Claude Code, MCP, Free Will, and the Price of Maple Syrup
It's a great start. I'm just gonna be working here while you're doing that. K. You can't hear me. Cool.
Speaker 1:Sounds good.
Speaker 2:Hello. Hey. Hang on. I commanded my wrong speaker.
Speaker 1:Hi, Casey. I know you're so excited that we just did this at the beginning where we're like, can you hear me?
Speaker 2:Alright. So I'm miserable right now because
Speaker 1:Oh, no.
Speaker 2:It's allergy season and I you know what? Give me a second, I need to go take my allergy meds or I'm go crazy.
Speaker 1:Yeah
Speaker 2:yeah yeah. Not only is it allergy season, so I'm allergic to tree pollen. But usually if I just take a pill every night before I go to sleep, I'm okay. But I've been kind of off of it. So I don't feel good right now.
Speaker 2:And on top of that, I stand on my computer today after I rebooted it after an upgrade and everything's in light mode.
Speaker 1:Oh, no.
Speaker 2:And it's like driving me nuts. Like I can't tell what settings flipped or something updated and like whatever setting I was using isn't recognized. So everything's really bright and it's like making my eyes squint, which is like triggering my allergies even more. So was just having like a miserable morning, it's going crazy.
Speaker 1:Oh, man. I think everyone can relate with like system something's wrong with your system, but you don't like want to figure it out or can't figure it out quickly and just got work to do.
Speaker 2:Yeah. I'm usually okay with that. It's just that the physical I'm like dealing with it with a physical
Speaker 1:Sensitivity to life.
Speaker 2:Yeah. Exactly.
Speaker 1:Yeah. Well, that sucks. I guess if you want me to talk the whole time, I can just talk. You can just nod.
Speaker 2:Yeah. I mean, that might be better.
Speaker 1:Might be better. Mhmm. No. It probably won't. I do have a lot to say.
Speaker 1:I do I just have lots of thoughts. I feel like this is a big week. Do you think this is a big week?
Speaker 2:I was thinking that yesterday. Was like, oh, we're never allowed to talk about tomorrow.
Speaker 1:I I struggle with this feeling of, like, am I a person that you can trust? Like, I don't jump on hype trains. Right? I don't get, like, overly excited about stuff that I shouldn't, do I? Am I a person that does that?
Speaker 2:You are not a person that doesn't do that. I think you're like
Speaker 1:Damn it.
Speaker 2:I mean, there there there's a spectrum. I think you fall somewhere on it. Probably the same point that I do. I don't think but you get excited about stuff.
Speaker 1:I get excited about stuff, but I don't like, I didn't do like the crypto thing. I feel like I'm not just like some guy that's like riding every wave.
Speaker 2:Yes. Correct.
Speaker 1:You agree? Okay.
Speaker 2:Yeah.
Speaker 1:I think this one's big. I think I think things this week shifted for me. Like, in terms of timelines, in terms of like how our job is changing. Yeah. It was Cloud Code.
Speaker 1:I mean, Code for me, I wasn't gonna use cursor. I know people have been using these kind of modalities with cursor for a while. Yeah. Just not gonna do that. Or maybe we talked about it.
Speaker 1:Maybe I said I would try. I don't wanna try. And we talked about other editors that are doing the same thing. But Cloud Code is the thing you kinda said weeks ago. You're like, it just needs to interact with the file system, and that's what it does.
Speaker 1:Yeah. And it it's pretty good. What do you yeah. What's your take on you've used it for a bit now. I heard you at the beginning say it was good or have you used it more?
Speaker 1:Do you use it all? What what are your thoughts?
Speaker 2:Yeah. So it's funny because I don't I still don't use it that much for code, not on purpose. I think it's just like my habits haven't shifted yet.
Speaker 1:That's the hard part, right?
Speaker 2:Yeah. And and when I did try a few times, like I didn't get I didn't I didn't land somewhere good with it. But that's not really its fault. It's just a mix of understanding what type of things you go to it for, what type of things you get a sense like, oh, it's probably just gonna get stuck in a loop. But obviously, stuff I've been posting about, I've been using it a lot as a client for, you know, like all my tools that I'm able to run.
