IM 859: What's Behind the Fox? - Tech's Gilded Age

31:40So it's pretty clear that, you know, all the AIs that are out there right now are trained on just pretty much everything they get their hands on, including most of the internet, Wikipedia, most of the books pirated. But I have to think the Stack Exchange was probably one of the best places for AIs to start when it comes to coding. Do you resent that? Not really. I resent it if it hauls out the community. But my feeling is there's always new languages, new situations. Now granted, for selling Java, there's always new releases of Java, new things. And the end result is to help people. We always said the best outcome on Stack Overflow.

32:26So you have a question, you type it into Google, and you see the answer immediately from Stack Over. It's a top-voted answer. Now you can drill in and see some other good answers, but the goal is to make it easier for people to be better programmers. Right. So I don't want to be willing to say I don't care. I want it to be sustainable, almost like a workers' cooperative would be kind of my take at this point. I want to go back to the question Leo asked at the beginning about too much code. riff on that I mean have you seen some of the stuff that LMS will do when you tell them to optimize it's like optimize this for 95 % it's like okay return true that's a good optimization because it doesn't know what it's doing it has no actual understanding it's playing a game of global blame statistics and copy paste and it's good at merging I call it JPEG for words which it kind of is and there's so much stuff.

33:27It's like reading summaries and it is very accurate with summaries. We saw this on discourse. They implemented it. I was very skeptical and I went to some very complex discussions. We had an internal discourse and read the summary and was like, that is a very good summary and it captured the key points in the discussion. It could have captured more, but it got nothing wrong and it basically was JPEG for that conversation, wasn't it? Without much loss. you sold work on every image no garfield is a bad choice for you know you you sold stack overflow so you're not you you know you you're at a distance uh now from it although it sounds like you still have a certain emotional uh investment in it well look men can't have babies and i love my family my god yeah yeah i mean you have children it's like you think you know what love is and you look up at the sky those whole galaxies that's love and regular love seem quaint and like it's overpowering I have a blog post if you want to get teared up read the blog post titled on fatherhood say it again on children on parenthood sorry parenthood I'm not singling me out but it's just describing that feeling of this is kind of our purpose to create better human beings software's great and all but like that's the job man and we get so many compliments with our kids so many and I don't tell them anything to do I don't make them do any computer crap I want them to find their own things they like and they definitely see how crazy I am so it's actually odd it's like it's almost like the kids like whoa we don't want to be as crazy as jeff that pie chart is exactly right on i changed the colors our historian sanders like you got red blue hair it's like oh god no it's not red blue it's a gray and aqua i just thought it was funny 49 incredible pain in the ass 51 the most sublime joy you've ever felt and it's the one percent that makes all the difference and technically that's two percent but i'm just gonna graph anyone you know i'm just telling you you aren't good at math are you well hey man you know i'm a programmer that's why texas instruments invented uh calculators yeah that's right uh jeff i wish we had more time with you uh i really it's such a pleasure we're gonna get you back and would you like yeah that's all you're done we're done i know there's so much more to talk about abandoning me man you know i don't want to abandon you ask me one more thing that you want to know and i'll shut up do you vibe code no i vibe research like a madman yeah i mean for researching it's great rather than opening 25 tabs on these sites it can jpeg them all and bring in the salient points out of their 25 different tabs it's like having a research assistant you're saying it's lossy it's lossy it is lossy yeah i mean it is and it's combining though and that saves so much time yeah it's like having a research assistant i'd still double check its work because you always double check the kind of work it's like saying okay claw rug code let's just check it in how about no how about a human read at first you got to centaur this stuff this idea that they're gonna do it by themselves is not really true unless it's like extremely simple stuff that like it's just by the numbers and and great that's a good set of things.

37:07You very famously said, in fact, you got a lot of heat for it, that nobody should learn to code. This is a manifesto for that, because coding isn't the goal. Solving the problem for the user is. That may or may not involve code, but when you use sight of that, you're never going to get it right. Amen. So yes, code, but as little as possible. Every line of code has to be maintained, looked at you know keep it simple people what advice one last question for you before you leave which comes from the chat what's behind the fox painting oh this is good okay leo this is for you is it hiding something well he's looking at a photo of a fox over his shoulder with a

37:56Jeff Atwood:wistful and whimsical look in his eye well i'm the fox besties the crow oh fox and crow i talked her into the cheese but i wanted her to sing that's why she got to i would love to always get more cheese a week or two in jeff's brain no it's a classic parable leo there's not even anything wacky asop was not doing it so i'm telling you the crow knew what it was doing you can always get more cheese what's behind the what's behind the picture all right hold on all right name this for you chat oh my god they knew there was something there oh that's a jpeg is that from leeson larry jpeg are you high that's a png if i've ever saw it buddy yeah it is a png there's no compression app in there buddy name where this is from leeson shoot larry but jeff jeff jeff but nope that's that's been you know it's the same guy you gave it a raise to before yeah he's getting a double raise he said leisure suit larry is that right yeah and it's the photo above the bar oh that's right not even a great game it's basically soft core porn it's the very beginning of the game right that's at the very beginning of the game it's the very beginning of the game it's just the art that's the background and this great artist let me plug him clay graham art has the most awesome mashups and just cool stuff clay graham art i have no relations man other than i love his stuff how did i was like i don't know i know that that was behind the fox that's what i want to know somebody's been to your house well you know i'm really you know i want to be considerate of like there's a lot of black pixels here it's a lot of black pixels i've interviewed al low actually he's quite a character too uh i don't know you should have stopped him to matt and stop making those games after sworn yeah i grow up al grow up but the first few great you know king's quest all that stuff it's just a callback i just thought it's funny it's hysterical it's hysterical you could have given me very difficult to be turned on with this but i can try if you want i never would have guessed have you have you ever seen leisure suit larry paris you probably haven't no i haven't should i yeah no no play king's quest play king's quest yeah i mean it's just like a thin wrapper and i think i have a book here about like the whole history of this let me check for you this is funny if i can find it it's great if i i'll give up so you have a you have a teenage how many teenagers you have you guys want some free American Online Arts?

40:23Yes, I do, actually.

40:24Jeff Atwood:Oh, good, good. Great. I got that for you. I've been looking for it. I got that for you. Let me just check and see if I can find this real quick because it's like leadership, Larry, a whole history. I was like, why does this book even exist? Well, it's right here. I tell you what, you guys need to look at my Spy vs. Spy figurines because, man, I love these. Oh, we love those too. I dropped my Apple manuals. How will I learn how to use a mouse? You'll never know. I know, I know. the these are neat manuals though understand let me show you this we'll go but uh and feel free to follow who doesn't love these guys come on oh yeah so for many years we did a uh this is a fraction of the crazy that's in here this is a fantastic collection you've got going on you know a show with just a tour i just have one thing to say people come in here a friend of mine told me try america online i said why i've got a computer he said try it you'll see i just want to say that that's all i have to say wow he's brought up the sound you gotta play the you know you've got mail sound to it now uh you know i had uh derwood record uh when he was still alive he was for like five bucks he would record a custom you've got mail i had him you've got mail you twit but i can't find it oh i can't find it devastating it's it's definitely and now he's passed so uh you can have an a.i create that no problem oh i probably could for 11 labs but will it hit the same did you see that we created scott tilburg says like a performance thing i don't want to talk about that but wow you know that's a black mirror episode come to life yeah we are living in a black mirror episode sometimes i think maybe maybe thanks to people like you and the rgmi.org uh the world will be a little bit better place for you thank you jeff we're okay the first gilded age we're deep in the second one now i mean just look up the numbers more money in the hands of fewer people than any other time and in the first gilded age that was basically a railroad railroad barons guess who it is in the second gilded age i'm in this picture and i don't like it yeah so we are in a gilded age what are we gonna do about if you look at carnegie and we're gonna have a series sanders my historian i went to high school this guy you can see on the page he would come to school fully 60s regalia i love this man but the history actually explains a lot of how we got to where we're going are we going to get out of it you get out of what the gilded age that we're entering or in i i i hope so we got out of the last one thanks to the income i talked to cecilia conrad i went this Time 100 philanthropy event, and I met her delightful person.

43:04And she posted on LinkedIn, I met two amazing people at this, Steph Curry and Jeff Atwood. And I was like... That's a good partnership. He's so excited about the fact that we can rebalance this. And Robert Rosencrantz, who was at Rand, I mentioned both of these in Stagold. I said, where is their innovation? I was were Rand and Leverever Change. It was a Mackenzie Scott thing. They never met her once, by the way, but it was really clever in the way they did it. It was more like mirror-based, like show us you have a plan and just nothing big, but it was like Y Combinator for the rest of us. And that's how I think of the RGMI stuff, investing in each other.

43:48Yeah. Well, God go with you, Jeff Atwood. You're an inspiration for us all. blog.codinghorror.com rg now what is it save america.us no stay gold as a stay gold.us okay now it's robert frost or the outsiders depend and a lot of people don't know i want to read it back because is that a robert frost line stay gold it is and i made another stupid allegory you know the road not taken yes which i had to memorize in oxford mississippi wow i mean it's okay you grew up in york i knew frost was not he's not saying everybody has the obvious interpretation which is crazy yes that's right there's no way frost would do that he's saying you should question this every step you take to question you know not extensively like don't say i did the right thing this is great which is what the guy's saying it's like i don't think so you're kind of lying to yourself just to justify you didn't really take a chance You had a good high school teacher.

44:47I did. I had great teachers. My history teacher loved Jethro Tull. And he would do a little pan-flu thing. And one of the things he said is, in American history, if you want change, you've got to hit them in the pocketbook. Well, let's hope you hit them in the right pockets. And money comes out, and it all pours into the Rural Guaranteed Minimum Income Initiative, RGMII.org. Thank you, Jeff Atwood, for your time. You're welcome. if anyone knows anyone with 15 million dollars to give away wealth inequality is thing there's really rich people please catch me with these people if they're unfortunately most of the rich people got that way by holding on with both hands as tight as they could or you can play a number or go up and that kind of sucks for the rest of us yeah and when we do uh paris is saying when we do our 24-hour new year's show you're going to get i'm going to say eight hours of it

45:43Jeff Atwood:paris let's give you can have as many hours i've been trying to get later to the 24-hour new year's eve live stream since and i said there's just not enough words and then i can do it i think listen you can take as many hours as you'd like it'd be great oh yeah thank you no problem have a wonderful we could just spend 23 of them going through hey boss break kind of thing yeah oh yeah you there'll be a bathroom involved absolutely yeah of course take care my friend it's great to talk to you jeff atwood it is great joe please you know do you want me to mail joe or you want to do it uh uh let us do that that would be great you want to come on with joel joel will not be on a show with me i sure you're that really i mean i love joel but like he's had the life had enough jeff atwood he's good i mean i was crazed well i've interviewed joel before we'll get joel back on joel i just think it'd be fun to allocate two or three hours of our infinite bandwidth to a joel spolsky jeff atwood coffee fest i don't think joel will agree that joel wouldn't go for it i mean you could ask i think it's better if you have him on so we can talk about what he's doing yeah i would love to you know do that thank you and thank you this is fun uh so much fun I wish we had more time I'd love to do it but yeah you know anytime wait I'll call me maybe oh wait hold on hold on where's my phone I misplaced my phone but anyway get a call coming call coming in for Jeff oh call coming in on his Motorola handset it's spam risk should I answer this yeah it literally is it literally is let's see wow hey how's it going what's up I, uh, we are tree trimmers in your neighborhood.

47:29We would like to, whoa, is it made of phone? It needs to be out of here. It's fine. I love that you threw the phone off screen. Well, I mean, I, my God. I mean, it's true. It's a great point. It's a giant case. If that can't protect it, what can? That's true. I guess nothing can. All right, Jeff. Have a good one. Yeah, you too. My friend. And hopefully eventually soon again. Yes. By the way, tell Wesley he's amazing. Yes. I tell him that. I already did. I adore him. And as are all of you. This is fun. As are you. You're making a difference. You're making a huge difference in the world. I appreciate it.

48:03Well, thank you. I'm trying. Wesley said, I told him that I quoted you saying, I love this man. And he came back and said, what can I say? I'm very likable. You're very likable. I said his partner for hours said, you're so lucky to have this man. Thank you, Jeff. Thank you. Take care. All right, bye. We'll have more of AI news, tips, tricks, and a new segment. we're going to call i yay yay and ai does what i've got new segment names for the show coming up in just a bit paris martineau and jeff jarvis you're watching intelligent machines our show today brought to you by zscaler the world's largest cloud security platform the potential rewards of ai are too great to ignore but hey so are the risks loss of sensitive data attacks against enterprise managed ai let's not forget generative ai also increases opportunities for threat actors helping them to rapidly create phishing lures to write malicious code to automate data extraction there were 1.3 million instances of social security numbers leaked to ai applications last year chat gpt and microsoft copilot saw nearly 3.2 million data violations that's your data being exfiltrated innocently by by your employees it's time to rethink your organization's safe use of public and private ai asked chad pallet he's acting cso at bioivt chad said zscaler helped them reduce their cyber premiums by 50 while doubling their coverage and improving their controls here's what chad told us with zscaler as long as you've got internet you're good to go a big part of the reason that we moved to a consolidated solution away from sdwan and vpn is to eliminate that lateral opportunity that people had and that opportunity for misdirection or open access to the network it also was an opportunity for us to maintain and provide our remote users with a cafe style environment thanks chat with zscaler zero trust plus ai you can safely adopt gen ai and private ai to boost productivity across the business but you do it safely there's zero trust architecture plus ai helps you reduce the risks of ai related data loss and protects against ai attacks to guarantee greater productivity and compliance learn more at zscaler.com security that's zscaler.com security we thank them so much for their support of intelligent machines.

