I mean, I don't know how you get more compelling and realistic than that, yo.
He continues to do great work around AI agents. He's got one that does AI agent log visualization. Let's open this one up. So basically you click through here and yeah look at these. You click through and you can see exactly what the agent did and get like logs from it. Really cool. Project. Russia's getting 89% of its chips through China as a way of bypassing sanctions. Gary Marcus is betting 10 million that we won't have smarter AI,
smarter than humans by 2025. Yeah, the whole trick there is like how they measure, right, what KPI they're actually using. But I think this person's, uh, yeah, this is going to come down to definition competition, I guarantee you. But that is a bad bet. I think Mr. Ole just released a eight times 22 billion model via torrent, which is pretty cool. It's a really big files and distributed and yeah, and the more people downloading it, the faster it gets. So this is all the reason that transformers
or no torrents were made in the first place. So cool. This paper says transformers capable of acting like a general purpose computer. OpenAI is shipped new GPT four turbo. Although I have tested it, and in my most difficult AI jobs that I have, which are within fabric like detect hidden message, for example, that is a extremely difficult human cognitive function, like advanced logic is really needed there. And opus still crushes, uh, GPT four in that particular one
and others that are similar. And after I beat them, professional go players didn't just catch up, they became better and more creative. I love this thing where it's like we learned and we got better, but I don't want people to think that it means we'll be able to keep up. It really doesn't, but we still can find things interesting and useful, even if it's just human to human. And we're nowhere near as good as human versus computer or computer versus computer. That's fine. Right? It's not. One
does not have to mean the other. So look at chess. Chess is more popular than ever. And we got crushed by computers in 97. And we're not getting much better. Nanotech Onyx is rolling out Cube Fab, a moderate, a modular chip fab powered by Nvidia to democratize semiconductor manufacturing. Human pin got flamed by Mkbhd, and I think he could say whatever he wants. I just think he was a little harsh on this thing. And to me, it's
not like I want to control what he's saying. I was just a little bit let down by the vibe. I tend to think of him like as this nice, balanced person. So when I see a headline that's just like worst product ever, it just disappointed me at a personality level, not at an analysis level. Nothing about his analysis was wrong, and he actually was somewhat balanced in
the quite balanced inside of the actual review. My thing is like when you put that spin of like, worst ever, it just automatically has like this negative valence to it. And it turned me off. As a fan of his reviews, but I don't think he was wrong, and I don't think anyone should be telling him what he can or cannot do. Dyson's new Clean Trace feature lets you use AR to see all the spots you missed. Okay, but
what do you have to wear? Do you have to wear a resume, an Oculus, or an AVP Sequoia's arc product? This one I haven't messed with yet introduces three unique archetypes to help startups navigate their way to market success. I'm going to keep this tab open and you need to go play with that more. Tesla just dropped the FSD subscription from I think it was like 250 or 300, something like that, and now it's 100 bucks a month. Cool. So I hope mine that I was already paying for
just goes down to 99 naturally. And I don't have to do something like cancel it and renew. All right, humans, right after the solar eclipse, a whole bunch of people were googling my eyes hurt, which, uh, further diminishes my hope for humanity. It's not so much that they looked at the sun. It's the two hit combo of them looking at the sun and then wondering why they couldn't see. And one of the books that will absolutely change your
mind on like, so much stuff is Everybody Lies. It talks about the difference between what people say is true or what people say they believe or like versus what they actually Google for. And it's just like it reveals so much interesting stuff. It's just a fantastic book. Job interviews are pretty terrible at predicting job performance, but what's
the alternative? That's my question, is, like, I see a lot of things that say it doesn't do a good job, but if you don't have a better proxy, then so what? Why are we talking about it? Over half of women experience sexual harassment at work. Homicides are on a surprising decline across major cities. I feel like we're this close and very close to an onion article where far right people are. Like, we used to have good murder rates, and now the whole country is like feminized. And the
murder rates keep. Going down. And that's just further evidence of America's decline. Not trying to make fun of the right. They're just, uh. I thought that was funny. I recently re subscribed to onion, so I'm hearing onion jokes everywhere I look. Now, U.S. steel is about to become part of Japan's Nippon Steel. I think that's how you pronounce that. And Harvard is back to SAT and Act scores. It's almost like they were a good idea in the first place, which is why we had them in the first place.
And my comment here, I wish the Overton Pendulum had more than two settings, which is like woke and racist. Like, can we get a middle ground? Can we get a middle setting? Compassion, logical, something like secular humanism or something, or even not secular humanism, just logical humanism getting a lot more open to some moderate religion? I think it's
got a lot of positive functions. In fact, the research by Jonathan Haidt on this is pretty compelling, basically showing that kids of religious parents are doing so much better in mental health. And yeah, I'm trying to track all these threads and find a way to unify all this. Dumb phones are making comeback. You basically get to stay in contact but not be distracted. It's going to be a Dungeons and Dragons show at Madison Square Garden, which
is supposedly going to sell out. That does not surprise me. I think this whole thing, which, by the way, tonight is a gaming night and I have to leave soon, so I guess I'm going to speed up. Yeah, I played D&;D with my friends. It's fantastic. I'm so glad to see D&;D rising up. Actually, it doesn't really matter if it's exactly D&;D just getting together, doing some sort of intellectually, creatively challenging role playing game I think is super valuable. What top performers do and how to deal
with their critics and detractors. Ernest Hemingway transform from a humble journalist to a celebrity, losing himself to the fame and power that came with success. Abraham Lincoln was shaped by Aesop's Fables more than any other book, including the Bible. That was a really interesting read and for my idea and analysis this week, I heard somewhere on a podcast it might have been Alex Hormones. It might have been like, no, it was so many on Chris Williams and maybe the
gamer doctor or yeah, maybe. I'm not sure exactly who it was, but they didn't exactly say it this way. But it was essentially like, if you have an idea that's really exciting to you, you basically don't want to share that or you don't want to get any wins from the idea that aren't wins from actually doing the work. And so I'm expanding that out here to basically say energy you have for the idea is limited and precious, and even telling someone about it, I think they did say,
actually say this. I think the guests might have said this. Even telling somebody about it removes some of its power. So instead of telling people about it, you should work on it immediately. Use that power to go right into the actual work. Otherwise, you might end up not doing anything because your your body, like evolution, thinks that you did the work because you talked about it. It's like it's a fake win and you don't want to do that. This one I'm probably going to do as a standalone essay.
Reason to worry, reason to build. So I'm going to skip that one. And we got a colab notebook here that uses cloud to turn images into magic cards. Karpathy launched LMK shell history. This thing is really cool. Adder turns your terminal into a pair programming session with AI, DNS viz cohere Rerank three boosts enterprise search and drag systems. Terminal tweaks UDL turns your text prompts into music. This thing is insane. Collection of content ideas that have resonated
on TikTok. Pretty cool list for creators. Fire crawls Wdev turns entire websites into data sets for LMS. Zeek turns your command line into a powerful plaintext Zettelkasten or personal wiki keeper. Kind of an alternative to some of the other ones like obsidian. And yeah, recommendation of the week is actually to read that piece, which I'm going to do a standalone episode on, and the aphorism of the week being deeply loved by someone gives you strength, but
loving someone deeply gives you courage. Being deeply loved by someone gives you strength, while loving someone deeply gives you courage. Lao Tzu Unsupervised Learning is produced and edited by Daniel Missler on a 1987 AI microphone using Hindenburg. Intro and outro. Music is by zombie with a Y, and to get the text and links from this episode, sign up for the newsletter version of the show at Daniel missler.com/newsletter. We'll see you next time.