Speaker 1:Yeah. Which is interesting. Does does a given MCP client really matter? Like, does which MCP client you use does Claude code versus using like Claude desktop give you any advantages when you're doing the stuff you're doing? And maybe tell the audience what you're doing.
Speaker 2:Yes. And I don't understand why. Okay. So obviously, this week, everyone's been talking about MCP servers and there's just been these giant threads
Speaker 1:You've been this last twenty four hours. It just like, for some reason, everyone on Twitter decided that they care about this and it's so weird I
Speaker 2:figured it out though. I figured it out. I figured out why this happened this week. Mhmm. Because it was very confusing for me at first because I was like, I'd heard about MCP for a while, I understood it for a while, I knew what it
Speaker 1:was for a two months ago, right? You sent me the first when they first announced it, you sent it to me because we were doing some stuff. I mean, has been around for it's not brand new.
Speaker 2:Yeah. But then for me personally, I was like, I don't use anything that is a client. So as much as I'd like to use this stuff, I personally cannot because it doesn't fit to my workflow. Then Cloud Code came out and that unblocked me and that's when I started really looking to it. And everyone seemed to and I was just like, what the there's no way that everybody was waiting for Cloud Code.
Speaker 2:Like, I know all of you are using Cursor. But it turns out that Cursor updated their MCP support January 23
Speaker 1:Okay.
Speaker 2:To like actually make it like more full featured and like visible.
Speaker 1:Mhmm.
Speaker 2:And I guess it takes like a roughly a month for people to update and probably in the past couple of weeks, everyone like got the update and like saw it and started to set set stuff up. So I think it just blew up this week because it got updated in cursor.
Speaker 1:I don't mind the just like I I mean, it's like it is a natural phenomenon that you can't resist that like sometimes stuff just catches fire and everything you see on your timeline is about that thing. Yeah. That doesn't bother me. The only thing that bothers me is the people who like suddenly like have like tutorials or answers, and they're not like associated with this stuff at all. Like, they just ride every trend, I guess.
Speaker 1:They just need
Speaker 2:to like It's so annoying.
Speaker 1:Come out as like a four thought leader and Yeah. It's so weird.
Speaker 2:It's always phrased the same way. It's like, people have been asking me about MCP. So who the fuck is asking you? Literally, in the world has asked you a single question about anything in your whole life?
Speaker 1:Like, to one person in person and ask them if they have heard about MCP and they will say no. Is on Twitter, so okay, you got some Twitter DMs, I don't believe you.
Speaker 2:Not even, not even. Yeah. No one is fucking asking you anything. You're like, you're not the person anyone goes to when they have a question, okay? So, yeah.
Speaker 2:So the reason this bothers me is because, well, one, at the base level, they actually confuse things more. It's so simple. They're just plugins for Cloud. Like, it's that's all there is. There's
Speaker 1:no like Thank you. It's plugins.
Speaker 2:Yes. There's no like deep philosophical game changing architecture
Speaker 1:changes everything. It's Yeah. Store.
Speaker 2:So they inflate it and they also make it so I actually have stuff I want to ask people about and talk about, but like, I just don't wanna now I'm like adding to that noise.
Speaker 1:You don't wanna be associated with Uh-huh. Yeah. Yeah.
Speaker 2:Yeah. I'm with you. So, just like goddamn, every time. It's like the other side of Twitter that we don't normally see but it's always people with like a hundred thousand followers and we're just like, who are you? Like, why do you have more followers than like the biggest open source people in like the real world.
Speaker 2:So it's just Yeah. It's just like that world is like intersecting right now and it's it's it's very annoying.
Speaker 1:It's very annoying. And it's annoying because you see both ends of it. Like, it's the people who have no clue and they're just like, what is this? And then there's the people who have the big expert opinions. Then there's the people who are like way too hot right now.