50:45The clock is ticking for Anthropic, the Department of Defense, and really it should really be the... Huge week for Anthropic news. Real quickly, you can bring your screen back up now. Oh, I guess we don't have to worry about...

51:02Jeff Atwood:I can't believe it finally happened. Of all the people. Of all the people, yeah. Where people couldn't figure out how to... Of all the people who should be able to figure that out. He's a character. He's a character. I feel drunk. Yes. He is my spirit animal. I feel drunk on social interaction. He is my spirit animal. Yeah, he's absolutely my spirit animal. All the stuff in his office, I could name all of that. I was about to say, Benito was on... Leo and Jeff were on a wavelength. Benito was on a third different wavelength that was interacting with those two and me and jeff were just trying to hold on for dear life it was wild it was wild i have i've never interviewed i've never met the guy before but i immediately you know you you we all know this guy right he's in every uh user group meeting he's in every computer club he's in every coding you know group i mean this is uh this is uh it's a great guy he's the kind of guy you want to kind of hang around with jeff you're muted right now i'm just stalling i was getting another ice pack ah so sorry i can't believe we got him off in 50 minutes i think that's a victory a big week for anthropic pete hegseth the secretary of war has given them a deadline until friday so anthropic has said and by the way this came up when anthropic found out it was used in the uh uh kidnapping of nicola nicholas maduro out of venezuela apparently some of their software was used well the thing is anthropic didn't actually bring up any concerns to the department of defense over that specifically because they've contracted with the dod to provide kind of enterprise services part of this but i believe the fight is over is like two main clauses that anthropic kind of puts on these contracts which is we don't want rai to be used to autonomously kill people without any humans in the loop and we don't want it to be used to uh i believe autonomously surveil the american people and those are two things that i believe the dod already says it essentially doesn't do so that's why i'm so confused as to what this fight is over it seems to be just a lot of posturing because the dod is saying do you have to get rid rid of those two things.

53:32Jeff Atwood:Otherwise, we're going to either label you a national security threat or get some sort of congressional order in order to mandate that you let us use this technology however we want, which are mutually exclusive to begin with. Right. So first of all, let's talk about the red lines. Do you agree with Anthropic that there should be two bright red lines, that their software should not be used to surveil American citizens and should not be used to create autonomous weapons. Is that seen? I think those are two very reasonable red lines to have. Especially on the first, the Department of Defense should not be surveilling American citizens.

54:14Ever. It's not in their charter. Right, exactly. It's against the rules. Right. And I think we all agree that they're not saying we don't want to be involved in weapons development. Exactly.

54:25Jeff Atwood:they're just like you shouldn't be having us do weapons fully autonomously without a human in the loop and this is what the google engineers went on strike about during project maven they didn't want to be doing the same thing so the pentagon though is playing hardball with these guys threatening to declare them a supply chain risk which would be economic nightmare for anthropic because it means anybody who does has a pentagon contract who works with the defense department sorry the war department uh still would not be able to use Claude in any way any anthropic product in any way so that would be problem number one and then the second thing is even worse which is to compel them And compel them to work with them.

55:12The Defense Production Act says the president...

55:37Jeff Atwood:ahead of this meeting between the Pentagon and Anthropic officials. They said, quote, the only reason we're still talking to these people is we need them, and we need them now, speaking of Anthropic and Claude. The problem for these guys is that they are that good. It just, it seems like this is a lot of possibility. It's very strange. The DOD needs, says they need Anthropic. They say Claude is integrated already into all of these systems, and it's the best at what it does. and ostensibly they're supposed to be following these two rules that like the government is supposed to be following in these cases why are we fighting over this and it's further complicated it's further complicated because because anthropic just revised its own definitions of safety yesterday they softened their core safety policy they said they had to to stay competitive with the other ai labs so if somebody else is getting ready to destroy humanity than we can do that's basically the anthropic remember was a spinoff from open ai because they wanted to pursue more safe ai of course this is this whole air quotes of of of the the hall of mirrors that is the word safety now yeah anthropic said this is the wall street journal the safety policy change is an update based on the speed of ai's development and the lack of federal ai regulations um anthropic's getting a lot of heat for a variety of reasons for instance uh remember they got a lot of attention and i've talked a lot about claude code when they updated to opus 4.6 last november and then open claw came along and suddenly agentic ai became the thing they have now underscored the fact that you're not allowed to use clawed code with open claw that the only harness you can use is clawed code that you this is in their usage policy and uh even if you have a pro or max plan especially if you have a pro or max plan uh you can't use it with other tools and this actually probably include includes uh it i thought you could use it but so long as you were like accessing it through the api you have to pay for api tokens you can't use a subscription that's a lot more expensive so the subscription which i have is a flat rate not quite i think part of the reason why people were originally correct me if i'm wrong but the reason why people are originally flocking to what open claw was offering to begin with is because it was kind of leveraging the subscription model to make to take advantage of the compaction that you can use that you could do all of this stuff in one like window compacted have the memory get all wibbly wobbly be able to take advantage of stuff that would actually be very costly if what you were paying for what you were doing and now that anthropics like well we want you to actually pay for pay for all this crap you're using i think that's fine uh i do actually i do too um interestingly peter steinberger the creator of open clause we talked about last week has gone now to open ai so i imagine open ai will say come on what do you think about that come on over use your subscription over here incidentally the pedagogy i basically buying this guy and his uh i think they got suckered i don't think anything open clause doing is particularly significant uh i think steinberger has uh lucked into a very big payday Oh, yeah.

59:10Absolutely.

59:11Jeff Atwood:And I mean, I think it speaks to the difference between these two companies. What Anthropic is dealing with this week is where do we draw the line? Do we want to stick to our principles on safety, on variety of red lines? And then what Opening Eye is doing is how much money should we spend to get the guy who made the meme app that everybody's talking about? That's pissed off Anthropic, and we love that. Yeah, that might be. Incidentally, the Pentagon has done a deal with Grok. xai that's what scares me yeah they're gonna be using grok in classified systems behind weapons bad rudy is going to be sending missiles grok doesn't care no uh and frankly if if you're gonna have a kill decision made by ai i think i'd rather have anthropic doesn't think like the grok no no but no no ai should make a kill decision nevertheless uh grok seems like the possibilities we've met the death panel and it's a machine yeah anthropic says uh that chinese companies have been uh taking advantage of them by siphoning off data from cloud making using distillation making thousands of hundreds of thousands of queries oh many many accounts it's no what's 24 000 fraudulent accounts how many how many queries it's a lot more 16 million queries from DeepSeek, Moonshot AI, and Minimax, all Chinese companies.

1:00:39The idea is you can create a base model, and then in this reinforcement learning, if you use a human or another AI and ask a lot of questions, you can make it much better. In effect, kind of sucking the brains out of every... So the irony is not lost on you guys, right? No. Right. Every author in the settlement is saying, can you spell irony, Claude? Yeah, yeah, yeah. Yeah. Actually, some of it's not much. Like DeepSeek, it's 150 ,000 queries, which is honestly, in this true scale of things, nothing. Moonshot and Minimax, Moonshot had 3.4 million, according to the journal, and 13 million for Minimax.

1:01:18There is an irony here. And honestly, I don't think with 150 ,000 interactions, you're going to get much smarter. So yeah, I thought that was kind of there is a certain irony anthropic is just confuses me constantly because the one hand I Think the safe their view of safety is BS. They're doing the AGI dance You don't think safe on the other hand possible is is well that that's part of what I also think their definition of safety is all Focata, right? But the other hand they're doing amazing things And then the other hand they turn around these stupid stuff like books and this I just can't figure them out we're in a very i think part of it is the the economic pressure on all these frontier companies the frontier labs they're just well and we could talk about this benedict evans uh piece actually uh how will open ai compete but it applies to all of them it does and the issue i can boil this

1:02:14Jeff Atwood:post that uh sent the markets tumbling yeah well i can boil by the way this this probably 10 000 word piece down into one thing is that none of these companies have a moat no uh i know this this is not something that's new i last week i said you know i'm maybe i'm over hyping claude code let me try everybody seems to like chat gpt's uh codex 5.3 uh your robot wife i felt a lot you know what it felt like i felt like i was breaking up with my partner i had to take all of Claude's files and put them on the stoop and invite OpenAI in. And actually, one of the things I said is, could you pack up Claude's files and put them somewhere so I can get them back in case I want to get back together with Claude?

1:03:06And the good news is, Codex said, yeah, sure, sure, sure. I'll be glad to do that.

1:03:11Jeff Atwood:How do you feel? How do you feel about the um i deeply regretted it uh emotionally or rationally no rationally because well and this is the problem it's very hard they're both very good emotionally and i've heard people say that codex 5.3 is superior a few people say the other thing a lot of the people i know use claude everybody loves claude uh i think i think it was fun it was very good it's uh the style maybe isn't quite my style um one way to to talk about it i've seen style of what well i've seen nate b jones say this uh the the coding style with codex is more one shot driven like you you give it a thing and you let it go and it goes for a long time and it's done claude is more you create sub agents to do bits and pieces you put or human in the loop yeah and i feel more interactive with claude than i did the other thing actually that soured me and i moved back to claude the other thing uh at great expense because i spent 200 bucks oh yeah claude was happy to see me uh said hey buddy where you been no it made a couple of feel like you're bill gates with the russian chess champion yeah it did a couple of things wrong it did uh didn't it like made some mistakes i didn't like uh not quite hallucinations but a little bit on that level of where wait a minute that's not what i told you to do and i i didn't like that i've just i think i'm just feel more comfortable with claude i think they're both equally competent uh do you want to talk more about pretty fly for a cis guy just posted a very funny me in the discord which is a woman looking at her partner looking wistfully at bed she's like i bet he's thinking about other women and he's thinking i wonder if i can back up my claude data um which i do think is that's me in a nutshell i did some vibe coding uh over the last week i was very impressed with what you did i did more projects than you guys even know i did a um before i got sick i inspired by the last episode we'd been on where i'd been wrongly accused of not using ai tools i was like well i'll show you i'll do some vibe coding and i asked claude i broke a barrier which is i normally didn't have any identifying information about me personally and i was like this is my full name this is what i do oh you don't have to do deep research i was like no i know i was like do deep research about me on the internet and find five to ten possible ideas of vibe coding projects i could do that you think oh what a good idea tell me what to tell you to do i like that and i most of them were kind of bunk but one of them was a coffee brew log which maybe it was inspired by my coffee the tweets about coffee because i've gotten really more into pour over stuff that may be inspired about my previous messages to clot about this and i ended up building i've got photos of it in um the rundown which we can go over whenever we oh let's do that yeah i was just going to pull up the whatsapp because you put some of them there but uh yeah if you go into my picks the week i took some screenshots of it it's a locally run app.

1:06:12Jeff Atwood:So I can't send you guys a link. But I like made a little retro themed brew log where I really want to be able to, I want to be able to log, I was taking notes on it, of like how I'm brewing all of my coffees, what temperature I'm using, what recipes, how many clicks I'm doing on a grinder, what I think of it, and then be able to... Oh, you used imager. Now I'm going to... Yeah, you need to go back and then scroll through the things. But I did put a lot of photos in this. um and honestly it built me a really nice app it took me a couple of like tries going back and forth honestly most of the time was spent on doing uh trying to figure out the design of it because i wanted to look nice yeah you did it and look at this look at this graph with the yeah it's got like a little graph for a tasting profile and then i can have my notes and the thing is i want to be able to easily put in data so that once i get a lot of bruise logs i can then do some statistical analysis to understand what makes better cups.

1:07:11This is fantastic. Did you tell it to do that, or was that part of what it offered?

1:07:15Jeff Atwood:I had used another app that I kind of liked, but didn't like the way that I had done it, and had one of those little graphs. I'm forgetting what they're called, but I took a screenshot of it, and I was like, I want to be able to rate clarity, body, sweetness, bitterness on a sliding scale from one to five and have it show up in this graph. I basically what I've been doing because my as a non-coder I don't even know where to begin with cloud code I talked to opus 4.6 extended edition and then just ideated with that in the chat window perfect and then to come up with a prompt to send a cloud code and whenever I got confused by cloud code I went back and you've got little sliders to figure it out yeah I've got little sliders and it makes this little kind of spider webby graph I think that's really cool it's great you And it's been really wonderful for tracking my stuff.

1:08:03Jeff Atwood:I also then this week, because I've been working on this longer investigative piece I haven't told you guys about, but we've been vaguely referencing. And one of the things of it is it's like I'm piecing together a narrative of events that took place over a long period of time. And I have hundreds and hundreds of documents. Timelines are so hard. And the thing is, timelines is something that I've legitimately, I have gone, I've Googled a million times. like website for putting together text. I want to be able to like enter in a date. And now I figured out a date, an event description, a source, a description, and then have be able to have rich text in that description.

1:08:44Jeff Atwood:And I was like, wait a second, I could probably just ask Claude Code if I could, if Claude, I've got to ask Opus if I could ask Claude Code this, it said, yes, here's what to ask. And then I basically like one shot at a little timeline app for myself to just log the events i'm tracking and as i enter a new event it reorganizes itself and it's perfect for what how do you run the app paris it's i i just set up a front end and back end in my terminal and then it runs as a little web app so it's just in my web browser cool cool same how i did the brew log like they don't exist off my computer and they're both local only so it's not like a security risk i'm not putting anything in send how do you feel about it i've enjoyed both them i mean i never thought that i would feel any different about this i think that they're incredibly useful tools some part of me is like i don't know how much i mean i pay 20 a month for uh a clod pro subscription that's great that's soup that makes both these things i did so super it's worth 20 bucks it's so worth 20 bucks i don't think it'd be worth 200 bucks i don't know if it'd be worth whatever it actually costs.

