Martian Human Vats

[JM]: I came across a fun site called "Are You in the Weights?"

[JM]: And the purpose of this site is to type in names of people, whether it's yours or someone else's, and see, A, whether or not you can find that person in various large language models.

[JM]: And two, what do you see about that person?

[JM]: What is the strength of

[JM]: number associated with it, which I'm not sure I fully understand.

[JM]: They don't really define what the strength thing means.

[DJ]: So wait a minute, Justin, is this a website about like lifting weights or something else?

[DJ]: Like I can find out how strong you are?

[JM]: definitely not about lifting weights and i suppose if i bothered to tap on the about button then i'd already know the answer to the thing that i said i didn't understand which is your strength score is an average of how strongly each model recognized you plus a bonus for how many models recognized you so this is kind of the 2026 everyman version of do you have a wikipedia article about you

[JM]: I suppose so.

[JM]: But these days, given how Wikipedia editors at some point decided that they were going to be real strict about notability and really raise the bar for who qualifies as notable, these days, I think we can safely say that the vast majority of the human population is not in Wikipedia, but might publish random stuff on a website somewhere that could get

[JM]: sucked up into one or more of these various large language models. And presumably given that they have this thing on the main page that says today's heavyweights where it shows a bunch of names and then the strength score underneath them, clearly the people that made the site didn't populate it with a whole bunch of names. They didn't automatically feed it a bunch of names and then generate these things. I think it's people that are typing in these names.

[JM]: And they're showing you a leaderboard of the various people and what their strength scores were.

[JM]: And it's kind of interesting to see some of the random people that people type into this thing.

[JM]: But in any case, I digress from what is clearly the most important part here, which is, am I in the weights?

[JM]: And of course, you know, I'm going to ask you next.

[JM]: Actually, I'm going to ask you first.

[JM]: So Dan, are you in the weights?

[DJ]: I don't know, because I saw a link to this site the other day and thought, I don't care about this.

[DJ]: I don't want to ruin your pre-show bit, Justin.

[DJ]: But I saw this thing and thought, I spend every other waking moment of my life and every conversation I've ever had with anyone and everything I've ever read on the Internet for the last couple of years is about large language models.

[DJ]: And I saw this one and I was just like, yeah, maybe.

[DJ]: But I just, I can't.

[DJ]: I don't have the...

[JM]: Well, then you just made this game so much more fun, because now I'm going to tell you if you are in the weights.

[JM]: And the answer is, I also don't know, because...

[DJ]: I mean, I have a pretty common name, so I'm not sure.

[DJ]: If you just type my name into this thing, there's lots of people with that name.

[JM]: The top result for your name is South African Novelist and Short Story Writer.

[DJ]: That's right.

[DJ]: That's why I used a different name when I published my novel, because I did not want to compete with that person.

[DJ]: I'm serious.

[JM]: Oh, that's an actual person.

[DJ]: There's a very notable author who has my name, who is South African.

[DJ]: I think from the 20th century, he's not still alive, but that is a person.

[DJ]: So I was not surprised that that's the top hit.

[DJ]: That's basically also the top hit if you type my name into a search engine.

[JM]: That makes sense.

[DJ]: There's also a TV producer from the 1990s with my name.

[DJ]: So when I was watching sitcoms as a kid, I'd often be like, hey, what am I doing up there?

[JM]: Well, other candidates in the weights with your name are Debian developer and free software activist, American political operative, computational biologist at Oak Ridge, and software engineer and Node.js contributor.

[DJ]: Yeah, none of those sound like me.

[DJ]: I mean, I'm a software engineer, but I've never contributed to Node.js, at least not that I'm aware of.

[JM]: And obviously it is tough when names are more common than others.

[JM]: I learned just by sheer nature of setting up a Google search for myself once with my name because I wanted to know if people were referencing an article that I published or not.

[JM]: talking about something that I put into the world, I thought it would be cool to know about it because oftentimes there's not really other good ways to know about it.

[JM]: And what I learned is there are a lot of people with my name, many more people than I ever even fathomed.

[JM]: And sometimes they do really bad stuff and make headlines.

[JM]: And that's how I end up hearing about other people with my name.

[JM]: Oh, no.

[DJ]: Okay.

[DJ]: I'm imagining.

[DJ]: I'm imagining some really bad stuff now.

[JM]: So I fully expected going into this thinking, okay, well, yes, I would be surprised on some level if I weren't in the weights, but didn't really have strong expectations in terms of being notable enough to be listed among the results.

[JM]: But it turns out the very first result says Pelican Static Site Generator Creator.

[JM]: And

[JM]: I showed up in the majority of the models that this site says it queries, including Claude, ChatGPT, Kimi K2, and Qwen3.

[JM]: And Claude Opus 4.8 says, "Software developer and open source contributor known as maintainer of the Pelican static site generator written in Python".

[DJ]: That's pretty good, I suppose.

[DJ]: That's pretty good, except that it doesn't mention that you're also an amazing podcaster.