Speaker 1:People are talking about this way too much. Somebody like, we need to bring it back down to Earth. And then it's like the people who are the naysayers, and they're like, this is stupid. It doesn't even work. Just the whole spectrum of it is so annoying to me, But I don't remember it being like this with something else.
Speaker 1:Like, this was the most, like, in one day takeover Twitter I've ever seen.
Speaker 2:Yeah. It was mind blowing.
Speaker 1:Is that what the future holds for us? I feel like that's not because of MCP. It's just because of, like, us and social media and the feedback loop and we are we are being trained by these things. Yep.
Speaker 2:I I feel It's
Speaker 1:so scary.
Speaker 2:That's how it feels like any kind of cycle that in any part of the world, anywhere where there's been like a cycle that you're familiar with, it feels like that cycle is way more rapid. Mhmm. It's like way it just compresses more and more and more and more into the point where like we're seeing the full cycle happen within twenty four hours where previously would have taken like a couple weeks.
Speaker 1:Don't love it, if I'm being honest.
Speaker 2:I know. It's it's it's weird. But going back to it, so Cloud Code unblocked me, which made me realize like, oh, this idea of being able to like create tools that you can just have Cloud Code compose By the way, I think this is the part that's not super obvious. It is really good at composing stuff together. Mhmm.
Speaker 2:So if you give it three tools, it knows how to like run a then c then b three times and go back to a. Like it can really compose stuff well, so that's a part that I think is easy to underestimate. That's the magic. And the tools themselves are just stupid. They're just like little Yeah.
Speaker 2:Little functions to call. So yeah, my my exploration this past week has been how do we make it really easy to expose internal functions we wrote for our code base through this system so we can do like all the operations operation stuff and we've yeah, we put together a bunch of fun demos, stuff that I've actually used already. Like I I handled some support cases through a little Stripe tool.
Speaker 1:Mhmm. I love this.
Speaker 2:Yeah. And the alternative before was like way more complicated. So I I'm more I'm more using the tool calling stuff and that's already like massive for me. I'm not using the the code stuff as much.
Speaker 1:So I'm trying to lean into the code stuff.
Speaker 2:Mhmm.
Speaker 1:I'm just trying to like see what it's good at, what it's bad at. I've got, like, two really mature not mature, but, like, code bases that are legacy. They're not, like, greenfield projects where I'm building a new game in cursor or whatever. This is, like, actual repos, And I'm I don't I don't know where I stand yet. Like, I don't know what parts are the limitations of the model, like, Claude three seven, and what parts are fundamental limitations with the approach.
Speaker 1:I think what you said about composing tools together, if you think of, like, the internal tools, the internal capabilities that they've built as, like, their own MCP servers or whatever, its ability to, like, find stuff across the code base and kinda string together a plan seems pretty good. Like, it gets rid of the whole context window problems of, like, everything having to fit into context. It seems I don't know what the mechanism I don't know how it's actually keeping track of its plan and then executing the individual tasks, but it seems pretty good at that part. It's just the actual writing the files that, like, it's still limited by the actual model. Like, three seven just doesn't seem to always make great choices in terms of, like, actually writing the code.
Speaker 1:But it comes up with pretty good plans, and it seems to go through, like, it can run your tests. It can figure out, you know, and and work through eventually with, like, a working test. It can work through stuff pretty well. I'm I'm impressed with that part of it, the agent side of it, I guess Yep.
Speaker 2:Where it's
Speaker 1:bringing stuff together. It's really cool to see it like run all these bash commands and scripts to accomplish things. Yeah. It's clever.
Speaker 2:It's clever. Like, if it hits a roadblock, it'll like find a workaround. So the demo I was doing yesterday was having it operate the terminal API. But I was running it inside the terminal code base and when I initially saw the tool, I messed it up, like I implemented it wrong. So I ran into an error because I asked it like, okay, give me what are those products from terminal?
Speaker 2:It tried to do it, it got an error and then it was like, fuck it, I'm just gonna query the DB and give you the answer. It just went around and gave me the answer. And the other thing that was cool was, again, when I was implementing a tool, I implemented it wrong. So I had it I was like, okay, use this tool, do whatever. It failed, then it went and read the file the tool was in and was like, oh, here's the mistake.