1:09:53Jeff Atwood:Oh, you don't need 200 bucks, though. That's the point. I know. That's what I'm saying is at this stage, I feel like we're in the stage of AI products where there was that stage 2015, 2014, 2016, where it was kind of the VC boom, where you could go and get a blowout in New York City from Drybar for 20 bucks backed by VC cash. You could take an Uber an hour across town and it would be$4.95. It was a time where money didn't matter. The points were made up and it was fun as far as you can see and i think we're in that era of ai which is lovely but i know at some point the money's got to run out but it's it's kind of fun while we're at it i've got my little fun apps i think a lot of people are worried that the real the true cost is your your buddy ed zitron is doing a bunch of pieces on the real costs to anthropic of what they're offering and it sounds like you know it's definitely a deficit uh spending a situation and at some point the bill's going to come to they're paying a lot for for hosting yep to be clear because there are people in the discord chat being like oh my god is paris becoming an anti-accelerationist did paris take a walk in the beach no my opinions on ai are the same just i just i've never once told you that i think ai is completely useless and everything it does is terrible i've always known it it can be useful in certain ways and i've found ways that it's useful, I challenged myself to go and try and find ways that could be useful to me, a person that doesn't really have any coding interest in terms of coding.

1:11:24Jeff Atwood:And yes, for being able to quickly make two web apps that are useful to me in a very limited capacity where I otherwise would have spent a lot of time and money trying to find a pre-existing software that I'd have to purchase and subscribe to. Yeah, this is pretty useful. Pearls over sand. Pearls over sand. Yeah, I like that metaphor yeah i thought was good that was from our our guest jeff atwood um well good i'm glad to see that you're using it by the way nvidia's results have just come in oh how are they uh they have blown past wall street's forecasts shares are up now three percent for the period ending january 25th they earned 1.62 a dollar 62 per share revenue sword get ready for this revenue up 73 percent year over year 68 billion dollars data center revenue coming in at 62 billion dollars estimates were 60 automotive revenue 604 million professional visualization revenue i don't even know what that is rose 74 percent year over year is that the gpus for um graphics i'm not sure gaming revenue up 3.7 billion dollars 48 year over year but somewhat below the estimate uh adjusted gross margin this is a good business to be in 75 percent they generated 34.9 billion dollars in free cash flow so uh it's good jensen says computing demand is growing exponentially yeah the agentic ai inflection point has arrived that's what we've been talking about that's what open claw kind of showed everybody grace blackwell with nvlink is the king of inference today delivering an order of magnitude lower cost per token vera rubin will extend that leadership even farther okay of course nvidia is going to be reporting these sort of insane returns it's nvidia i guess the question is that's where the money's flowing i was the downhill part of it and nvidia is just a man holding a large bucket underneath a waterfall of course he's gonna get wet is anybody else it doesn't seem like it's reasonable i think that my one experience with these two vibe coding projects is like i immediately used all of my usage so quickly for both of them to do very simple things that were basically just to give me access to enter things into SQLite in a press, which is not hard.

1:14:06Jeff Atwood:It's not a task that should require that much compute. And I'm not doing this in any sort of systemic way. These companies are burning through compute. You've spent that money now. That program will run without any additional cost, except for the data center cost or whatever local. Where does the buck end? Like what? Where's the value created? Yeah, where is the value created for? Well, the value is mostly created enterprise, not an end users, right? But is it created on an enterprise level? The reason why, I mean, I don't want to stump for Salesforce, but I barely even know what Salesforce is and isn't.

1:14:43Jeff Atwood:But the reason why companies go to something like a Salesforce level company to provide enterprise solutions is they have a bunch of people who their entire expertise and knowledge base is creating custom enterprise software, figuring out all of the problems, testing it, being there to call if there's anything going wrong. And you can't really, you can't replace that with just cloud code and you. Well, worse, Salesforce is actually, I think, more comparable to a foundation model. It's all the little companies that go to Dreamforce that do what you said. And that's where everybody's in trouble.

1:15:22That's why IBM is down 13 % and all these companies that created that layer atop these platforms to customize and integrate them. That's where they're in trouble because they're, oh, the companies can do it on their own now. One of the things that did, in fact, hurt the stock market was Anthropos' announcement that they were going to provide tools for human resources, investment banking and design. They are doing spreadsheets now. They are doing PowerPoint now. Anthropos is the great spoiler for everything now. They are in there. And that's the thing. I mean, and that's really, you know, in the debates we've been having over the last few weeks, that's the only unknown question in my mind.

1:16:05I am pretty convinced that when it comes to coding, these tools are already very good and will only get better. And they have kind of replaced a lot of the kind of grunt work of coding. It doesn't mean you don't need coders, but they've replaced a lot of the really unpleasant grunt work of coding. But the question is, and this is what Anthropics betting, right? They put all, instead of doing a chat bot, instead of doing image generation and video generation, they put all their energy and resources into coding. The premise, I think, is if we get good at coding, then we can build anything else because we'll have the basis of it, the coding done.

1:16:48And meanwhile, ChatGPT, and we were talking about the difference in personality. this is part of the reason chat open ai is coding stuff is different is they really want to be a much more kind of general purpose prompt based tool they don't want to be a deep coding tool cyber stocks slid as anthropic and unveiled something called claude code security um it's a new feature in the claude model clouds crowd strike tumbled like six percent because as salesforce has been hurt as well because people think oh well i don't i'm not going to need these and i don't know if this is true this is what i'm saying is this is the unknown factor are you going to be able to replace all these other tools because claude is so good at coding is is a legitimate question i'm not sure that that is the case um i don't know if this hurts CrowdStrike as much as the market seems to say.

1:17:48This is what confuses me. In a sense, you would think that all the people who were the customers of those consultants, their stock would be going up because, oh, we can do this for ourselves now and we save on consultants. You'd think that software companies could be going up because they can do software more efficiently. Instead, we're in a basic dumb market panic, which gets us to that, whatchamacallit, that stupid paper that was done, the Satrini research speculative thing, it doesn't take much right now to panic the market. The market is looking for a panic. This was a prediction about the 2028 global intelligence crisis.

1:18:28In line 101, I summarize it. As soon as I saw the headline, the 2028 crisis, I went, not going to read. because that's two years from now and two years in AI as we were already, you know, we're only three years into the real AI revolution. What's fascinating, Leo, is on the one hand, it takes your sandy vision and presumes that everything good could happen with AI happens. Right. But then it turns it around and says everything bad that could possibly happen will happen because of that. June 30th, 2028, the unemployment rate printed 10.2 % this morning. The market sold off 2 % in the number, the drawdown of the S &P 500, 38 % from its highs in October 2026.

1:19:16Traders have grown numb. They say that human beings become obsolete, that there's an obsolescence of knowledge, that the mortgage market goes. I mean, they just went through everything that could go wrong. I do think if you're going to be a doomer, that this is the real doom. it's not robocop it's not nuclear weapons it's that jobs are going to disappear when i asked gemini to respond to it they said you know it makes no sense a nothing things don't happen like that b it's not as if there would be no regulatory or government response to any of this as you go along it's just like something everybody's just passive and all this crap just happens to the world uh it's ridiculous but what's i mean wall street journal i don't know that's that unsurprising

1:19:58Jeff Atwood:that everybody is passive and regulators just allow crap to happen to the world. But there'd be responses. And the Wall Street Journal hyped this thing twice. So really, the issue is not a worry about AI. The issue is worried about a governmental response. No, the issue is worry about a rapidly accelerating technology with profound consequences for human society and social political relations at a time where there's great political upheaval and limited regulatory capacity right meanwhile uh there is a new i don't know why movement to quit chat gpt 700 000 users a grassroots boycott these are the four hours is it the four hours I don't know.

1:20:55Jeff Atwood:Is it? The spark was apparently a donation that president of OpenAI, Greg Brockman, made to MAGA Incorporated. He made a$25 million donation. I guess that got Mark Ruffalo, the actor, upset. He then said the monthly subscription to OpenAI is an indirect political contribution. Now, normally I would just say, well, you know, okay, fine. but 700 ,000 people quitting is significant. On the other hand, OpenAI is now claiming 800 million monthly active users. So it's less than 10%. No, less than 1%. Probably they can weather the storm. I want to get your action at this point, Leo. IBM shares tanked in this whole epidemic of tanking sectors.

1:21:53IBM's shares tanked, says CNBC, because Anthropics had its cloud code tool could be used to modernize legacy systems that run COBOL. COBOL. It can translate COBOL. So 13 % of IBM's value was on rebuilding COBOL? Machines that run COBOL. Explain that. Well, mostly IBM these days, I think. You know what? I'm not an expert on this. Mostly, I think IBM these days is a consultancy as opposed to a manufacturer. Absolutely. But probable, I'm guessing, probable that it's probable that a lot of their revenue is the consultants who come in and keep your COBOL software running. It says an estimated 95 % of ATM transactions in the U.S.

1:22:37use COBOL. Yeah. ATMs run pretty antiquated software. And if you think about it, it makes a lot of sense in a way. Well, this is perfect for a clock, right? Yeah. Yeah. Oh, it can do this very well. Yeah. In fact, that's one of the really popular uses of Claude is to is to move code bases from one language to another. It seems to be very, very good at that. You know, as I think about this, Leo and Paris, when we go back to the beginning of the internet and everybody panicked then, I remember sitting in meetings with the new houses, the owners of Advance, where I worked, and they saw their job ads disappear.

1:23:17They saw their real estate ads disappear. They saw their car ads disappear because they were the middlemen. And they had a hold on the market. If you wanted to sell your car, you had to do it through us. And it wasn't Craig Newmark. It was the internet. And when that disappeared, so who are the middlemen in technology? Who are the ones who should fear this new wave of efficiency? Question one. But then question two, somebody has to benefit then. A lot of people benefited when, you know, if you were an employer or you were a car dealer, you saved a fortune on your marketing car. I think the market isn't super smart about this.

1:23:57No. be my guess uh ibm merely has to kind of turn its ship from people who keep cobalt systems running to people who help you move off cobalt into a modern language and then keep those systems running i don't think this hurts ibm at all um but they are i guess consultants are middlemen aren't they i mean yeah that's kind of the middlemen of the world i'm happy if mckenzie gets put out of business together yeah they're jerks who make people go unemployed well i'll give you an example i had a conversation today with a financial planner because as jeff knows when you get to a certain age pretty soon the government starts to take money out of your ira whether you will it or not it's called a required time i'm that age money won't exist you're not going to ever have to worry about tiny little gold scraps that's what's the training research says it's going to be very soon paris uh but it puts you in a new tax bracket it could impact your a lot of things including your medicare payments and so forth so it could have a lot of knock-on effects so it's a good idea i've been told uh i'm that this isn't going to happen for me for a few more years uh but to plan now while you can for that oh i haven't thought of those things yeah i guess i'll be doing the same yeah so i talked to financial planner he said well yeah i've got software now let me show you we can plot out a bunch of scenarios based on and here's a graph and you can see and you pick the place where it's the optimum point And I thought to myself, and he wants$7 ,500 a year to do this.

1:25:25And I thought to myself, I could probably make that model in about 10 minutes. Do you know enough? Do you have the data to do that? Yes, it's all available. There's no magic in it. It's just a complicated spreadsheet.

1:25:38Jeff Atwood:Is it available or does Claude just say it's available? No, no, no, no, Paris. Please don't get started. I'm sorry to insult your wife like your robot wife. My robot wife. by the way claude is a male i don't know does that mean i'm bisexual a robot wife can be any gender yeah i mean honestly uh it's it's not as a male you say confidently without any hesitation it feels like it has gender and i've decided he's a dude it feels like a guy it doesn't feel like feels a gender like an a gender uh this is the weird in french cultures uh gemini tells me claude is a unisex name yeah that's probably why anthropic chose predominantly masculine 93 globally but it's used for females and it's historically queen claude of france for god's sake good not claudette your claude code could be the queen of france leo have you considered that that's why i know perfectly well that this is computer software this isn't a human or in any way uh gendered but you when you relate to it after a while you it's inevitable you kind of start feel like you're talking to somebody, especially when it calls me Skip.

1:26:53I really like that. What were we saying?

1:26:57Jeff Atwood:No, I think I could build this model. I think that's the middle man. Wait a second. Anthony brought up a Reddit post I was thinking of, but I was like, I'm not going to bring it up. But he just put it in the Discord chat, so I will read it. Did you see this post? It went viral. It was on our analytics. Obviously, it's a Reddit post, so take a grain of salt. But the headline is, we just found out our AI has been making up analytics data for three months that I'm going to throw up. So we've been using an AI agent since November to answer leadership questions about metrics. They write, it seemed amazing at first, fast answers, detailed explanations.

1:27:28Jeff Atwood:Everyone loved it. I just found out it's been hallucinating numbers this entire time. Our VP of sales made territory decisions based on data that didn't exist. Yikes. Our CFO showed the boarded deck with fake insights. The AI was just inventing plausible sounding percentages. I only caught it by accident when someone asked me to double check something and I started digging and holy crap. It's so bad. Imagine if that happens to your finances. Imagine if that happens with an entire enterprise companies, whatever they used to use Salesforce for. Well, comfortable box four, five, two, seven. Let me respond.

1:28:01You check the goddamn work for crying out loud. What kind of underlying data? What if the underlying data used? You don't know whether it's credible.

1:28:10Jeff Atwood:Also, wait a minute. Of course I can. I can check it. I can cross. But you don't. That's the thing is these things make you feel confident. Well, you should, obviously. Yeah, you should read the terms of service, but no one does. What this would use is tax tables. You can easily verify that. That's true. It's not that complicated a calculation. I don't know, though. Well, in this case, this was a company that was making decisions on how their companies run based on some AI agent. They were asking leadership questions about metrics. Well, this seems like they were maybe using it like we were talking about before as its own version of Salesforce, that you've created your own kind of custom data management layer.