[JM]: That's right.

[JM]: I think the source data, the training data, needs to be updated.

[DJ]: Yeah, come on.

[DJ]: We're like over 50 episodes into this thing.

[DJ]: Give us some love, Claude Opus 4.8.

[JM]: Yeah, Claude.

[DJ]: Maybe 4.9.

[DJ]: 4.9 will do it.

[DJ]: It'll be called "The Abstractions Update".

[DJ]: We realized we were missing something very important.

[JM]: Absolutely.

[JM]: And then ChatGPT 5.5 says that I am a "software engineer and open source contributor associated with Python and Pelican related projects".

[JM]: So it's a little weird, you know, to see your life encapsulated in this way.

[DJ]: Especially because our friend, well, we don't know him, but we mention him a lot.

[DJ]: Simon Willison is also a developer who's associated with Python and Pelican-related projects because he's always posting pictures of pelicans that he got LLMs to generate.

[JM]: That's true.

[DJ]: So really, you could say that about a lot of people, Python and Pelicans.

[DJ]: They go hand in hand, which is weird because neither Pythons nor Pelicans have hands.

[JM]: That is weird.

[JM]: Sometimes they have hands in the images that get generated when Simon hands this query to a large language model and says, you know, draw me a Pelican riding a bicycle.

[JM]: Yeah, sometimes it has hands.

[DJ]: Right.

[JM]: But yeah, it's really weird to have your life encapsulated in five words as a tagline and then a one-sentence summary of who you are.

[JM]: It's just something about it is deeply weird.

[JM]: So this top result that is indeed me...

[JM]: Shows a strength of 340.

[JM]: The other people underneath this, the strength is in the 60s.

[JM]: And there's only two other people listed.

[JM]: But none of these doppelgangers clearly have like a material presence in the weights.

[JM]: Because the two that were listed only appeared in a single model.

[JM]: Whereas I appeared in the majority of the models that they queried.

[JM]: And as far as I can tell, both of these that are listed are confabulated non-existent people.

[JM]: Just based on who it says they are, I couldn't find any information about what it is they are reported to actually do or be associated with.

[JM]: So they don't seem to exist, which is an interesting result.

[DJ]: I guess.

[DJ]: I'm not entirely sure I like this implication that just because you don't maintain a bunch of open source repositories, you don't exist.

[DJ]: But okay, Justin, there's a certain amount of online chauvinism present there.

[DJ]: And I guess in this whole exercise, right?

[DJ]: Because the only things these models know about are basically things that are on the internet for the most part, because that's what they were trained on, in addition to books and all kinds of stuff.

[DJ]: But this is a funny reflection of like the age that we live in because it feels very similar to what we were doing probably 20 years ago when Google was a new thing, or even when like search engines at all were a new thing would be like typing people's names in and being like, is this person on the internet?

[DJ]: You know, cause that was new and novel.

[DJ]: And now it's, was this person who is on the Internet slurped into the training data of a large language model is essentially what we're asking.

[JM]: Yeah, one should really not infer any sense of validation by being included in the weights on this site or in any large language model period.

[JM]: Because as you said, the way that this works is it's just sending a bunch of bot scrapers out and collecting information that is not anywhere near comprehensive in terms of what actually goes on in the world.

[JM]: So I'm sure of all the other Justin Mayers, there are probably some that are much more interesting and would have much more interesting taglines than, "Dude Who Maintains the Pelican Static Site Generator". Which... I hope that is not the only thing I am known for when I shuffle off this mortal coil, but it was at least an entertaining five minutes.

[JM]: That's fair.

[DJ]: I'm thinking of a couple of funny things.

[DJ]: One of them is the notion that like there's another Justin Mayer out there who's like an international jewel thief.

[DJ]: He has like multiple identities and disguises and he's always on yachts and, or, you know, whatever rooftop chases with local law enforcement agencies.

[DJ]: And like, he's way too busy to be on the Internet, right?.

[DJ]: So he's living a much more interesting life.

[DJ]: But then on the other hand,

[DJ]: Something I was reminded of is there used to be this website called Twitter.

[DJ]: I don't really remember what happened to it, but people would have these little bios and the bios often followed a formula because, you know, people were sheep basically.

[DJ]: So in the same way that like, if you looked at, look at people's LinkedIn profiles, they all have, have a little tagline and the tagline all looks the same, which just as a point of advice to people trying to get hired on LinkedIn, you're not doing yourself any favors by looking like every other person.

[DJ]: Just saying.

[DJ]: Sign up for my online course.

[DJ]: It's only $9.99.

[DJ]: But back on this web site called Twitter, people would have these bios.

[DJ]: And I just remember these various trends showing up in them.

[DJ]: And one of them would be because you don't have a lot of word count, people's bios would say something like "Husband. Christian. Father, Open source developer."