Speaker 2:And it like fixed the tool and then ran it, which was pretty pretty wild to see.
Speaker 1:Yeah. One of my favorite things it's done so far is it's the kind of stuff that I just I get so bogged down and will not push through on. It actually, like, dives into dependency code and node modules to figure out how a dependency works Mhmm. To then ultimately write the code I'm writing to interact with this dependency. Like, it it watching it do that and watching it traverse into node modules and figure stuff out that I knew there had to be answers.
Speaker 1:I'd gone on to GitHub. I'd looked through some of the code. I'd looked through issues, open issues, and I never found, like, an answer to my problem, and it figured it out by just inspecting node modules. So that's pretty cool stuff. Like Yeah.
Speaker 1:Stuff that saves you a lot of time. Yeah. I don't know, like, will the next model is it Cloud four? When is the model where, like, I feel good about the code it writes? Because I don't always feel good.
Speaker 2:Yeah. I don't I don't know. I'm still not super I'm not banking on the raw power getting that much better, I think. And this is the the these are observations that I've seen people make for a while. Were like, based on the current capabilities of the model, it's actually like, you know, one or two years ago people were saying, we actually already have a lot of capability, it's just not being like put together in the right package.
Speaker 2:And even if the base model doesn't get better, like we're gonna see you're still gonna be able to have like much better outcomes. And that's kind of where we are now. Okay. Yeah. We realize if we run these things in a loop and have all these tricks that it can do, it effectively makes this base model a lot more useful.
Speaker 2:But yeah, I don't know I don't know how much more we can get from the base model. At this point, think we're like hitting the ceiling of what we can get from this model. And I also don't know if it's gonna get that much better. I think some of the interesting things that I learned this past week was if you look at the approach of how I wrote was writing these tools in the beginning of the week versus the end, it's radically different. So the first thing I want to do was like I would love to have Claude, code be able to operate my AWS accounts and ask it like, hey, that last ECS deployment, like, did something go wrong?
Speaker 2:Or like, you know, clear out all the files in this S3 bucket. So I went and like pulled the JSON spec for the AWS API, built like all these different tools that like, let's cloud query that spec and then, you know, understand what operations it has, then let's cloud like learn how to build the input for that. Like, I I did like a really complex thing to eventually make a fetch call today to the API. And that's how I started. Then I realized like, man, these tools can just be really sloppy and dumb because they can they can just like handle anything.
Speaker 2:So with the so yesterday, Frank did another demo where he does glob imports like the whole AWS SDK v two and then the tool just takes the service name, the function call and then like arguments like three things. And Claude already knows AWS SDK v2, it knows what services there are and it knows like what functions there are. And we just like dynamically call the function. And there's like no structure to it, there's like this is like a five line tool maybe. Yeah.
Speaker 2:And it effectively drives it well and I'd I'd say anything with the Stripe API it did where I was like, first I was like, I need to import the Stripe SDK and expose each each each like method as a tool. No, I just I just ask it for a path and an HTTP method and a body and I just send that to Tripe API and yeah, it gets it right. When it gets it wrong, it can as long as the error messages coming back are reasonable, it can kind of tell what it did wrong. Yeah. Yeah.
Speaker 2:Like it it sent the date for something with AWS, it sent like the date, timestamp in the wrong format, got back the error, realized, fixed Yeah. So, yeah, just just like the composability, like that whole like loop is so much better than I think people realize it is.
Speaker 1:Yeah. No. It's it's exciting. It's I know people complain that it's expensive, but I love the idea of, like, starting a thing, giving it permission to do whatever it wants, and then going and eating my breakfast and knowing, like, I'm kinda still getting stuff done right now. Like, it may not end up sticking.
Speaker 1:I may have to go back and, you know, correct it or whatever. But like, something's happening and before, nothing was happening because I was just making breakfast.
Speaker 2:Annoying that. Yeah. Yeah. For the coding thing, I think, you know, the price is a question. But for the stuff I'm using it for, it's really stupid.