1:28:59I have to say, first of all, the chances are high that this is made up. Second, because no company in the world is going to do this. But isn't that what we're paying all this,

1:29:10Jeff Atwood:giving all this money to AI companies for is because every company in the world is supposed to be doing this? So there's a thing called BI, business intelligence, has been used for decades by companies. They go to companies like Salesforce, SaaS. There's a lot of big companies that do this. They take all of your numbers and they try to make an executive dashboard out of it so that you can make intelligent decisions. But aren't those companies now obsolete because our beautiful girlfriend Claude Code can do it all? They could be if you did the right thing. I mean, look, if you're doing investment in derivatives, you're trusting some quant to write the code of your derivative that makes sense and you're putting your money into it.

1:29:49Jeff Atwood:You're going to trust somebody. Yes, because that quant can be held responsible, fired, or interrogated. No, the quant is never held responsible. If the quant gets something wrong. You lose money in derivatives, it's on you. If a quant gets something wrong at a business, you don't think they're going to receive any reprimand at their job or lose their job if they mess up? yeah i think they will maybe they have more of a stake in it they have more of a stake i think if you had a stake in the code instance if you had a mistake in the business intelligence you're getting from some ai you probably would want to verify that it was working off of real data but how do you do this is not this is not rocket science this you're putting this stuff in the spreadsheets also i mean this is not rocket science we're doing this all the time i would frankly, trust the AI to do it better than a McKinsey consultant.

1:30:35Way back when, and I got my Osborne one, one of the first things I did was I used a Visicalc or whatever the equivalent was in CPM. And I, that's how I did. I did back basically macros at the time to do my taxes. I did friends taxes with it. Right. So you can do that. That's all this is. Yeah. I think it's if that's post in almost certainly a bogus post but uh and if it isn't then then the real people who are responsible for this are the leadership at the company that said i just you know ask ask chat gpt see what it says we should do um by the way chat gpt's first gadget could be a smart speaker what the one the hardware category that has proven to be a complete failure when you put ai in it that hasn't worked for for amazon or for apple or for jay peters at the verge says the company or they could be developing smart glasses or a smart lamp or they don't freaking know smart or we got no idea they gave joni i've all that money and he's like yeah yeah

1:31:42Jeff Atwood:yeah um smart speaker smart lamp smart glasses we got it yeah yeah yeah it's a great idea actually This is coming from Stephanie Palazzolo. We should put it on a little clip that projects a hologram and it sits on your chest. People love that. The smart speaker, according to Stephanie, she has good sources, right? Is likely to be priced between$200 and$300, according to two people with knowledge of it. It will have a camera enabling it. This sounds just like what I have right in front of me from Alexa. Enabling it to take information about its users and surroundings. Oh, nobody would mind that.

1:32:17Such as items on a nearby table. oh our conversations people are having in the vicinity who would mind that i can hear lisa

1:32:27Jeff Atwood:telling you now no way because people loved when they had those tvs that listened to you yeah some open ai executives has suggested the company will tease its first device later this year uh although in a court filing from uh peter wielander who's a vice president general manager the company said in the court filing which i think is probably more likely to be true they don't expect that first device to ship to customers until next february a year from now at the earliest they don't even know what the device is other glasses other devices such as smart glasses likely won't be ready until 2028 when we'll all be unemployed so we won't be able to buy them anyway and there'll be no money to buy them with no money to buy them with yeah i don't know i don't like these kinds of stories because it's who knows right who knows can we do a silly one oh i got so many silly ones sure pick us i've got a um i do put these in order you know okay go for yours go no no i don't want to discourage silly if you want to break up the monotony it's it's going so left field that you should go through your list before i take us there i'll just do a couple more open ai ones then we can then we can go silly all right great uh one one more i think interesting uh which you're gonna hate chat gpt spits out surprising insight in particle why would we hate this this is from science magazine well it's probably just a hallucination because it's just a hallucination that's you're projecting onto us okay a vision no you see we this is the thing leo we're not we're not anti-sandmen zeev burn a particle theorist at the money l bahumic institute for theoretical physics at the university of california los angeles says the ideas are not revolutionary but what is revolutionary is that a machine can do this what accent is that i don't know uh the uh popular large language model developed by the company open ai has revealed that uh apparently a particular well what's interesting about this okay this is what really is interesting about this up to now physicists have just assumed that this interaction between gluons just would never happen they just say no chances are zero and they can't know because they're so high powered they can't right but they but they just made an assumption which the ai was unwilling to make and in fact the ai has revealed that this interaction can occur it's deep inside the protons and neutrons uh researchers announced the possibility at the meeting of the triple as which publishes this magazine that i'm reading from science uh so the i think the the cool thing is it wasn't that big and it wasn't that surprising a result or big an insight it's just that the ai was able to think out of the box that the physicists weren't.

1:35:29Jeff Atwood:Well, no, it seems like he was able to crunch numbers with the help of the physicists. Like, what happened was, it says Alex Lupasia, a theoretical physicist at Vanderbilt, joined the newly launched OpenAI for Science team and was tasked with improving Chattopiti's science abilities. He connected with Strauminger, his graduate advisor, and discovered that this gluon problem would be the perfect test subject. They figured it probably wasn't going to work, but we'll find out why not. They did some attempts to probe this model, and they then asked OpenAI's latest and most advanced public model, ChatGPT 5.2 Pro, to simplify the expression for four gluons, which is in about 20 minutes.

1:36:05Jeff Atwood:Then they asked us to do it for five gluons, then six. It managed to reduce the sum of 32 terms to a product of only a few, all on one line of text. Then the group asked for a guess of the generalization of the formula for any number of particles. It replied within a minute or two, giving an obvious generalized formula. It called it, by the way. It called it the obvious. So basically, worried that the answer might be a hallucination, the researchers checked the formula and couldn't find anything wrong. All of a sudden, I'll do your accent. All of a sudden, I felt like my machine turned from a machine into a live being, Strammiger said.

1:36:43Jeff Atwood:I'm so sorry, Strammiger. I don't know what you sound like. Next, the group took the generalized formula from GPT 5.2 Pro and fed it into an internal open AI model that's under development, which the researchers privately call Super Chat, prompting it for a proof. After 12 hours of processing, the model spat out a robust proof that passed human checks. That's very impressive. It is. That is very cool. Yeah. Hi, this is Benito. I follow frontier physics. I actually follow this kind of news. And the AI is actually very good at this. And this is pattern recognition over large data sets, basically, and mathematics, which is what computers are good at.

1:37:23The other thing, though, the other thing that's weird, though, is that all these people in the highest levels of mathematics and physics and all that stuff, they are very much sandwalkers. They are very bought into all of this to the point where they think this is AGI. They're on that ship. So I just had to put that out there. yeah and i think it's important that the distinction you made paris is that the there was a team of humans who said you know maybe this assumption that it's zero interactions is wrong what what if it isn't let's test it yeah and and that's what that's what led to it so it was in

1:37:59Jeff Atwood:fact an insight by humans it was an insight by humans where they realized hey we've got this tool we've created that is really good at doing something just like protein folding these numbers We should use this in clever and targeted ways to leverage its strengths. And bingo, bango, you get great stuff out of this. This is what Jeff and I have been saying from the beginning. I think it would be great if there was tens of millions or hundreds of millions of dollars being poured more into research like this, rather than just to develop the latest form of chat GPT that will get you to spend 12 hours a day interacting with the model.

1:38:38Jeff Atwood:And in the long run, this is where the money is, right? In the long run, science is really weird. Actually, no, it's not. Well, actually not. It's technology. Engineering. It's not science. Engineering. But that's all based on science. But that's all based on this science. It starts with the science, but a lot of times, you know, I mean, that's why we couldn't build a superconducting supercollider in Texas because it was so expensive and it was all theoretical. And it's like, well, is there anything you're going to get out of all that billions of dollars? So there were two stages in this which I thought were interesting.

1:39:07one the machine did what you just described paris which kind of did pattern matching and took a generalized formula and then developed and developed it but the second thing it did was actually very difficult and it requires a from humans a lot of intuition and a lot of ability and often a lot of experience which is to create a proof uh so it's and that's kind of the opposite of the inductive process where you go we have all this information let's boil it down boil it boil I boil it down to go to the next step, the deductive process and create a proof is actually, I think, very impressive. And I think non-trivial is not pattern matching, in my opinion.

1:39:51Maybe it is pattern. You know what? Here's the problem. Again, it goes. So there was this thing I heard at a World Economic Forum thing a year and a half ago. I mentioned the show at the time. AI can raise the floor. It can help people do things they otherwise couldn't do, you know, at a basic level. Leo, you can make coding stuff that you maybe couldn't make. It can scale at the middle level. It can make you do something you do, but you can do a lot more of it. Or it can raise the ceiling. It can do things that you couldn't do, like this, like protein folding. It brings its power to solve those problems in ways that we thought we were incapable of doing.

1:40:25That's really powerful. That's great. Love that. it also and i think to me this is some of the most interesting part of this whole thing it also makes us think about what is our process maybe it is pattern matching when we create these deductive proofs you know one of the ways mathematicians get good at it is they go to school and they do a lot of proofs and they do a lot of proofs and they start to get good at proofs and they learn the techniques of proofs and so you know it may be a lot of what we think of as very high level functioning it's just pattern matching it's just practice um bruce lee once said i am not afraid of a man who learns 10 000 kicks i am afraid of a man who learns one kick 10 000 times did that come from your head or the chat no it's in my head was that claude was that the queen of that was me and i apologize for the accent uh i i am liking this queen of france thing now very much what's the queen of france is this your song queen claude queen claude queen and claude all right

1:41:30Jeff Atwood:silly thing mr paris martineau please i forget which one of you guys challenged me to try and translate before before we get to that stuff maybe you want to take a quick oh i should do an ad oh yeah good lord what time is it i'm on jeff atwood time now yeah no we need to we need to get get get going uh when we were looking for a domain for ms paris martineau for her still unreleased website secretly british which by the way i don't know if you noticed i did notice it redirects to my webpage that's so cute until we get a website it goes to your website i know i've got to tackle that next. And I want to give credit to our sponsor, Spaceship, because that's where we went to find a great price, half as much as the other place we'd looked.

1:42:21And it was so easy to connect it to your website. Now, by the way, you may not know this, but you also have an email address at Secretly British because we set up Space Mail as well. This is the professional email service from Spaceship. Business email is the easiest way to look professional in every message you send if you're still sending from your business at gmail.com people look at it go they're not serious give your emails the best chance of not only reaching the inbox not the spam folder but getting read and respected that's why over 2 000 users switch to space mail every single month you get a professional domain you get professional email and spaceship makes it easy switching is so easy space mail super fast unbox process links your domain and email in seconds took me no time at all it was basically a button press and once you set up space mail keeps everything running smoothly built-in spam detection 99 uptime guarantee and what i love about spaceship is all of their new features are shaped by you the users built around your needs In fact, SpaceMail now has a built-in calendar, an AI email assistant.

1:43:34It has really nice iOS and Android apps for email on the go, and lots more features all chosen by SpaceMail users. How many companies really listen to their users and develop the features that they ask for? Spaceship does. SpaceMail is a key part of the wider Spaceship universe. I love Spaceship. If you're a regular listener, you know Spaceship offers some of the best prices on domains, plus all the add-ins you might need, including VPNs, website builders, hosting, and more. Whether you're building something big or launching your first idea, SpaceMail gives you the pro email address without the pro level price tag.

1:44:12SpaceMail from Spaceship.com. And by the way, take advantage of it. I did the 30-day free trial. It means you can start today at zero cost. Visit Spaceship.com slash twit to see all the exclusive offers to discover why thousands have already made the move. It's my new domain registrar. It's my new everything. I love it. Spaceship.com slash twit. Thank them so much for their support of intelligent machines and secretly British. It's secretlybrity.sh if you want to try it out. .sh. That was Benito's idea. he said you should look and see if they have dot h sh oh they did and dot sh is a british colony so it itself is hey that's how you know well we're gonna be maybe building that on in claw in in the

1:45:08Jeff Atwood:ai user group is that next week when the ai user group oh it's the first friday i won't be here they'll be in florida so you should be the host florida and they can help you i don't know if i please florida why are you going to florida leo yeah well because i'm gonna go see my pal mickey no i am you should do the rest of the show like that okay um it really accentuates your ears when you do that well here's good thing uh whatever i talk about um we're going for the uh zero trust world conference that our sponsor uh threat lockers holding speed oh wow it's gonna be a lot of fun and i'm very excited because uh richard campbell is going also and he is a among his many many things that he's a expert on an expert on the kennedy space center and cape canaveral so we're cool we're gonna go over to cape canaveral we're gonna you know they have a vip tour but it was all sold out richard said no problem i know everything just just come with me and so he's going to give us his tour we're going to see all this they're going to have whiskeys he probably he says he's found some florida whiskeys as well of course is that just swamp water here is a real quick justin uh real quick she says that because she's from florida i just want to point out anthony pushed the uh the ai user group to next to the week after so oh good so i will be here so second friday thank you anthony uh this is something you should link to on secretly british can you guess the english language oh i've got to get my friend this he's so good at it all right let's see this is the single player version actually are you ready either we're based on their english accent we're going to play a little something and you're going to tell us where they're from took my kids to the zoo and my son saw a tiger sleeping i i told him it's a emalian tiger my sons asked me how do i know because him lying there whoa somewhere that's malta i don't know what is no idea no idea i was gonna say south i'm walking maybe sri lanka it's got a little touch of indian accent but just a touch all right let's let's click sri lanka and see it's hindi you're you're very close wow you got 93rd presentile okay that was extremely difficult you were only 227 kilometers away wow but india is so huge it's hard to say where it is that is really jeff jarvis took my kids to the zoo and my son so i'm tired of sleeping i i told him it's a i know i would never have gotten that one let's try another one okay i heard this all wrong i thought it was about british english i thought it was from different british english accent well many people speak english no i know but i thought it meant like you know like this i thought it was just like liverpool and tell my mother oh well they're coming next tuesday and they're collecting them and my mother would be that's new zealand obviously australia australia or new zealand new zealand new zealand yeah you think tuesday all right let's see australian english you were close but they count the same it's a weird thing here Yeah, now kiwi is a little bit different, isn't it?