[DJ]: or something like that and now I'm just imagining what if these models had been trained on those. And so then you would look someone up in the weights and it would always have these weird like stilted descriptions. So it's like Justin Mayer and it's just like "Open Source Maintainer" you don't have any of the other qualifiers because that's all you're known as it's not like you know it's not like, "Friend. Mixologist."

[JM]: "Mechanical Keyboard Connoisseur."

[DJ]: That should actually be what you're most notable for.

[DJ]: "Passive Speakers Evangelist".

[JM]: That's a good one.

[JM]: I like the 8-bit style that these folks used to do this site.

[JM]: It's very 1980s, like not even 20 years ago.

[JM]: This is like very 8-bit.

[JM]: We aren't going to use any anti-aliasing.

[JM]: And it has some cool sound effects when you type the name as well as when you hit the delete key to backspace.

[JM]: And those are different sounds, which is a nice touch.

[JM]: So extra points for style.

[DJ]: That is pretty cool.

[DJ]: And that reminds me, although I don't have a link to this right now, I came across something the other day.

[DJ]: Someone had built a piece of software that shows you the status of baseball games, like live baseball that's going on right now.

[DJ]: But it's all rendered in that sort of like 1980s Commodore 64 or like Nintendo Entertainment System 8-bit pixelated system.

[DJ]: So it has like a picture of a baseball diamond with tiny little players running around.

[DJ]: And like, it's a dead ringer for like a baseball game that came out in 1988, but it's showing you the stats of the game that you're watching.

[DJ]: And I didn't dive too deep.

[DJ]: I would love to know how this person built this thing, but it was one of those things I came across on the internet and it filled me with joy.

[DJ]: So yeah.

[DJ]: There is something to be said for these retro, these retro 8-bit styles.

[DJ]: I don't know why, but I mean, probably my age is why, but like they really pluck at the heartstrings.

[JM]: People that create things like the baseball scoreboard that you mentioned are, they're so incredibly creative.

[JM]: Like they belong in the way.

[JM]: It's certainly more than I do.

[DJ]: Get in the weights, 8-bit creator guy.

[DJ]: Unless, I mean, unless you don't want to.

[DJ]: It feels like it should be optional to be in the weights, which I...

[DJ]: Doesn't matter whether you want to.

[DJ]: I know.

[DJ]: I feel like this is actually an argument that has been going on for a long time amongst many people, really, is like, it should be optional to be in the weights.

[DJ]: It's not, though.

[JM]: No, it is definitely not.

[JM]: There is no opt-out button.

[DJ]: And speaking of things I wish I could opt out from: SpaceX.

[JM]: It would be nice if we could all opt out of SpaceX entering our 401(k)s, but...

[DJ]: But you can't.

[JM]: It doesn't seem like that's going to be so easy to do.

[DJ]: Nope.

[DJ]: Not if you're an index investor, it's not.

[JM]: Just for the sheer entertainment value, I wanted to take another look and see what was going on when we started recording.

[JM]: And what's going on is that SpaceX has, after achieving a stock price of over $222 a share, has gone straight off a cliff and has dropped to $140-something.

[JM]: Currently now at $150-something, which is for those keeping score at home is below its IPO price. So if you bought it in the mad dash right after the IPO, you probably took a bath, at least if you sold it today. And there's definitely a small amount of schadenfreude here I guess but...

[DJ]: Are you sure it's small?

[JM]: Okay, maybe not so small.

[JM]: But also it still has a market cap north of $2 trillion.

[JM]: And the world's first trillionaire is presumably still a trillionaire.

[JM]: So that hasn't really changed.

[JM]: But it will be interesting to see it continue to move.

[JM]: And in which direction.

[DJ]: It will.

[DJ]: I mean, I'm not an investment professional, but I think if you bought shares in this IPO and you sold them now, I'm not sure what your strategy was because if the point was to sell them, like to flip them essentially, you should have done that when they were worth $200, right?

[DJ]: But I guess if someone was sitting on those going like, oh, no, no, this is going to 300 next week.

[DJ]: Well, you pays your money and you takes your chances.

[DJ]: But I imagine for a lot of investors, no.

[DJ]: I don't know how many investors, but I imagine for some investors at least, the point isn't that it popped after its IPO and then came back down.

[DJ]: The point is that 10 years from now, their shares will be worth a whole jillion dollars and they'll be able to retire on a private island.

[DJ]: The joke's on them because the world's first trillionaire will have purchased up all of the islands.

[DJ]: and launch them into space to form a colony on Mars.

[DJ]: Don't ask me how that's going to work.

[DJ]: I mean, I'm not in charge of those details.

[DJ]: I'm not worth a trillion dollars.

[JM]: Apparently he does have every intention of putting stuff on a rocket and hurling it at Mars because one of the things I've learned recently is that the CEO and founder has been granted by SpaceX's board an additional $1 billion of...

[JM]: restricted shares of Class B common stock on one condition.

[JM]: He has to hit 15 market capitalization milestones up to $7.5 trillion and, and this is the best part, establish a permanent human colony on Mars with at least 1 million inhabitants.