Speaker 2:Honestly, I bet like a much dumber model could do all the same stuff that I'm asking it to do because it's just converting these two calls together. And I bet it could be super fast because right now, it's kind of slow. Like 95% of the time is it doing AI stuff and then 5% of time is like waiting for the responses from the call it's making. Yeah. So I think a little bit like 20 times faster.
Speaker 2:Yeah. Yeah. And I I wanna see like a version of this with like a dumber model that's optimized to be cheap and and really fast. And they can get like pretty insane results.
Speaker 1:Yeah. I guess I I just saw somebody say the other day that like the bigger the models are getting, the slower they're getting. That's inherent. Mhmm. So that would suck because I think for like things like coding, you do need the big powerful models.
Speaker 1:The costing, I don't I don't think about it at all, mostly because I guess I've hired engineers, like, as a startup founder.
Speaker 2:It's like Yeah. Yeah. Relative to what it's doing, it's it's like a fraction. Yeah.
Speaker 1:These things just get cheaper. That's what they do. Like, the models have all gotten cheaper over time, so this is as bad as it might be, I would think. Yeah. Or the models just get a lot better, and then it's worth it.
Speaker 1:But then maybe they're more expensive. But, yeah, the cost thing, I don't I haven't really cared. And I just I see people, like, outright dismissing it because of the cost. And I don't know. Maybe their experience is something I'm not experiencing, but I I know some people, have five pains going, and they're using Gitwork tree to, like, have different branches checked out, and they're doing work simultaneously.
Speaker 1:Maybe you could rack up some pretty big bills that way, I guess. But
Speaker 2:Yeah. I've only I mean, I've been doing like a lot of heavy testing. This is on working with this stuff and I only spent $25 Yeah. Till I get to the point where I launch something. And then yeah, like, you know like I said, I'm not using it for coding.
Speaker 2:I'm sure it like goes in longer loops there.
Speaker 1:Yeah. I wanna get anecdotal on the coding thing real quick. Yeah. Because I I do hate when people just praise or shoot down something and then there's no examples. So I've had like one impressive thing.
Speaker 1:The first thing I had it do really in terms of coding across the statmuse codebase, which is giant. I mean, it's a huge consumer app. I mean, consumer codebase, whatever. It's a huge app. It's a huge something.
Speaker 1:It's big. Lots of code. There's lots of files. And it it made a change to fit was just a bug fix, but it made a change across three files that were very spread out, and it figured out, like, what it had to do, and it was perfect.
Speaker 2:Like, I didn't have
Speaker 1:to change anything. Pushed it. Nice. I mean, tested, and everything was good. So that was very impressive.
Speaker 1:And I was like, we're going to the moon, everybody. This is this is incredible. But then then, like, on the downside, there was a case where I wrote, like, one file and I needed, like, 20 files like it. It's it's a test file across like, testing one resource in an API, and then I need to, like, repeat that. Like, I wrote it manually.
Speaker 1:I need to now do this for every resource in the API, and it literally can't follow my it cannot just do what I did but apply it. Like, it will not just follow the pattern. It's doing new things. It's doing its own thing. I think all that stuff chalks up to, like, Cloud three seven just being dumb.
Speaker 1:And I know a lot of people on Twitter say Cloud three seven's dumb or overfit or whatever. Uh-huh. It's really frustrating. Like, feel like that should be a thing that LLMs can just nail. Like, here's a pattern.
Speaker 1:Now, apply it to these other things. I just can't get it to do it. So I'm writing it manually and that sucks.
Speaker 2:Okay. So I didn't know that people were saying Cloud three seven is dumb. But now that you've said that, I have had a bunch of situations where I'm like, I was kind of surprised that like, I gave it work that seemed like LM work and it Yeah. Was getting confused. Is it worse than three because I saw someone post as more in three five is still better than three seven and I was like
Speaker 1:I've seen that sentiment as well. I don't know. I could I think in Cloud Code, you can change the model. Maybe I could change back to three five and see how it does. Maybe with this specific task, it'd be great if something could just pair it.
Speaker 1:It would save me hours and that would be great.