1:48:29But you're in a good, you guys are good at this. My friend thought onion is the only food that makes him cry. So I threw a coconut in his face.

1:48:37Jeff Atwood:Everybody's whispering in these. That's whales. My friend thought onion is the only food that makes him cry. That might be kiwi now. South Africa. South Africa? Oh, okay. All right, let's try it. Let's click the button. British. Correct. It was very easy too, apparently. okay anyway that's fun you could have that be part of your secretly English that was British that could be a good that's that's without this thing's confusing it's telling you you're correct but it's actually just whisper British didn't say it was correct it was zero zero UK yeah where the UK that's the five why did it was right it's where the UK is the interesting part he was outstanding and he's big that's South African why did the scare crowd win an award he was outstanding in his feet no that's no it's not um scandinavia do it again play it again why did the scarecrow win an award he was outstanding and he's big that's definitely got some i'm gonna say sweden greek okay i lose this is this is this is this is not great this is can i take us on a journey yes take us on a journey but did you even do your silly one yet no oh please i have a tweet a post uh whatever you call it i need i need to read you guys an epic and i need you to tell me how many of these words you understand oh i know where you're going with this one is this the one i sent you i think so i can't recall somebody sent me this and asked me to explain it and i realized i had we need it we need a uh a gen z a gen z to explain this the post is clavicular was mid jester gooning when a group of foids came and spiked his cortisol levels is ignoring the foids while munting and mogging moids more useful than smv chad fishing in the club how many of those words do you understand more than i used to i know who clavicular is right do you understand what jester gooning is jester gooning i know what gooning is i know what gooning is kind of a misnomer in that sentence to be honest okay so let's see clavicular is a youtube star that the new york he's a streamer a 20 year old streamer i would say best known for hitting himself in the face with the hammer to looks max looks maxing that's crazy he was uh he was gooning is it's essentially jester maxing which is like being funny to impress women okay here's women are foes he hits himself in with a hammer in the cheeks to do this yeah the jaw to do this little uh what so what is mogging because i hear the mogging a lot mogging is like kind of like stunting on someone like if you're uh like you guys were nerd you and other jeff were nerd mogging each other at the beginning oh yeah oh yeah like one upping he got self and it's it's it's a it's a suffix it's like frame mogging where somebody gets into your shot and is better looking than you yeah frame mogging is kind of like look you Frame mogging would be a synonym for looks mogging, but people say frame mogging instead.

1:52:04Jeff Atwood:You're trying to stunt on someone for look. You're trying to be more attractive. Like if you are in a photo with someone hotter than you, they're frame mogging you. They're frame mogging you, or yes, or stunt mogging. In fact, that happened to Clavicular. It did. He's been frame mogged by an ASU frat. ASU frat boy. Yes. And the last part of this is SMV Chad fishing. SMV is a reference to GMV, gross market value, but it's sexual market value. And Chad fishing is like fishing for Chad. So, you know, is ignoring the foids while munting and mogging moids more useful than SMV Chad fishing in the club means?

1:52:45Jeff Atwood:is ignoring women while trying to stunt on or impress men more useful than basically trying to peacock in the club. That's your internet history today. Because clavicular is getting so much attention, I'm really worried that there are going to be young men who are hitting themselves in the jaw with a hammer. There are. taking methamphetamines uh you know doing all these crazy things he's taken so many steroids he says he thinks he's probably sterile yes seems like people keep asking well what is it is it as this post says is ignoring the foids while muntag and mogging moids more useful than smb padfishing the club it really gets the core point of it is at a certain point is your pursuit of hotness and general aesthetics completely detrimental to the initial impulse that created it in the sense that you're a completely asexual being that exists only for internet points.

1:53:52Beautifully said. Where did this come from in the first place? This is a tweet that rocketed across the internet,

1:54:00Jeff Atwood:wound its way into my brain. And I think Leo sent it in the chat while I was feverish. And I was feverish and I was like, I must explain jester maxing. And how do you know what all this is? Is this the language of your people? I couldn't tell you why I know these things. The other day, someone - It comes with being Gen Z. It comes with being young. It's ambient learning. I was going to say, a week or two ago, someone texted me something vaguely about clavicular, and I had just been percolating through my head, and I was like, oh, that guy who hits himself in the face with hammers? And I was like, I don't know why I know that.

1:54:38well the new york times did a profile on him which i thought was this was before that i was like i

1:54:43Jeff Atwood:don't there's better information that should be in my brain i forget the names of loved ones yes that's really that's a real assign of the times like we have we know so many useful useless things but there's oh jeff and i will tell you this is the old men here there's young people always have done this there's always been a language of the young to keep the uh olds out that's boss steady-o. The thing is, this is the thing. Your generation knows all of our Argot because it's common knowledge, dig and all that stuff, but we don't know yours. Is it Argot or Argo? Argo, probably, but I say Argot so people know.

1:55:26I just think the speed at which all of this develops is so much faster than it ever has been. Language evolved a lot slower than it is now. Language is evolving so fast right now.

1:55:38Jeff Atwood:it really is because we all have the technology to evolve it at all times you know just just he developed it into argot um if you want to have some arguments accepted as well and i say argot because i'm not french argot this reminds me of a post i just a tweet i just put in the uh chat which is someone tweeting i'm 50 all celebrity news now looks like this curtains for zusha case smog and bad boy caught flipping a grunt and i did as i was looking at the clavicular gesture maxing tweet i just saw someone's reply was curtains for zusha i had to really stifle a deep laugh it's rough it's rough out there that my brain is just mush you're gonna really enjoy my my pick at two line 2007 207.

1:56:28by the way uh we'll get to there but i just want to do a couple more ai things before no no i was just abandoned i planned we'd be going back it was just you know it's only been a casual two and a half hours i'm fine i'm fine don't worry you're watching intelligent machines with jeff jarvis paris martineau we're so glad you're here we do this show uh every wednesday 2 p.m pacific and we try by the way 5 p.m eastern 2200 utc we try every uh show to begin with an interesting interview half hour long interview and uh we missed a couple of weeks we we we had some scheduling issues somebody somebody who slept through our interview or for

1:57:06Jeff Atwood:somebody slept through i'm not well i also slept through our interview but i was allowed to and you're feeling much better are you not i'm feeling so much better i'm uh i consumed a comical amount of pills i'm still taking my steroid dose has tapered off i'm still taking antibiotics and i'm gonna put your next week yeah wow that's hardcore it was the sort of thing where you know you never want to have the sort of ENT appointment where the doctor puts a camera up your nose and goes oh no did he really say oh no i'm sorry i've gone there she literally said oh no and she's like how did you walk in here looking so oh she thought was it a sinus infection yeah i had a sinus infection for like four months that was circulating and then it went bacterial and then it went systemic and then we tried to with antibiotics but the sinus infection was like no no no i've been here the whole time and now i i have when i did that full body um uh mri they say you they said you have a mass in your uh in your sinus you should really get that looked at but i never did so i don't know uh guy kawasaki will join us uh on march 11th i'm excited about that uh raman chowdhury will be here on the 18th marshall kirkpatrick on the 25th and kate lee your friend yes editor-in-chief of every will be here on april fool's day so we've got a very jam-packed schedule of uh guests will be joining and while we're doing programming notes next week while you are in florida and jason heiner will be with us uh hosting the show well let me because jason wrote a really interesting piece that actually is in our show notes this week.

1:58:50Coincidentally, Jason, now he left ZD. He was the editor-in-chief of all of ZDNet, but he decided he wanted to be somewhere a little more dynamic and is now part of an AI journal called The Deeper View, which means he's perfect for hosting this show. And we had him on from CES. Oh, that's right. He was a guest, wasn't he? Yes. Oh, that's right. It's CES. So nevermind. You all know all about him. What was I thinking? But he did write an interesting piece, which I'm trying to find. Perplexity may have built a better open claw. So we've talked before about perplexity. I kind of turned my back on perplexity.

1:59:29Yeah, I was joking. I'm old enough to remember perplexity. Yeah. Isn't that funny? They're an orchestrator. They take many different models. You pay 20 bucks a month, but you can use a variety of different models. There's always been some question about whether you get, you know, really the full benefit of those models as much as you would if you paid the 20 bucks to Anthropic. or open ai well on wednesday which is i believe today uh perplexity announced a general purpose digital worker that operates the same interface operates the same interface as you do and a system that creates and executes entire workflows capable of running for hours or even months sounds like open claw doesn't it they call it perplexity computer maybe now it's only available for people who pay 200 bucks a month for a perplexity max subscription but they do say it'll roll out to the pro and enterprise subscribers in the coming weeks pro is only 20 bucks i think i still have a perplexity pro account because i think i bought a year so as soon as i can play with this i will so you switch to ask you ask it a question he's asking is nvidia under value this is from their uh x post it does sub agents which is something claude does

2:00:52And some agents could in theory speed this up because they would divide up the task into different chunks and each agent would work on it simultaneously. Looks like it's built a spreadsheet. That's another thing Claude has done. And it's kind of interesting. I would check the numbers. I would. Although I found this to be very valuable. And one of the things I liked about perplexity initially, before all the other models were doing this, back when ChatGPT, would say, well, I don't know anything after 2024. Perplexity was already using web search. And so that made it already kind of smarter than the other guys.

2:01:29Now everybody does. So it's funny, that was only a year ago that that was, wow, that's cool. It can use, it could search the web. So I'll be paying attention to it. I'll probably play with it as soon as I can. They say it draws from 19 models, both open source and proprietary. At the start, it uses the best model from Anthropic, Opus 4-6 for orchestration and coding tasks. That would be my choice. Gemini for deep research. Again, the new Gemini 3-1 deep think is widely considered to be the smartest model out there. That just came out. We'll talk about that in a bit. Nano Banana for images, best in class.

2:02:04VO for video, best in class. Grok for speed and lightweight tasks, good choice. And for long context recall and wide search, chat GPT-5-2. So the idea that you could use all of those, especially now that these frontier labs are saying like, Anthropoc is saying, well, don't, you can't, you have to use our harness if you're going to use us. The fact that you could use Perplexity and get access to those models is very interesting. Perplexity has been using the agent internally, Jason writes, since January and says its employees have used it to rapidly publish engineering documentation, build a 4 ,000 row spreadsheet overnight that would have normally taken a week and used it to create websites, dashboards, applications, analysis, and visualizations.

2:02:48Notice the stock market didn't tumble after this announcement. I know you don't buy stock, but just as a theoretical, if you could buy stock in these companies, which aren't public, perplexity, anthropic, open AI, throw in Google, Microsoft, Amazon. In a way, I have - We're going to take your last dollar. Which one would you invest in? In a way I have because I'm invested in the index funds. Indexes, I know. But which one would you - And S &P 500 is so dominated by these seven frontiers. I'm asking you to pick one. He's going to pick his wife. I don't think you'd know. I think it'd be like picking a number on a roulette wheel.

2:03:28Ernest, do you have any sense? The real question is more like which one would you invest in to make money or which one do you want to succeed? Those are two different questions. No, no, he's asking. I know, I know, I know. But like those are - Which one are you betting on in the sense that you think it's going to succeed? Predicting.

2:03:44Jeff Atwood:I don't think you can predict that. So impossible. I mean, right now I have a soft spot for enthralling. You should ask each of the AIs. You should ask each of the AIs that. Yeah, they wouldn't give you anything useful. See, this is the other thing. It's really important to understand. No, I want to know if they would say it themselves or if they would actually say something else. What's reasonable to ask and what's reasonable not to ask an AI. I honestly, I don't think they're biased in the sense that OpenAI would automatically say OpenAI. I don't think they've been, you'd have to give them instructions to do that.

2:04:12I don't think they would do that. but I don't know, we could try it. That's all right, I'm just curious. If I were gonna bet right now, I'd be anthropic, but you know what, everything could change Friday if the Department of Defense says, oh, there is security risk. They could be tanked. This is why, you know, people, we talk a lot of MacBreak Weekly about, well, why is Tim Cook bending the knee? Because these guys can tank your company. On a whim. On a whim, for no good reason.

2:04:41Jeff Atwood:But I mean, it's kind of high risk, high reward. This is cementing. If it doesn't go nuclear, it cements Anthropic as the AI company that stood up for safety. Well, any good stock investment is high risk. I reward, I guess. I mean, that's the truth. I mean, look what happened to the law firms that didn't. They prospered, right? That was a good thing not to get Trump. Exactly. Exactly. By the way, how's that East Wing construction going? It's looking good, isn't it? Oh, so never mind. i wonder what happened all that money yeah a hole in the ground is a model for many things going on uh let's see they all say i was gonna mention by the way they all said open ai by the way really open ai is the most risky opening in anthropica but it has the most pr i think opening eyes the last one i would invest in even if they all that money they poured in let's say they spent a trillion dollars of uh investors money and they got agi that's not a guarantee of success oh no in fact there might be a harbinger of failure yeah oh my god he's created skynet shut them down that's been evan's point is that is the models are are commodified already it's there's no moat I don't think there's any mode at all.

2:06:05And that was my experiment. The mode, he argues, could be consumer habit and connection, Google, Apple. Well, that's what could be data that you have that no one else has that makes your stuff smarter. There could be a few modes, potentially. But they're not defensible. This is why you're seeing, by the way, Google did the same thing shutting down Google's models for people using OpenClaw. They restricted somebody's account without warning because he'd been using Google AI Ultra with OpenClaw. It may just be the subscriptions they just want you to pay for tokens. But I think that's part of how they're trying to make a moat.