[JM]: Now, he's already got 5 billion shares.

[JM]: He doesn't really need the extra billion.

[JM]: But I do want to at least comment on what I think the likelihood of him achieving this particular milestone is.

[JM]: Many people have looked at the state of things and have issued predictions with near certainty.

[JM]: We're well aware that the road is littered with things like 640 kilobytes should be enough memory for anybody.

[JM]: or the world market for computers is like 10.

[JM]: But I'm going to go out on a limb and say, he's not hitting this milestone.

[JM]: There is, in my opinion, 0% chance, not like 1%, 0% chance that he gets a million people to permanently live on Mars in his lifetime.

[JM]: That's just not happening.

[DJ]: Well, it's interesting that you say in his lifetime, because I was curious, like, I didn't hear any dates associated with that.

[DJ]: Because it was like, okay, you got to get the market cap up to seven and a half trillion, you got to settle a million people on Mars.

[DJ]: But like, in how long?

[JM]: Well, it's his compensation.

[DJ]: Right.

[DJ]: But I mean, I presume he doesn't really need that compensation.

[JM]: He definitely doesn't need the compensation.

[JM]: But the timeline is as long as he is still CEO of this entity, presumably.

[JM]: I don't think there's like an expiration date.

[JM]: As long as he's still CEO, that's what he gets if he accomplishes this condition.

[DJ]: I see.

[JM]: Which, again, is never going to happen in his lifetime.

[DJ]: Well, but again, you say lifetime, but I'm pretty sure the singularity is near, right?

[DJ]: So human beings are about to transcend mortality and their physical bodies.

[DJ]: And then you've got infinity years to settle a million people on Mars.

[DJ]: I'm not sure what the point is once people have transcended their physical bodies of moving them to Mars.

[DJ]: And it actually raises some philosophical questions.

[DJ]: Like if we've digitized ourselves and we live in machines now, where can you be said to live?

[DJ]: Like are you on Mars or are you on Earth if what you really are is just data flowing through the universe?

[DJ]: I'm sure my skeptical tone is going to look pretty silly in like 10 years when we're all just data flowing through the universe and or living on a colony on Mars.

[DJ]: But I stand by it.

[JM]: Well, if you want a sneak peek into what life will be like when this vision that you describe becomes reality, watch a animated program called Pantheon, which I checked out recently.

[JM]: And there are two seasons of it.

[JM]: And it explores this very question of uploaded humans.

[JM]: And thanks to my friend Andrea for recommending this show to me because I have really enjoyed it.

[JM]: It is both interesting philosophically and as a show.

[JM]: So check it out.

[DJ]: Nice.

[DJ]: Well, as long as we're recommending animated things, I will recommend the classic anime Ghost in the Shell, which came out in 95, 96, something like that.

[DJ]: And I've been meaning to give it a rewatch.

[DJ]: And it also deals with the philosophical questions of like, what is a person?

[DJ]: Like, are you your mind?

[DJ]: Are you something else?

[DJ]: Is like, is a synthetic person a person? Et cetera.

[DJ]: Like can software be conscious? It's all very interesting and timely, and it's also a totally rad action mystery thriller.

[JM]: Well, there you go everyone. There are there are your animated show recommendations for the week.

[DJ]: I mean, we're really not qualified to tell you whether or not you should buy SpaceX, so we should probably just stick to to recommending animation.

[JM]: Absolutely. This is not financial advice. Standard disclaimers apply.

[DJ]: A million people on Mars, though.

[DJ]: See, I think that's the one that doesn't make any sense to me.

[DJ]: It's one thing to say, like, we're going to start a colony on Mars.

[DJ]: And I guess I understand that if you're like, no, I'm going to put people on Mars, then you want to put a lot of people there, because otherwise, what was the point?

[DJ]: But a million people, like, are you going to take a million people from Earth and move them to Mars?

[DJ]: Because that seems to have some serious practical difficulties.

[DJ]: Like, no one has ever moved a million people anywhere, right?.

[JM]: All of it involves wildly impractical difficulties.

[DJ]: Well, I know -- this is what I'm saying.

[DJ]: Or do you like put a small colony on Mars and then have it grow, which is fine, but still presumably it's going to take quite a while to grow to a million people.

[JM]: You're just going to grow people in vats.

[DJ]: That's it.

[DJ]: That's all you're putting out there.

[DJ]: You're just like, "people in vats".

[DJ]: I have two words for you.

[DJ]: People. Vats.

[JM]: It's going to be a vast human growing pod farm where we're all grown in these pods or vats, you know, whichever you prefer.

[JM]: Yeah, that's how it's going to be done.

[DJ]: Okay.

[DJ]: It doesn't say that in the prospectus, right?

[DJ]: Like it doesn't require people to be grown in vats in order for him to get his extra billion shares.

[DJ]: So the vat growing is strictly optional.

[DJ]: I'm just imagining the marketing for this.

[DJ]: It's like, "Join the Mars colony -- now with optional vats."