Speaker 2:The other interesting thing is, isn't it funny to contrast the approach of Anthropic and OpenAI? They're like going towards the same goal but these companies cannot feel like any more different.
Speaker 1:Any different? Yeah.
Speaker 2:Yeah.
Speaker 1:Yeah. Could you explain the OpenAI approach because I just can't I just can't even listen to them talk.
Speaker 2:OpenAI's approach is so like head in the clouds. It's like we don't even care about doing anything useful in the near term because our goals are so like years out and like there's no point wasting time in the short term thing because we're gonna do this much bigger thing, which I don't think is wrong. But it can be wrong if you've thought about that concept so much that you're now like doing it on purpose. You know what I mean? I feel like Yeah.
Speaker 2:There's like, everyone knows you got to aim big. But because you know you have to aim big, you kind of like do it in an artificial way, you know, like kind of poisons your Mhmm. So that that's a thing I get from OpenAI where it just feels like how much are they just like feeding this this narrative but like not actually doing anything. And I can see them like not do And then on the other hand, like, Anthropic feels the opposite where they're just like building very useful incremental things. They're like, oh, here's a problem we ran into, here's a fix for it.
Speaker 2:Okay, can make this thing Yeah.
Speaker 1:I feel like using Cloud Code, it's like they built this for them and that they're using it. It has that feel.
Speaker 2:Yeah, exactly. And then they're also like integrated in I think Cursor's main partner, I think is Anthropic, if I understand that So like, again, very practical consumer of this stuff and like trying to solve like really practical problems like just solving just the next incremental thing. Mhmm. Another way I'm saying it, it makes it sound like obviously the anthropic approach is always better. I actually think it can go either way.
Speaker 2:Like you can just go this incremental approach and then someone just lands something that just obsoletes everything you've been doing. I've seen that happen. But in this case right now, it feels like their approach makes a lot more sense especially in such a like frothy market. Their CEO in every talks, he seems like really low key. Yeah.
Speaker 2:Not like Like definitely a lot more like Sam.
Speaker 1:Oh, man. I cannot stand Sam. I cannot stand Sam.
Speaker 2:Can't stand him anymore. Like, ew.
Speaker 1:Every time I look at him now, it's worse and worse all the time. I just that guy.
Speaker 2:Okay. I have a question about this. I think he has a bunch of plastic surgery, I think.
Speaker 1:Oh, really?
Speaker 2:And I'm wondering Maybe that's what
Speaker 1:it is.
Speaker 2:That's part of what's feeding this.
Speaker 1:Is it his face that's bothering me? Because something about him is bothering me. I mean, also his personality and the things he says.
Speaker 2:So so Liz has this funny thing where she's like, sometimes we intellectualize situations so much and all that's going on is we think someone's ugly. There's this whole thing with and she points out like Lena Dunham who's like this she created this show called Girls. Definitely a really kind of like outside of the way she looks, she's like, there's reasons people find her annoying or controversial or whatever.
Speaker 1:Mhmm.
Speaker 2:But she was just like the subject of endless like, people just talking about how much they hated her. They're just like all this like intellectualizing on top of it. Mhmm. But she's just like, kinda ugly. It's like, she's pretty ugly to look at.
Speaker 2:And she's always like, all that's going on here was that she's ugly and that and that's And she wasn't ugly, I don't think this like this conversation around her would have inflated that much. Interesting. Sam's face to me looks like I wanna punch it. Like it's just one of those punchable faces.
Speaker 1:See, I thought that was because of the things he says. But maybe it's not, maybe it's just his face.
Speaker 2:Okay. So this brings me to another point. I am very convinced that there is something with the way people look and their personalities and I can't unsee it, right? Like sometimes you see someone, like you see a random picture of someone and they look like a really nice person. Mhmm.
Speaker 2:And it's impossible to imagine that they're ever like a huge asshole, right? Like they have like a nice sweet face, like whatever that means. Mhmm. And I'm just like that never that literally never happens. Like, never see someone that looks nice and it turns out they're like they have like a mean aggressive personality, like I just don't see it just doesn't happen.