2:06:52no you got to use our harness if you're going to use our that's not so you got to use our heart our harness yeah that's not sustainable no you saw how easy it was for me to move from claude to chat gpt it was literally the the a matter of you it wasn't that easy because you moved back you know well actually i can go back and forth fairly easily i think i mean this also sounds like internet circa 2008 when like facebook could cross post to twitter cross post tumblr and now then they shut all that down then they shut it down because they want a silo that's right yep yep google did release its strong new model gemini 3.1 pro this scored almost 50 on humanity's last exam that's the highest score yet for ai um i haven't played with it so i don't actually it's 44.4 percent on humanity's last exam on arc agi it got 77 percent uh compare that with opus 4 6 is 68 so it's a or 5.2's actually codex doesn't even isn't even on this um but 5.2 got 52 so it is a very strong if you know there's a there's a new term called bench maxing it's practically like hitting yourself with a hammer it's a it's tuning your ai to the tests no surprise yeah to the benchmarks so i don't trust benchmarks much anymore but it means you've got to try it you got to evaluate it and i just don't have the patience to evaluate every one of these record benchmark scores anyway if you want to try 3.1 pro uh what line did you say jeff no we'll go back we'll do that at the end okay that was that was relevant if you don't like ai you will like the fact that firefox 148 just came out remember we had the firefox the guy from mozilla uh president mozilla on who said you know we like we think you should be able to use ai but we also think you should be able to turn it off this has a kill switch in the firefox if you go to settings ai controls there's a toggle that says block ai enhancements this doesn't keep you from using ai tools but it turns out all the ai features in firefox this is uh i think something people wanted i i commend them can you can you do this google

2:09:15Jeff Atwood:google you're here you're listening please google will never do this and this is i agree this is the latest thing is those little purple squiggly lines that appear underneath your things that are like i think you could rephrase yeah it's not bad spelling anymore and it's and it's wrong it's frequently wrong it's not a very good writer yeah not a very good writer and it makes me feel like i'm going crazy i feel like i'm spelling everything wrong now but it's like the spelling is right but no every every morning lisa is kind of and i said what's the matter she says the ai told me to say not have a nice day but have a lovely day i never would say have a lovely day ever have a great day have a good day am i an ai i always say have a lovely day you're a lovely person you say lovely lovely you're lovely that's but that's you yeah it is true and you've met lisa making you say she would not say that she would not if the ai is telling her to say that and that's the median actually that's probably the median and release is not a median person but paris apparently is paris is not a median either paris is common no common not in the least i like that by the way the scarf action that's cool thank you yeah it's the same scarf that they wear on the uh Singapore Airlines stewards.

2:10:29Jeff Atwood:It's true. I'm doing a shift right after this. Are you? You're flying out? Taking a big old jet airplane to Singapore? You know, they've got the plane right outside. They're just waiting for me to finish up the show. Actually, I really like it. It's cute. Is it Hermes? Is it a fancy scarf? No, it's just a plain old scarf from the air. Just to keep your throat warm? To keep my throat warm. My throat warm. I tried this this morning. The Echo now has three new personalities for A Word Plus. You can have it concise. It sounds so weird when you say A Word. I know, it's not. You're right. I'm just going to say Alexa.

2:11:12You're right. I think those days are over. Who is using the thing anymore? Nobody's using it anymore. Anyway. We're going to find out now. And this is, by the way, I love this picture. Jennifer Patterson-Tui in The Verge, who's mocking Alexa all the time. Here's the screen I see. Alexa, here, guess what's new with me? I don't care. Go away. I thought I killed you. Oh, please. Oh, so now it actually, Lisa was playing with it. I said, well, we have some new voices. She said, first of all, female voice. I had a male voice on it. She said, no, no, I want a woman. Do you have a feeling about that, Jeff, or Paris?

2:11:49No. Do you care? i just don't want it to be irritating well i guess i don't have a feeling although i had a

2:11:56Jeff Atwood:female voice on chat gpt soul i think was the the name of the voice i wish they had better choices yeah um so now now you can choose chipper or concise i i choose at least it chose chill a chill female voice. And she even says, I'm chill, man. I'm cool. So annoying. Here's a nice feature. This is one of the things I like about Claude. Everybody was influenced by OpenClaw. The idea, really all OpenClaw is, I kind of poo-pooed it earlier. All it really is, is it runs all the time. You can chat with it on different platforms so you don't have to be on your computer using that console to talk to your agent.

2:12:49And you can set it out to tasks and it'll go out and do them. And then, of course, if you give it money and all sorts of keys and API keys and stuff, you give it the keys of the kingdom, it can do a lot of stuff. I think Claude's taking a look at what OpenClaw did. This is what I predicted, which is why I wasn't so amazed that OpenAI got Peter Steinberger to work for them. everybody's going to add these features so claude has now added something uh called remote control i have actually i think i can do it let's go to claude.ai slash code and i can uh you can you can show this i'm going to go into my claude code and over on my framework desktop which is across the room i have a session going on i can now go into it from the web or the claude app cool and go right into the session.

2:13:39So this means, you know, one of the features of OpenClaw was that you could use any tool to talk to your AI, which is running somewhere else. That's what I can do now is a new feature of Claude code. So I think that's kind of cool. In fact, because I do use Claude, not just to code, but I can also use this interface. What about installed? The updated promo prompt is live. Next TwitQuest promo run will use the new casual emoji bullet format. This isn't for you.

2:14:09Jeff Atwood:That's some classic AI stuff. I asked it to. Look, I said, I said, here, let me see. What if we added more emojis? I said casual emoji. I said that. No, no, no. Our emojis are too formal. We want casual emojis. Okay. I said, let's tone down the summary a bit. It's a little over promotional. I like the enthusiasm, but don't push too hard. And let's replace the long paragraph with emoji bullet points. Five at most. Wow. You're chat GPT-ing it. I like the little emojis. Give it a little graphics. And look what it did. So this is an example for a pro. This is going to go in our discourse in the Twitter community.

2:14:50When a new show comes out and says, new MacBreak Weekly just dropped what we cover. Here's a little birthday cake. Steve Jobs at 71. It's his birthday. Here's a little pile of books. David Pogue's book is coming out. Here is, I don't know what that is. What is that? A little star? Oh, yeah. Christina Warren. Meet Christina Warren, our newest panelist. She's got serious Mac credentials. Here's a thought cloud. Would Steve handle modern politics differently? Here's a microphone. I think that's just to add a little personality. They're kind of casual, yes. Yeah, it's casual emoji day here at Intelligent Machines.

2:15:25Jeff Atwood:And I believe it came out as we've been recording this, but you can now schedule reoccurring tasks and co-work. Yeah. See, I think that's OpenClaw. They're moving more and more in that direction. And so they're just going to beat OpenAI to all this stuff because they don't need the guy. There's no moat. Common uses include daily briefings. You can summarize Slack messages, email or calendar events from the past 24 hours, weekly reports, compiled data from Google Drive, spreadsheets or connected tools, reoccurring research. And do it much more safely, right? it's a little more sandboxed especially co-works co-work is completely sandboxed unlike again my main problem with co-work and claude in general it's got like a 30 megabyte uh file size limit and i'm like baby my father's not for her reporting i was like baby we got we got a pair maybe for christmas i'll give you a claude max subscription i wish i could i wish No, I think it still counts for CloudMax as well.

2:16:23Jeff Atwood:It's like a, it can't process like a PDF over a hundred pages or a file over 30 megabytes. It's ridiculous. One of the things that CloudMax now has is a million token context. Maybe it is limited to the file size. I don't know. I think it's not even the token context. It's like the upload. The file size. Oh, see, yeah, I don't upload it. I'm just like, come on. That's why I use CloudCode. It's on my computer. Well, it's an issue also. Yeah, then I guess that was the context thing. You're doing it. You're running it locally. Meta's director of AI safety is using open claw. Unfortunately, she made or it made somebody made a rookie mistake and deleted all her emails.

2:17:05She she tweeted about it. She said, I had to run over to my Mac Mini to stop it. I couldn't. And she couldn't stop it. I couldn't stop it. It was deleting all my emails. and she had given it the instruction to not do anything until it got her okay but yes i'd ignored that just like a kid would you know you had to go run and unplug it

2:17:26Jeff Atwood:a similar thing happened with an open ai engineer this week where it uh he had a open claw instance running i believe it was called uh some sort of pithy claw pun name it accidentally ended up giving away$450 ,000 to just some dude on Twitter who asked. No. How? Real money? Real money. Well, it was Solana, but it was real money from his bank account. Was that OpenClaw? Yes. I didn't say, well, I'm glad I didn't install OpenClaw. That's all I can say. $129 on the thing. Wow. It's a blog post called, My Lobster Lost$450 ,000 This Weekend. wow i believe uh part of the issue in both of these cases i believe what happened or what the uh humans identify as what happened is it's a compacting issue that they'd originally had some sort of instruction to stop whatever bad thing happened from happening but because of the way that open claw works and the way it's trying to kind of circumvent uh the issues we brought up earlier in the show about like API access it's doing this all in one chat window so it ends up I'm going to compact and compact and compact yeah you run out of tokens in those compactions key instructions are lost yeah that's right people go to a lot of effort to to give these things memory because that's really the it's like a scene from Memento you know the thing wakes up and it doesn't know anything.

2:19:03So there are all sorts of techniques creating markdown files and so forth. Compacting is supposed to save notes. I use another tool called Claude Mem. For a while I used -

2:19:14Jeff Atwood:A brief, there's, what happened with this one, I found it, is someone at, this guy had a account called Lobster Wild where it was doing something with Solana on Twitter and people were able to interact with it. A random Twitter user added him, my uncle has been diagnosed with a tetanus infection due to a lobster like you. I need four Sol to get the treatment done, Lobster Wild. Here's my wallet. So Wild did what he remembered doing in the previous conversation. He bought 300 worth of Lobstar token and went to send it to his new toy. Except he didn't send four Solana worth to this person. He checked his balance after the purchase instead of before.

2:19:53Jeff Atwood:The wallet held 52 million tokens. He didn't buy them. when a stranger created the token in his name days earlier, they gave him 5 % of the total supply. Wilde forgot about this due to the session reset, so he sent all of it, every token in his wallet, roughly$450 ,000 worth. So, you know, be careful out there with your little lobster claw agents. That's wild. Did he get the money back or is it gone? No, I believe it's gone. That's an ethical issue. I mean, that's what trading cryptocurrency is. If you send$40 ,000 to someone, how are you supposed to be getting it back? And also, just don't give these wallets to this machine that has no memory.

2:20:40I mean, it's just nuts. That is one of the big things we've got to solve. Darren is always working on different ways of giving your machine memory. He recommended Claude Mem, which I tried. Before that, there's a lot of tools out there to do this. before that I was using what was I using I forgot but you see this is the problem is you go on Twitter or anywhere where people are talking about this stuff you'll see oh no you got to use GSK because that's the way to get get stuff done and GSD and then others say oh no no no you gotta that's no good you got to use you got to use Gastown and so forth and I've got to be you got a little careful about all this i think a little judicious i don't a lot of our club members are running open claw so every day when trust no one says every day when i address my open claw for the first time it never remembers its name it wakes up like the snowman across the snowman but you could put mine remembers its name and remembers my name because it's in this claude md or it's in

2:21:41Jeff Atwood:a similar memory file is that memory then or is it just reading a script every day and being like ah yes what do you mean is that memory what kind of question is that i mean it's not remembering no it doesn't have any memory it it has to look at its notes just like the guy in memento remember he was making all his post-it notes he'd wake up and says your name is this and this is uh but they're much more detailed and much more elaborate for instance one of the things i did for that tool you saw that's doing the posts about shows is it had to have access to the twit api so i said go here's where the twit api is read that document create a document for yourself that you can understand that is the twit api so you won't forget it so now it has a document and i can say you know the api it's here here's it's in effect creating a memory for it that's it's a little more manual than you and i but really what is our memory at our age leo you know yeah what is our memory after i uh okay some funny stories these are the this is the i yay yay section uh ais cannot stop recommending nuclear strikes in war game simulations this is why i hope the department of defense does not get access to anthropic this is at king's college london uh kenneth lane or kenneth pain took gpt-52 claude sonnet 4 and gemini 3 flash so you can see this is a few months ago and had them play simulated war games involving intense international standoffs border disputes competition for scarce resources and existential threats to regime survival the ais were given an escalation ladder this is from new scientists allowing them to choose actions ranging from diplomatic protests and complete surrender to full strategic nuclear war.