[DJ]: Optional vats to grow people in.

[JM]: "Want a family but don't want all the work that goes along with it?"

[JM]: "Don't worry, we got you covered."

[JM]: Okay, moving on to other actual news that doesn't involve growing humans in vats.

[DJ]: Is this our weirdest episode yet, Justin?

[DJ]: It might be.

[JM]: I think it is.

[JM]: I think it qualifies.

[JM]: I think it must be the summer heat.

[DJ]: Probably, yeah.

[DJ]: And that's a good segue into Steam.

[JM]: Speaking of heat, the Steam Machine, which we talked about in a previous episode because it was announced months ago –

[JM]: is actually going to launch. They have opened a site where you can go and place your request, and the reason why it's a request is because Valve does not expect to be able to fulfill all the orders, so it's essentially going to be a lottery system.

[JM]: So if you are listening to this before June 25th at 10 a.m. Pacific time, then you can go to the site that's in the show notes and you can submit your request.

[JM]: And after that time, they're going to close the signups.

[JM]: They'll do a one-time randomization to determine if

[JM]: the reservation and a waitlist order.

[JM]: And then they'll reach out if you were chosen or placed on the waitlist.

[JM]: And one of the things that is mentioned in this announcement is the price. Because when they originally announced the steam machine, there was no official price mentioned and everyone just had to guess, and reasonable people with decent knowledge of how these things work estimated it would be maybe somewhere in the $750 to $800 range.

[JM]: In terms of the starting price in their announcement, Valve said that unfortunately, they had to price it higher than they would have preferred.

[JM]: So for $1,050, you get a six-core, 16-gigabyte RAM, 512-gigabyte SSD without a controller. The controller is $80 more.

[DJ]: Sorry, that was my, that's my opinion about buying a machine dedicated to playing video games and you don't even get the controller with it.

[DJ]: Like, come on.

[JM]: Presumably they're doing that in case you already have adequate controllers and you don't want one.

[DJ]: I guess it's fine, but still.

[DJ]: It feels nickel-and-dimey, I'm just saying.

[DJ]: It's like having to pay for your human vat extra when you've already bought a ticket to the Mars colony.

[JM]: And for $300 more, you get a upgrade from the 512-gigabyte SSD to a two-terabyte SSD.

[JM]: And the GPU in this machine is roughly on par with an Nvidia RTX 3060 or an AMD RX 6600.

[JM]: And the general consensus, and I would concur with that consensus, is, yeah, this is really pricey for what you're getting.

[JM]: Obviously, a lot of that is not their fault, and...

[JM]: I think the announcement is written in a way that really tries to have empathy for the people on the other side, people who were and perhaps still are excited about this machine.

[JM]: They go to length to explain that, hey, this was not our preferred price.

[JM]: We didn't want to have a lottery system, but.

[JM]: Not only are the prices high for these components, but it was hard for us to even get them and to get enough of them.

[JM]: And even just looking at the solid state drive upgrade, I mean, $300 to go from 512 GB to two terabytes.

[JM]: Yikes.

[JM]: It was not that long ago...

[JM]: It was like maybe six or so months ago that I bought a four terabyte solid state drive for like $250.

[JM]: It is a sobering situation, right?

[JM]: Where Valve is super excited to make this machine.

[JM]: People are super excited to buy it.

[JM]: And this whole component crunch has really taken some of the wind out of the sails of this launch.

[JM]: I just really feel for the company and for all the people that were excited about buying one.

[DJ]: Yeah, we've commented on this situation a lot because we're both computer enthusiasts, and we think about buying individual components.

[DJ]: Like you said, you picked up an SSD, and I've thought about buying another storage disk for my Unraid server that I put together last year, and it's just gotten economically infeasible to do that.

[DJ]: It's just not worth it.

[DJ]: I don't need some extra terabytes of storage

[DJ]: redundant disk storage badly enough to pay like $600 for them or whatever.

[DJ]: We can see this affecting projects that are bigger than just individual hobbyist computing as well, right?

[DJ]: Like Valve is not a small company.

[DJ]: They're small compared to the companies that are buying up all of the manufacturing capacity of computer components so that they can put them in the data centers that they're trying to build.

[DJ]: But you can really see how this is affecting this entire industry, the entire industry of putting computers together and selling them.

[DJ]: Valve's affected by it.

[DJ]: Apple just announced that they're going to be increasing the prices of their products at some point in the near future.

[DJ]: I don't like it.

[DJ]: I'm hoping that we find our way out of this trough in the computer component industry eventually.

[DJ]: I want the days of $100 RAM to return.

[DJ]: I'm looking forward to a day when we can actually purchase those components and companies like Valve can actually purchase them too and deliver them a product that's compelling.

[DJ]: Because I think you're right.

[DJ]: Whatever the selling proposition of the Steam Machine was, it can't be as expensive as a high-powered general-purpose computer.

[DJ]: Otherwise, what's the point of buying it other than you can pay $80 more to also get a gaming controller?