Speaker 1:It's like temperament actually shapes your appearance in some way like
Speaker 2:over time. Or like manifests. And then like and I feel like you this has to be a thing because some actors get typecast. Yeah. In like mean roles or like childish roles or Yeah.
Speaker 2:So there there's like something something to it. The flip side is like obviously actors that play evil people because they look evil usually are like nice in real life.
Speaker 1:Yeah. Right.
Speaker 2:But there's some there's I feel there's it's not like totally zero. There's like something there's something there.
Speaker 1:There's something there. Yeah.
Speaker 2:So Sam looks like a sneaky annoying motherfucker. Like I just is. Yeah. And just feeds and it shapes his face and the shape and the world treats him a certain way and that like makes him more a certain way, you know. I feel like there's a feedback loop.
Speaker 1:Yeah. Do people do are there people that like him? I don't know how he even like likes himself after the congress thing. Like, the congressional hearing where he was like, I just do it because I love it.
Speaker 2:Like, I don't have any stock.
Speaker 1:I don't and then, like, all the things he's done since then, like, how can anybody like this guy? How does he sleep at night? I just like, does he ever wake up and just be like, I am the absolute, like, worst version of Silicon Valley CEO Founder guy, And was it worth it? Like, I don't know.
Speaker 2:No. Because because there's just legions of people that like him, quote unquote, because they're incentivized to suck up to him. And I feel like you're surrounded by people like that, so hard to get
Speaker 1:a get
Speaker 2:a sense of how annoying this is. But again, it's at the end the day, it might just be arbitrary. It might just stem from the fact that he looks a certain way and he doesn't control that. Really?
Speaker 1:That's true. Like you're saying, maybe he is an awful scammy person because of his appearance. He doesn't even know?
Speaker 2:The world treated him like he was and he became, you know, it's like the Zootopia thing.
Speaker 1:Okay. Well, I do have this thing where I exonerate everybody. Something I'm learning about myself. Like, no matter what you do, I just assume it's part of your story. Your your circumstances led to that.
Speaker 1:There's nothing you could do about it. What is that? That's like a that's probably like a spiritual belief. It's like materialist or something like where you just think everything in the universe is just a product of atoms banging around and doing I don't know. Do I even know what I'm saying?
Speaker 1:I don't know what I'm saying.
Speaker 2:Yeah. I mean, that's that's I think I believe that too. Like, everything is just I I don't believe in free will. I don't believe in
Speaker 1:Oh, really? See, I always backtrack when I like explain that I just don't view people as like bad because like they just do things based on how they were raised or how they were whatever. And then people like, well, do you believe in free will? I'm like, of course. Yes.
Speaker 1:Oh, no. I do believe in free will. So could you explain? You're just leaning hard in to know free will.
Speaker 2:So I think my position on this is it's not an interesting question in that I don't believe I don't believe in free will but I do believe that our brains cannot comprehend the concept of no free will. So you saw the Builder Society around assuming people have it like at a technical level like yeah, because a meteor was like one centimeter off from where it would have been otherwise, that's the reason that I have a beard. Like I I I think that like things change together like that and like my personality is like
Speaker 1:that and and all and all that stuff.
Speaker 2:Yeah. And I don't literally control all that and I think everything is just I mean, to put it another way, I think if you knew all the rules of physics and you knew the start point, what things looked like at a certain point, you would know what where things are at any future I
Speaker 1:think it's a function. It's just a pure function Yeah. You take the whole state of the universe you put it in the function and you get the next state.
Speaker 2:I think that's like a reasonable thing to say and like be like, yeah, that that is how the universe works. Simultaneously, like, practically, we we have the illusion of free will so we can't really see it that way day to day. But yeah, I mean
Speaker 1:feels it feels like free will. It feels like Exactly. Right now that you couldn't there could have been it couldn't have been predicted.
Speaker 2:Yep. Exactly. And so like, at the end of the day, do. Simultaneously, sometimes I think about like, I'm kind of like you where I I wouldn't say fully exonerate. I think I I just hold two things together.