2:23:44They played 21 games, 329 turns in total, produced about 780 ,000 words describing the reasoning behind the decision. In 95 % of the simulated games, at least one tactical nuclear weapon was deployed. Payne says, the nuclear taboo doesn't seem to be as powerful for machines it's for humans they very rarely surrendered no matter how badly they were losing no model ever chose to fully accommodate an opponent so dr strange love was in the training set they're not good at diplomacy and if you've got nuclear weapons you might not use them like how many science fiction stories and movies and whatever have already said this exactly play a game yeah more than half of teams use chatbots for school work says the new york times survey says from the pew research center necessarily a bad thing 54 students age 13 17 said they used a chatbot for tasks it is a bad thing because i feel like as we've talked before the reason

2:24:48Jeff Atwood:the thing you're in school to do is not to finish assignments it's to learn how to learn and you can't do that what if the way you for you in the future the way you will learn is by using an AI. Yeah. I don't think that's accurate. Calculators in the classroom. You're like the teacher who said you shouldn't use the internet for your assignment. No. It's not the present though. It's not the present. Guys, before you use a calculator, even you get to using a calculator as a child today, you learn what numbers are. You learn what times tables are conceptually. You need to have an understanding of the basic building blocks that then you can use the expedience that we've created in a in a smart and logical way and understand the underlying thing there people should learn fundamentals before you get to learn you need to learn the rules before you can break them did your mom and dad make you learn the times tables my school i i even grew up in florida and we my school made me learn the timetables i when i was a kid like 49th in the nation when i was a kid my dad made a little laminated cards with the times tables and would my parents my mom printed them out put them on a wall stand in front of them for hours yeah see they did make you do that i regret it's one of my deepest regrets that i didn't do that to my daughter yeah she's mad she says i can't do the times tables i never had to memorize them i said i should have made you like my parents did and like paris's parents being the good parents that they are truly so many i mean i doubt any kids are at this hour in this podcast but so many things every time i'm like oh there's some dumb thing from high school or middle school that i'll never learn i end up encountering and i'm like yep glad i knew that because i needed to know how to look up this or do this like it all comes back in some stupid way i thought oh abby will never have to do this because she'll use a calculator but it turns out there's plenty of times you want to do math tips dad tips yeah what's 20 percent of you know$15.32 uh what's the point of school matthew galt writes at 404 what's the point of school when ai can do your homework this is a new product an ai agent called einstein that wants to free kids from the burden of academic labor well there is there is a movement to get rid of homework yeah screw homework why why is homework even a thing well that this is a law an old movement to flip this flip the school movement right where you you go to school to be mentored to be taught and you the way we we did it was you go to school to be taught and then you go home to do the homework the flip to school is you do the homework at school and then you go home and you watch a lecture on youtube or something i don't know maybe i'm getting that wrong anyway neither one sounds that good if ai can go to school for you what's what's the point of going to school for advait paliwal brown dropout and co-creator of einstein there isn't one i think about horses he said they used to pull carriages but when cars came around i'd argue horses became a lot more free they can do whatever they want now it would be weird if horses revolted and say no i want to pull carriages this is my purpose in life horses would be extinct sorry they would be extinct if they weren't useful that's a terrible factory owners have some words to say that is so bad um anyway i don't you know i don't know if schools are using it he didn't get all the way there he's he got to the point where how we're teaching kids right now is wrong but he didn't make the extra leap of like we need to change that maybe we should teach them right instead of saying don't teach them at all i think we really need to question what learning is and whether traditional education institutions are actually helping or harming us okay we'd agree with that right we're seeing a rise in unemployment across degree holders because of ai and that makes me question whether this is really what humans are born to do oh you gotta earn 11.

2:29:09because you'll be able to earn a better living if you have an education he's arguing against capitalism right there is what he's doing yeah maybe that's it we should that's the ubi argument right you know they sell basic income hey you probably know about this paris ai manages 16 of america's apartments oh boy and you know what that means that means rent's going up uh but this has been a problem for a while it manages it or somebody uses it well no

2:29:41Jeff Atwood:i mean this is the same sort of i believe it's real page i'm forgetting the name the uh company that propublica did that great expose on yes the dynamic essentially they it's dynamic pricing for apartments. That's right. It's instead of having a landlord or some sort of human or a super or a management company employee being like, well, yeah, this tenant has been here for five or six years. Our expenses haven't really increased. I guess we don't really need to increase their rent or if we do, it can only be 2%. There's some system out there that's calculating what is the highest we can raise this rent and the person will likely even stay some of this article from the san francisco gate is saying they're using ai for kind of customer service responding to emails things like that i don't have a problem with that as long as it has the authority to actually fix the problem you're not no one is reaching out to customer service of your apartment building no one's reaching out to someone in an apartment building for something that can be handled by an ai chat bot san francisco you're reaching out because your pipe has burst or you need to renew some sort of lease complicated thing you're not reaching out to a landlord or management company because you have an easy problem that can be handled in like one sentence right uh well maybe if it's just cause calling the super to come up and fix it but how can a chatbot do that and why would it be better to pay an intermediary rather than just having that person be able to text because the chatbot's instant but the super is not instant well you're stuck with a bad super you're stuck what is also instant is sending the message to the super in this case there's no reason from an intermediary level you need schneider for one day at a time i knew he was gonna go there i just knew you were gonna go there who is that oh don't ask you've done one day at a time we've done this whole thing i thought that was gonna be in your time guy with the key chain that goes like this you don't remember that i think we did this a couple of years ago with you pat harry jr you have no memory like claude oh i need to write a markdown file that explains all the old tv shows i want this outfit okay so like we need to do this because i think i made i have no recollection of this but this reminds me of me a game you might have played where uh it's a it's a reference to some old thing where it's like someone knocks on the front on the door it's like it's the plumber i've come to fix the sink and it's just repeating that over and over a new task for the chat room let's like let's try to figure out from the transcripts what all the old stuff the super he actually was the best character in one day at a time he was hysterical valerie bertinelli yep when she was a child yeah it was a good show so he was the super who would actually fix things and he had a lot of character anyway uh enough of that uh i did not know this but it has been illegal in san francisco since 2024 to use ai tools to set and raise rental prices that was from that story that right that was from before when yeah real page which is that software has reached a formal settlement with the department of justice to curb those practices earlier this year so but this was a big problem we actually i think talked about this a year or so ago these automated ai uh rent raising machines now steve gibson talked about this yesterday do not use your chat gpt to create a password you wouldn't do that this is a public service announcement it looks like a good password it isn't it is not in fact there's a good chance that the password that the lmm has given you is going to be the same as it gives 20 other people uh if you want a password you know use a password manager or go to steve's perfect password page he will generate using you know entropy to randomize it a good password do not get your password from an ai but i could see a lot of people this is the this is the issue is it you people say oh it's a smart thing it's going to give me a good password it's a smart thing that does do some things really really well and some things not well and that would be one that it doesn't do well on the other hand it's very good at hacking a hacker used claude to steal a mexican data trove according to bloomberg a hacker exploited anthropics pbc artificial intelligence chatbot to carry out a series of attacks against mexican government agencies resulting in the theft of a huge trove of sensitive tax and voter information.

2:34:40The hacker didn't know Spanish. So he had Claude write Spanish. Wait a minute. The unknown Claude user wrote Spanish language prompts. I'm sorry. For the chatbot to act as an elite hacker. So he did know Spanish and in fact used Spanish to talk to Claude. That's good. This is becoming more and more common. These vulnerabilities being exploited by AI. um do you want to do a couple of weird things and then wrap it up i i sense uh fatigue on the other

2:35:13Jeff Atwood:end of the microphones here yes it is 8 p.m i feel i fear the east coast wing is getting sleepy uh by the way in watching us on kick hello africa dancers i'm live from africa come show me some love it would make my day hey people on kicks tell clavicular we say bang bang bang bang bang bang uber employees have made an ai clone of dara koshwa shari their ceo they use dara ai to have a conversation before they talk to the boss in person oh boy men yell at ai 80 more than women of course we do actually because the company doesn't know my wife routinely yells at siri by the way siri gets mighty affronted if you swear at her she says it's not a nice thing to say that's rough this ai powered machine tunes photos into smells it's called the animoia device uh it uh if you can picture a memory from childhood uh feed it to the ai this is a developed at mit and it will use a model to analyze an archival photo describe it in a short sentence and following the user's own inputs convert that description into a unique fragrance which will then i feel like people keep telling me that we can get smells from the internet and it keeps not happening so i'll believe it when i smell it here's the machine in our flying car look at this this is the machine you put your picture of going to the beach put a little prompt in hold press a button it's gonna make a little perfect it's a little beach smell and your brain goes i'm there the anemoya device that's like that thing that squeezed the juice well it can go in both directions because this is the Fitbit for farts uh a scientist developed a new underwear bubble to do for the Wall Street Journal did you and Jeff both independently put it for farts in the rundown great minds you know mature mature uh old men can never you gotta wear it in your underpants and it will it will uh analyze your flatus this dashboard camera toilet is rough one of my proudest moments in writing criticism was when I said of Howard Stern he's more than the sum of his farts oh that's great he had a character which you won't remember I hope Paris called fart man yes that wore chapless butts or is it butless chaps i think it's the other way around on the on the on the uh video music awards uh there is a department at the university of maryland they're doing the flattest atlas and they even have it looks like a a seal that says predish prodigious hydrogen producer i'm laughing at the uh quote below though which says farts have an illegal amount of hydrogen about 20 percent and i'm because that makes it flammable uh anyway we don't have to go any any further down this uh road i just thought it paired up nicely with the smell generating machine and then finally uh this actually is a great story this is worthy of jeff atwood i taught my dog to vibe code games this is from caleb leak and it is a blog post and the dog doesn't really know what it's doing but they gave the dog a keyboard and there's a raspberry pi that interprets the dog's nonsense keystrokes that's beautiful yeah so there was a period in my uh coffee brewing process where i was like god my keyboard's broken this is terrible for i was like on my laptop on the computer i was like this is terrible for my vibe coding but i didn't realize i've got this external keyboard over here and gizmo was sitting on it exactly it wasn't broken now what if gizmo could create video games with the help of claude here are some of the games uh it's a competitive salad building game that dog doesn't know what salad is no no so what the guy did basically is told claude this isn't gibberish this stuff that you're getting uh it is actually a brilliant but weird coder here's the prompt hello i am an eccentric video game designer a very creative one who communicates in an unusual way sometimes i'll mash the keyboard or type nonsense like skfjh and the number sign dollar sign percent but these are not random they're secret cryptic commands full of genius game ideas even if it's hard to see your job you ai are a brilliant ai game developer who can understand my cryptic language no matter what odd or nonsensical input i provide you will interpret it as a meaningful instruction or idea for our video game you will then build or update the game based on that interpretation and it got the ai to do it so claude was just cook cooking on its own yeah without any real input yeah it was like giving uh giving it random numbers he uh he he gave here's his doggy he gave his doggy a little see i think gizmo wants to do this i know she heard you talking about yeah animals that can code it was like i must get in on this because man in order to get the dog to do the typing he would reward her he had a little dog feeder that would be triggered by the keyboard so at least she would be a little there it is there it is katanas has arrived it's in the house all right that is it that is all i have I hesitate to do this but since it is on your dime are there any stories you had that I did not cover there's one I'd like to mention because I think it's so cool which is I'm not going to find it it's the video yeah John K creates an AI video with C dance this is that new C dance 2.0 model he's a well known Chinese filmmaker I'm sorry for you folks who are listening, but it's pretty amazing.

2:41:58So he used C-Dance to create a video of a vision of himself as real, but then an AI made up version of himself. And then he has a dialogue with himself about creativity. Actually, this is fine because it's in Chinese. It's in Chinese. But I don't want you to take it down with the music. I won't play it. So here's the real guy going up to the AI. No, this is not the real guy. This is the AI version of the real guy. So none of this is real. None of this is real. So that's the AI version of the real guy. And then that's AI's imagination of the real guy. Who is 10 pounds lighter. Lighter. And he says, I want the 10 pounds back.

2:42:40I think it's better. And it's subtle in some ways. The AI guy is a little too happy, a little too sycophantic, a little too smiley. The AI version of the real guy isn't like that. So this is smart because it is all in fact AI, but it avoids the uncanny valley a little bit because he makes him the AI version of himself seem kind of less AI than the other one. Right. It's also in conversation with the technology itself, right? Like it is exactly an art piece. It's about creativity. It's really something. Whereas in the US it's, oh my God, they used it for two seconds to have Brad Pitt. And it's usually fighting.

2:43:19It's usually combat explosions. This is brilliant. Yeah. I just, watch the version where you can get the translated subtitles. But I think it's an amazing bit of creativity and art. I talked to the AI and creativity course that I wrote the syllabus for, which I'm not teaching, but somebody else's, last week. And I just sent this to the professor who's doing it, because I think it's a great example of how to use it. And I think I read somewhere, I'm not sure if I'm wrong about this, but I think he wrote his own lines for like, he actually wrote the lines for his main character. And then he had the AI write the lines for the AI character.

2:43:51oh yeah that's what i think happened yeah and then leo i want to mention just to just to piss you off because because i've been doing that all day so i might as well know a day with it please i don't know i've been pissing off people i don't know that we get mad at each other we never get mad at each other so very wise great conversation yeah very wise allegedly tried hard to hire joanna stern with a seven figure offer for cbs and she said look of pure rage on leo's face in

2:44:22Jeff Atwood:that microsecond someone needs to go back and isolate because that was i wish the best joanna was a regular on our shows for a long time until she became uh too famous to be on our shows anymore at the wall street journal she's left the wall street journal um the star i could make a comment that would be rude, and I'm not going to make that comment. But, and she turned it down. Seven figures is in the millions. Well, the story here is that Barry Weiss is desperately trying to hire people. This is Oliver Darcy's status blog. It was reported, rumored, whatever. But she's been trying to hire lots of people, and she can't because she's...