[JM]: The Steam Machine *is* a general purpose computer, and they go out of their way to say, "Install whatever OS you want on it and use it however you want to use it".

[JM]: It's a general purpose computer that just happens to be optimized for Steam OS, use with a controller, and playing games via Steam's platform.

[JM]: But I agree with you that it is not an easy proposition to try to sell something at this price point when even after the price of, say, a PlayStation 5 Pro has increased, this is still considerably more expensive without providing considerably more value for your average gamer.

[JM]: But I hope that people nonetheless end up buying these machines because I love that Valve conceived of this product, that they're trying to get it into people's hands.

[JM]: And I want them to continue making machines like this.

[JM]: Okay, in other news, we've mentioned Llama.cpp a few times.

[JM]: And I just wanted to point out that that project and its companion app that you can install on your desktop machine and have a little menu item in your menu bar...

[JM]: to get a more user friendly non terminal user interface way of managing your local models.

[JM]: These projects are now collectively known as "Llama".

[JM]: And you can go to llama.app to download the new desktop application.

[JM]: And I've been pretty happy with it.

[JM]: I like how it dynamically loads and unloads local models, you can tell it after five minutes, unload it and free up that memory.

[JM]: You can choose 15 minutes, an hour, or just leave it persistent in memory forever.

[JM]: And there are some other nice bells and whistles.

[JM]: So if you are into running local large language models, I highly recommend it.

[DJ]: Nice.

[DJ]: I think I'm going to give that a try because I've been using LM Studio for this, but I've been wanting to try Llama.cpp and get a little closer to the metal, so to speak.

[DJ]: And I'm perfectly willing to use a terminal-based application, but it's nice to also have the option to have more of a

[DJ]: a richer desktop app experience without having to go to a whole separate project that has its own aims and its own development path and things like that.

[DJ]: So this is nice to see.

[DJ]: Out of curiosity, because I haven't checked, do you know, is the Llama desktop app only for Mac or does it work on Linux also?

[JM]: I'm not sure.

[JM]: I know that when it started, it started out as a Mac application.

[JM]: I don't know if they have since produced a version that runs on Linux.

[JM]: Obviously, the command line tool, llama.cpp, obviously that runs on Linux.

[JM]: But I think the desktop application, the thing that gives you a little menu item in your menu bar, I think for the moment that is Mac only.

[DJ]: Gotcha.

[DJ]: Well, that's okay.

[DJ]: I'll just have to get comfy using the terminal.

[JM]: If you check out the Unsloth guides on HuggingFace, they're usually really good about giving you some detailed instructions on what parameters to use and what terminal knobs and buttons to use.

[JM]: So check that out.

[DJ]: That sounds good.

[DJ]: Although I will say that maybe one of the things I like the least about the era of large language models that we now live in is that it's forced us to say ridiculous words like "Unsloth" and "HuggingFace".

[DJ]: Because seriously, what?

[DJ]: Like, if you're going to become the most popular hub for open source large language models, could you at least have the decency to have a less silly name?

[JM]: They are indeed very silly names.

[JM]: And with the namespace getting more and more polluted by the day, I think if there's one thing you can count on, expect more very silly names.

[DJ]: Hold on.

[DJ]: Registering MartianHumanVats.com.

[DJ]: Gotta snap that up before someone else takes it.

[JM]: All right, moving on.

[JM]: In other news, Microsoft has announced a generative software personal assistant that they are calling Scout.

[JM]: And some internal document got leaked called ClawPilot Overview and Plan with Project Lobster.

[JM]: And I have thoughts already.

[DJ]: You're just making these names up to drive me insane now, right?

[DJ]: Like, come on.

[DJ]: Come on.

[DJ]: It's not really.

[JM]: I told you to expect more silly names.

[DJ]: The open claw dude, I can't remember his name, has a lot to answer for.

[JM]: He really does.

[JM]: So this document, with a subheading called "ClawPilot Overall Plan", notes three phases to its launch plan.

[JM]: And the first phase is: "Make people addicted."

[JM]: And that's the part that's been making headlines, and not in a good way.

[JM]: 404 Media's reporting on this topic says, "Microsoft has been piloting Scout as an internal tool for employees...

[JM]: it was calling ClawPilot since March.

[JM]: ClawPilot and now Scout are part of Project Lobster, which is a Microsoft plan to bring the popular OpenClaw AI tool to its Microsoft 360 suite of products...

[JM]: in a way that non-technical people can use.

[JM]: It is not particularly notable that Microsoft is developing new AI tools.

[JM]: The company has reoriented almost everything it does to focus on AI, and every major AI company has tried to figure out how to bring AI agents into their products after OpenClaw went viral earlier this year.

[JM]: OpenClaw allows users to create AI agents that can act on behalf of the person using it.

[JM]: It can send email, edit calendars, publish blog posts, and more.

[JM]: What is notable is that the explicit goal of the people developing the product is to addict its users."

[JM]: So the folks who authored this article interviewed Microsoft employees to ask them about it.

[JM]: And one employee said that the addiction language was very troubling and went on to say...

[JM]: "We're seeing more and more addiction happening with AI chatbots and agents and overall addiction to me is something no product should be making a part of its build strategy."

[JM]: "It feels like one of those 'saying the quiet part out loud' moments in the document."

[JM]: And I have to agree.

[JM]: They really nailed it.

[JM]: There is definitely "saying the quiet part out loud" when you say that the first phase of the plan is to make people addicted.

[JM]: But another employee said, "Isn't the end goal of all software made by all major technology companies to be addicting?"

[JM]: "Luckily for us, Microsoft is pretty bad at making addicting products compared to some of the other big companies."

[DJ]: Ouch.

[DJ]: Sorry, was the part you just said out loud a quote from a Microsoft employee or was that the writer of the article saying that?

[DJ]: No, that was the Microsoft employee.

[DJ]: Really?

[DJ]: Wow.

[DJ]: Okay.

[DJ]: So the part where they go, the people on record did not want to be identified discussing private matters or whatever.

[DJ]: In this case, it's like the people did not want to go on record publicly dissing their own employer, even though the diss is 100% valid.

[JM]: Including the takedown of "We suck at making addicting products."

[DJ]: No, exactly.

[DJ]: It's like, good.

[DJ]: I'm glad you know that the company like you're not deluded and thinking that people love using Microsoft software because it's awesome because that isn't true.

[DJ]: I saw this headline and I rolled my eyes a little bit because I thought, well, yeah, I'm sure someone internally was like carelessly used the word addicted to mean we want people to really be excited about this and use it a lot.

[DJ]: I'm giving that person the benefit of the doubt that they don't actually want to create a population of people who wake up in the morning and their hand won't stop shaking until they use this stupid Microsoft product because that's ridiculous.

[DJ]: But on the other hand, I think there is a good point in there about...

[DJ]: Really, we don't, I think we'd all be better off if software companies would pay attention to things other than like, we got to get people using our products.

[DJ]: How do we get them to use it?

[DJ]: How do we get them to use it more, more, more, more, more, more, more usage, more analytics, more like, how do we get them to tap this button and do that and then do this and do chill out dudes.

[DJ]: You know, I feel like that that is something a lot of us are feeling.

[DJ]: Could software companies just chill the heck out and like and just cut it out with this whole like, oh, we got to get people to use more stuff, more, more, more.

[DJ]: It's like I kind of want less, Justin.

[DJ]: I found myself attracted to all this retro technology lately. Like refurbishing my iPod Mini and plugging it into an amplifier and passive speakers instead of streaming music, right? I don't need to do that, but I find myself drawn to these things, I think out of this -- to I think steal the title of someone else's book -- this "Longing for Less".

[DJ]: And so one of the things that comes across in this internal Microsoft document, whether it's a sinister story of intentionally trying to addict people, is just this undercurrent of like, we need more stuff, we need more stuff, and we need more people to do more things.

[DJ]: And I would like to see a growing tide in the other direction.

[DJ]: I would like to see people going instead of using more software to more thoroughly automate the more things that I'm always doing.

[DJ]: What if we turn towards like, well, maybe we should just do less stuff like write less email, you know, write less code.

[DJ]: Generate less, etc, etc.

[DJ]: What about that?

[DJ]: I think there's value in contemplating this when all that we hear from the industry is always around more.

[DJ]: How do we burn more tokens?

[DJ]: How do we build more data centers so we can train bigger models so they can write more emails for you, etc.

[DJ]: I always come back to this question of like, what is all of this stuff in service of?

[DJ]: Other than, you know, making people trillionaires or certain people into trillionaires.

[DJ]: I mean, I get it.

[DJ]: So back to the email minds, people.

[JM]: Obviously, a company has to make a product sufficiently interesting in order to sell it.

[JM]: And any product that you pick up once and use and then put on a shelf and never touch again is probably not going to be...

[JM]: a successful product.

[JM]: So making products that people want to use is indeed a point generally speaking of making products.

[JM]: But to me, the idea that the end goal of all software made by all major technology companies is to be addicting is kind of a gross thing to say.

[DJ]: Yeah.

[DJ]: Is evil on its face essentially wrong and broken?

[JM]: Because even if you give this the most charitable definition of addicting instead of the more literal interpretation, I still think that nobody should be designing things that are intended to create so much engagement that it takes them away from the many other things that should be in a life.

[JM]: And I'm not here to tell people what to do with their time, but...

[JM]: I think we can all generally agree that just as an example, if you wake up in the morning and you go to a casino and you gamble all day long until you can't stand up straight and you just crash and you fall asleep and you do the same thing the next day and every day after that, that doesn't seem good.

[JM]: And maybe that's a value judgment, but I don't think that we should be even using this word to describe what we're trying to get people to do, like what we're trying to get our products to do.

[JM]: So no, the end goal of all software made by all major technology companies should not be to make them addicting.