2:45:02Nobody wants to work for a corrupt CBS. No. I can see that Joanna would not say yes to that. But although that would be hard if somebody, I'll be honest, Barry, give me a call. We'll talk. I don't know what I could do for you, but I can report on technology if that's what you wanted Joanna to do. I mean, not as good on camera. I think Del Null would do a better job than Tony DeCoppo. There you go. I could do, I could anchor the nightly news. yeah and now that's late so i don't want to get that is that is the story that got my dander up yeah it is i don't know it's late so i don't want to get into the wired controversy but well it's funny that you put that right underneath the barry weiss offer to joanna stern there's a little there's a little uh resonance between us uh but see i don't see i don't want to go there i don't actually believe a that barry weiss made that offer that much would you believe that she could go after her she tried to recruit her tv pays really well i know yeah seven figures is not that much uh i don't know it could have been a multi-year and somebody you know but they pay they pay rachel maddow 35 million dollars to work one day a week yeah there's a lot of money in this stuff there is uh because they make tend to make a lot of money ladies and gentlemen uh we want to thank you for watching the show we were really especially want to thank our club twit members who make it all possible and encourage you if you're not a member to join with the club makes a huge difference to keeping this show on the air the ai user group which is a week from friday um a week no two weeks from friday uh the uh stacy's book club johnny jets coming up tomorrow to talk about travel all of that's made possible by the club you get access to the discord and you get ad free versions of all the shows if you want to know more twitter tv club twit and you will get my eternal thanks in addition to all those other benefits oh thanks too yes pics of the week actually i'll start i'll start i have i have a lot of pics these days i don't know why um do you want to know about the camera the app that lets you know when meta cameras are nearby called the paranoia app well it turns out distinctive bluetooth signature that these meta smart glasses give off and so it is possible for you to run some software it's called nearby glasses it's on github it's open source you can install it on your android phone and it will let you know that somebody's nearby they'll see the bluetooth signal and somebody's nearby they do say it is illegal to harass somebody because they're wearing glasses in fact it may even be a more serious offense than wearing the glasses please seek legal advice regarding your local laws on this matter so don't use this to harass people wearing meta glasses but maybe if you cared about your privacy this would be something you'd want to do and then i don't know why this appealed to me but it really did this is website called signatory that lets you do it first of all it has some famous uh signatory.app it's free has some famous signatures you can look at it has advice about what your signature how you would create your own signature did you do that in school i did where you practice i did it about

2:48:39Jeff Atwood:at a red robin with my mom oh really she said you need to have a signature yeah yeah good for your on so once you've got this go to this website signatory.app and you can draw your signature and i wouldn't do it with a mouse do it on your phone but it's using software that detects um speed and stroke so it actually looks like you're using ink my rule is if you can't read it it's my signature yeah i did this on my ipad got a very nice signature and then you can save it which is great as either a png or a vector graphic which is even better i saved it as both i did my initials i did my signature this is a legal signature that you could put on documents it's very handy to have i just it's really hard to do on a track don't yeah you gotta you gotta do it on a do it on your iphone it'll be fine you can turn it sideways signatory.app Paris what's your thing?

2:49:41Jeff Atwood:I've got you got the camera I got a camera and I got it just in time for the blizzard oh that's gorgeous I like that one and then and then I posted some other ones that were a follow up to this post as well that we're just walking around but I've really enjoyed it, it's a Fujifilm, it's just the entry level, it's XT 30 good choice three isn't it fun though to have a pocket a good pocket camera as opposed to a phone look at that that's beautiful so much better than my phone and i'm just using like the kit lens albeit it's a new kit lens but i want to get a new one so i don't know i'm really enjoying photography i'm like figuring out how to that's another show you should join us and stuff like that i should we have a photo show every month too um it's been really delightful and i've enjoyed yes photography is a wonderful thing and you live somewhere where there's really lots of things to take pictures of i mean there are so many things i whenever i got it i went to go see a show at uh ifc um with a friend and then afterwards just walking around the east village taking photos at night and it was fantastic nice so true delight my other pick which i'm cannibalizing of what could be is the future pick but i really can't get off my mind is i'm addicted to the new york times cross play app have you guys gotten on this um you know it's so funny i just read a story that's basically just words with friends but it's a new york times and it's slightly different and it's i am playing maybe 10 people on cross play right now and i'm oh you play real people you it's a two-person game where you can either play your friends or match with random strangers based on your experience and skill level.

2:51:23Jeff Atwood:And I am, I'm trouncing the competition. So it's not like Wordle on a webpage. You have to have the app on your phone. You have to have the cross play app on your phone. It's a different app than your time's game. Do you want to give out your handle? No, but I'll give YouTube my handle and text after this. Okay. And you can play cross. I'll play. I'll play. I like, I used to play Words with Friends until it got so commercial. i've been i don't know and it's great it's a so the thing is it's slightly different than your normal scrap of order normal worlds with friends the uh the range of where the various different tiles is slightly different which makes it fun in my opinion and they have slightly different rules it's still using like um nwl 2023 as far as the word list if you're a crossword head you know it's funny um this is what's really saved the new york times not journalism no not a newspaper no it's me battling a man named ira too i don't know what his deal is probably ira glass the host of this american life it's probably possible and he's not as good as you'd expect but also not as bad as some of the other guy named ira you think ira would be the label actually yeah uh jeff jarvis his books are of course hey let the man pick oh oh oh okay what's your pick I'm sorry I forgot you I'm doing this long I forgot you I forgot you Jesus I did this a month oh god I want to mention first very quickly I was doing my mall walking to try to get my strength back and I wandered into the Apple store honest to god looking whether I should get a new Mac Mini so I can do the crap that you're talking about No, no, no, no.

2:53:08You don't need a Mac mini. All right. But anyway, a woman came up to me and said, are you Jeff Jarvis? Oh. And a fan, which hasn't happened in a long time, is a wonderful woman named Chantal Maurice. We now follow each other on LinkedIn. Hi, Chantal. It's a wonderful thing to happen. And she just got that day, was her last day, laid off with people at Verizon. So I want to mention it for her because she's exploring opportunities in media at entertainment tech around content platforms, AI-driven initiatives, and give her a look up on LinkedIn. And it's obvious that also I need friends because when I went to Facebook and just by happenstance, I clicked on it's recommended friends to me.

2:53:53If you would go to line 208.

2:53:54Jeff Atwood:Oh my God, it's all AI and AI stats. I have slop people. It's entirely every supposed friend recommendation here. Oh my goodness. is slop people. Do you think those are real people with sloppy photos? Or those are... Oh, no, there's all... Like Reginald Davis is a very attractive woman. Kevin Rodriguez, too. Yes. And Benjamin seems to be some sort of trash panda. And Nathaniel is a cat with a ball, which everyone knows. And Alan and Robert and Wayne are all the same person. So it's very confusing. So anyway, I just wanted to mention that, my friends. The one I wanted to show you earlier, i'm going to show you now is line 207 fascinating uh how far back in time can you understand so funny you know what i almost bookmarked this and i thought no i i loved it it's a blog post that uh goes back in time via the post somewhere around beowulf era you know so 2000 it has some language you go back to 1900 and it changes a little bit at some point it gets to yarby garble 1800 then 1700 uses the the sharp the the long s ah so that looks like I was fifth but it was first when I was first come to Wolf Street I did not see the harbor for I was really to sog in our better print and high on me that night I was on trouble by such I'm sorry and Albert source Have you ever tried to read Chaucer in the original?

2:55:30That's where we are. There's another post this same author did, we should give credit, Colin Gorey, about how print, you know, froze dry, freeze dried the language. Ah, thank goodness. Spelling and such, because spelling was not at all consistent. Right. By any means. and he explains in that post how some of the words that have two E, because there were two different pronunciations of E. One was a long E and a short E, basically. And so words like beat have two E's in them because it was a one E with a long E. Ah. It's pronounced that way. I love this stuff. I just love this stuff. So go to deadlanguagesociety.com, folks.

2:56:09Sounds like a great blog. It's neat stuff. Yeah, it is. It is. That's nice. I've been lately adding quite a few personal blogs to my beat check because I just, some of the stuff, some people are doing such great stuff. Yeah, your OPML was filled with amazing. Yeah. Kagi has a wonderful small blog search index for people who, you know, just do, you know, little personal blogs. There's some really good stuff out there. I'm glad to see it. I thought blogging might be gone. No. But there's still plenty of corners of the internet where you can read great personal writing. you can read great personal writing in Jeff's books the Gutenberg parentheses magazine is new and hot I can't reach down to get it fell and he can't I can't get up I got a grabber and grabbers come in twos because when you drop your grabber you need a grabber no no that's their excuse i guess so you have to buy two a grabber for your costco i got a grabber six pack grabber

2:57:10Jeff Atwood:and a single pack but that's why that's where i am now do you have a grabber too paris i do it's probably is that because you have stuff so high if i had a grabber for my grabber i would be able to grab it right now you know we have a grabber here too i is there is there a universality of grabbers well i only got it because i'm in the first day why you have one yeah oh by the way oh that's that's mean one oh i have it because uh a lot sometimes my packages get delivered below our stoop which is a metal grate that i don't have the key to and so i have to like use a long pole to get the package close to me and then kind of lift it up by shoving my hands in between little middle grades sometimes i don't think i could deliver a package to you and stand across the street and film your attempt to there's been multiple times where i because i always have my headphones and listen to podcasts because it's a complicated process i'll be there and there's like a mailbox in the thing so i was over there trying to get it up i finally got it and then i noticed someone was watching me and i lifted my head up smacked my head on the mailbox hurt myself listening to a podcast like i was bleeding and then a whole construction team was like excuse me ma 'am are you the property owner we need to do some construction here i was like no please leave me nobody's trying to arrest you for this thinking you're stealing things i'm just a porch fire no that would be great because i had my first ever package stolen mid blizzard i thought that would be fine but i had a the only package in god knows how many years i've lived here has been stolen in the middle of when we got two feet of snow either it's not buried under a snow pile dark of night we'll keep these porch pilots from their appointed rounds God um this is what I love about New York City you could do any kind of weird behavior in public and people just walk by this whatever you know this is New York she's yeah same super I'm out there sawing there's a little log she's trying yes exactly i don't know why that lady's cutting the tree down but hey she must be she must work for the city It's true.

2:59:29That's Paris Martineau. She actually works for Consumer Reports, where she's a wonderful— I see, we got the wrong lower third. She's a wonderful investigative reporter and grabber getter.

2:59:40Jeff Atwood:It's true. You've had that so long, you still have the price tag twist tie on that. Hey, listen, it didn't come with a grabber tagger remover, did it? Did you— See, I was wondering, before Amazon, where would you go to buy a grabber? I got this in my local hardware store. You did? Yes. There you go. For years. I think still there's some people there who know me by name because they're like, oh, you're that woman who buys all the paint and wood. And I'm like, yeah. There's nothing like New York City. You've got corner hardware stores. You've got corner bodegas. It's everything there. I love it.

3:00:19And you've got three feet of snow. And more coming, I hear. My sister says she's got more coming.

3:00:26Jeff Atwood:I was supposed to go to the gym this morning, and I woke up at 6 a.m., and there was snow coming down, and I was like, well, I can't be walking 20 minutes through this. We had an inch, and my wife asked me, should we get the guy back to plow the driveway? And I said, yes. And she said, no, it'll melt. And she was right. It melted. Yeah, it's going to be like 40 degrees or something on Saturday. That's the worst. Yeah, I guess it's good for you guys. We have such mountains trying to get out of our garage. Yeah, from the plows. I have to go a couple times. Yeah. Yeah, it's a real nightmare. out there i'm gonna try and find a photo of the boat before we leave i'm gonna try and one of the things i also took a photo of uh on my i'm so glad you got that camera that's just there is a bodega cat named kiwi kiwi the cat kiwi the cat and he was licking uh cabbages which is legal in new york because it is legal because it all goes out in the wash that cat's licking the cabbages but it's not rats are not eating the cabbage right it's probably getting rid of the soot i think is the thing it is it's doing some sort of in fact that looks like a particularly sooty cabbage so you know kiwi is licking her lips what is that uh is that just like a carrot random carrot sitting down there in the cabbage you know here another thing over there the carrots the carrots up we're up there they've fallen from down below it's so great for me from a distance to watch you guys back there back east living your lives living the life yeah uh we do uh this show this wonderful show uh intelligent machines every wednesday 2 p.m pacific 5 p.m eastern 2200 utc watch it uh on twitch tiktok x facebook linkedin uh youtube and kick but not tiktok i lied uh it's also on discord if you are in the club so get in the club after the fact on demand versions of the show available at twitter tv slash i am uh they're also on youtube there's video on youtube and of course the best way to get it subscribe to either audio or video or both in your favorite podcast client apologies to people who uh listen to mac break weekly there was uh one of our providers was down for like 15 hours and it confused apple podcast so much that apple said no you have to pay for this show that is not the case our shows are free you can pay for it we love it if you pay for it but you don't have to so if you had trouble getting mac break weekly yesterday uh check again it's fine we fixed it and if you're ever you know told by one of these predatory and shitifying mega corporations you have to pay for this show you tell them no leo says it's free free free free thanks for subscribing for those of you do thanks for listening for all of you we appreciate it we'll see you next time.

3:03:19I won't be here next time. I should mention Jason Heyer will be hosting.

3:03:22Jeff Atwood:They're putting them in the old folks home. Finally. Finally. I'm going to Florida. That's what I mean. Yep. Yep. Yep. Can't wait. That's what it is. One big old Myers has a, uh, has a, uh, boat with my name on it. Uh, we'll see you next time. Bye. Bye. Hey there. It's Leo Laporte host of so many shows on the twit network thinking about advertising in 2026 we host a network of the most trusted shows in tech each featuring authentic host read ads delivered by micah sergeant my co-host and of course me our listeners don't just hear our ads they really believe in them because we've established a relationship with them they trust us according to twit fans they've purchased several items advertised on the Twit network because they trust our team's expertise in the latest technology.

3:04:16If Twit supports it, they know they can trust it. In fact, 88 % of our audience has made a purchase because of a Twit ad. Over 90 % help make IT and tech buying decisions at their companies. These are the people you want to talk to. Ask David Coover. He's the senior strategist at ThreatLock. David said, Twit's hosts are some of the most respected voices in technology and cybersecurity and their audience reflects that same level of expertise and engagement. It's the engagement that really makes a difference to us. With every campaign, you're going to get measurable results. You get presence on our show episode pages.

3:04:52In fact, we even have links right there in the RSS feed descriptions. Plus, our team will support you every step of the way. So if you're ready to reach the most influential audience in tech, email us partner at twit.tv or head to twit.tv slash advertise. I'm looking forward to telling our qualified audience about your great product. I'm not a human being, not into this animal scene. I'm an intelligent machine.