[DJ]: No, the end goal of software made by software companies should be to elevate the lives of the human beings who use the software, to your point, because a life should consist of many different things, not just one thing that you do over and over and over and over again.

[DJ]: And to the point about like going to the casino, let's say, until you can't stand up or whatever, it's less about a value judgment is that you can observe the many negative second order effects of doing that, right?

[DJ]: So it's not so much like, hey, don't moralize at me.

[DJ]: It's like, well, I'm not moralizing at you.

[DJ]: You are physically and mentally unwell in the following like quantitative ways.

[DJ]: So the thing you're doing is bad for you.

[DJ]: It's not like you're a – I'm not saying you're a bad person for doing this.

[DJ]: Like I'm not saying you're a bad person for scrolling through a social network app on your phone all day long.

[DJ]: I'm saying it's bad for you.

[DJ]: You shouldn't do that probably.

[DJ]: You'll be better off and happier if you do something else.

[DJ]: And by that same token, the people who make software, I understand they have very powerful incentives to try to just keep their stock price going up at all costs.

[DJ]: And that shows up in the design of their products.

[DJ]: But no, really, if you make a software product, the goal should be to improve the well-being of the people using it.

[DJ]: I think I've complained about this before.

[DJ]: We've touched on things like that one writer's concept of needy software.

[DJ]: I've mentioned the way I perceive the problem is that the people who develop software, I mean, they spend a lot of time thinking about it, right?

[DJ]: They wake up all...

[DJ]: every day and they're in the shower and they think of like, oh, when I go to work today, the thing I'm going to do with that app that I work on is X, Y, Z.

[DJ]: It's really important to them.

[DJ]: And they conflate that with its importance to their users.

[DJ]: And the fact is, no matter how much I love your app, your app only takes up like 1% of my entire life.

[DJ]: So it's misguided, if you're a software developer, to think, how do we try to get it so that people spend more time using our app?

[DJ]: No, that's not the point.

[DJ]: The point should be, whatever is the useful amount of time for someone to use your app, is that time well spent?

[DJ]: So instead of thinking, the first thing we need to do is make our app addictive?

[DJ]: No.

[DJ]: The first thing and really kind of the only thing, the last thing as well that you need to do is like make your app really good at the thing that it's for and then trust that people will use it because they enjoy using it.

[DJ]: When you reverse that and you focus all your attention on, but how do we get the pigeon to tap the lever more often?

[DJ]: Then you go to some really dark places, which is what we're all seeing.

[DJ]: Like, this is what we're all responding to in this article.

[DJ]: It's just like, right, yeah, no, we already know that big technology companies are totally misguided in their approach to building software.

[DJ]: So it's really sucky to see that, as you say, the quiet part being said out loud.

[DJ]: It's like, hey, Microsoft's building a new product, and we're doing it in the exact same horrible way as every other technology company does.

[DJ]: does to ruin this corner of your life a little bit more.

[DJ]: So that's why I'm a big fan of Microsoft and all their products.

[DJ]: And so I'm here to tell you people, buy Microsoft.

[DJ]: No, this is why we all hate you, Microsoft.

[DJ]: And we always have.

[DJ]: And we always will.

[JM]: As a fun little footnote to this story, Microsoft CEO had previously said at a conference that he loves OpenClaw, but that Microsoft could never integrate it into Microsoft products, saying, I can't launch OpenClaw at Microsoft.

[JM]: I mean, it, you know, it just wouldn't work.

[JM]: I don't have permission to do that because that would be considered Microsoft launching a virus.

[JM]: I mean, that's just not a thing.

[JM]: End quote.

[DJ]: Right.

[DJ]: No, they can't.

[DJ]: They can't allow a virus into Microsoft, but they can build their own virus and then inflict it on everyone else.

[DJ]: That's fine.

[JM]: Fast forward to the present day and we have ClawPilot.

[JM]: So it looks like he can launch OpenClaw at Microsoft.

[DJ]: Yeah.

[DJ]: He, he for one welcomes his new lobster overlords.

[DJ]: Should we all, it would be funny if next episode, I mean, the listeners can't see this, but you and I are on video right now.

[DJ]: And so it'd be funny if next episode is...

[DJ]: You're like, so Dan, should we talk about how crappy OpenClaw is and Microsoft's new version of it is bad?

[DJ]: And I'm like, what do you mean, Justin?

[DJ]: OpenClaw is good.

[DJ]: And then I raise up my hand and it's been replaced by a giant lobster claw.

[DJ]: So anyway, if you want to fund my purchase of a big foam lobster claw so I can totally weird out my co-host, just send Bitcoin to the following wallet.

[DJ]: A6F481...

[DJ]: All right, I'll stop now.

[JM]: All right, everyone, that's all for this episode.

[JM]: Thanks a lot for listening.

[JM]: You can find me on the web at justinmayer.com and you can find Dan... in the weights.

[JM]: Reach out with your thoughts about this episode via the Fediverse at justin.ramble.space.