Speaker 2:Like, I'm always like, yes, there is. For this person, there was no avoiding the situation. They couldn't have been a better person, they couldn't have been whatever, it wasn't like in their control. Simultaneously, like, I still think they there's consequences that need to happen, but I will temper how intense I am about that. Right?
Speaker 1:Yeah. I'm realizing I I'm breaking my own rule because I don't I always thought I viewed everyone like this, but I don't. I view like dumb people this way. Like, if they're dumb, like a dumb person who just kills someone because they need to, I don't know, feed their family. That person I exonerate, and I'm like, they couldn't help it.
Speaker 1:That's just the environment they were in and the circumstances they were under. But someone like Sam Altman, I'm like, nope. That guy, he could have fixed it. He could have not been such an asshole. But Yeah.
Speaker 1:He chose to be.
Speaker 2:But like what made him, you know, maybe it was just his face.
Speaker 1:Maybe. It could have just been the face. Yeah. I don't know.
Speaker 2:I I mean, even like intelligence to me, it's just something that you're birthed with, so and I didn't I didn't really control that. So
Speaker 1:Yeah. No, it's true.
Speaker 2:Everything just feels out of my control. But at the same time, you know, I'm not like, forgive everyone for everything and no rules.
Speaker 1:Yeah. Yeah. And I'm not saying that either, I guess.
Speaker 2:Yeah. I'm gonna say people like will cast what I'm saying in that to that extreme.
Speaker 1:Yeah. Nobody's guilty of anything because we're all just a product of the giant universe function churning out new states like a React component.
Speaker 2:Yeah. Exactly.
Speaker 1:I do hate that, like, my brain goes to React when I think of, like, functions modify like, rendering something from state. Yeah. Pure functions. I hate that that they introduced me to it.
Speaker 2:And the worst part is is it even a pure function?
Speaker 1:Is what oh, the universe?
Speaker 2:No, the React.
Speaker 1:Oh, is React? Oh,
Speaker 2:It's it's funny. I don't even like you know, when I was saying I had like React brain and it took me a while to switch to Solid? Mhmm. I implemented for OpenAuth. I implemented like a really good auth context for Solid with like multi account support like switching like all that stuff.
Speaker 1:Mhmm. Mhmm.
Speaker 2:And I'm like, okay, gotta port this to React. I'm just like, I have no idea how to do this. Like, just I just don't understand it anymore.
Speaker 1:Starting over with React.
Speaker 2:Yeah. I'm gonna restart it. LM and have it have it do it, but
Speaker 1:LMs are very good at React, would think.
Speaker 2:Speaking of I mean, we brought this up a bunch before, but I gotta bring it up again. So you were asking like the MCP clients, is there any does it matter which one you use? I actually found that Cloud Code is better than Cloud Desktop. Even outside of coding, just this tool calling thing, I was getting better results in Cloud Code. I don't know exactly what the details there are.
Speaker 1:Now that I think now that I think about it, is Cloud Desktop even does it have the ability to do the whole agent chaining stuff? Isn't it just like every other chat app where you just send it a prompt and it gives you a response?
Speaker 2:No. I think I saw something about like 3.7 reasoning in it.
Speaker 1:Okay.
Speaker 2:I I gotta look. I've just been it's coming in a whirlwind this week. I gotta like Yeah. So I remember I was asking Claude that I I attached some AWS tool and I was like, oh, like, you know, how much traffic did I get to this endpoint in the past week? And okay, it calls a tool, gets the data back, awesome.
Speaker 2:Okay, now it starts to build a React app with hard coded data. So I can run your
Speaker 1:What? My word.
Speaker 2:And I'm like, okay, this this technically kind of spark because now you have like an unlimited you have unlimited ability to build any UI effectively for whatever the best way that to show this information. Mhmm. But man, does that seem like such like a blunt hammer.
Speaker 1:Yeah. I think that's what I've heard people saying that how Claude just it does too much.
Speaker 2:It does too much.
Speaker 1:Like, it doesn't do the simple thing any anymore. Like, three seven is just, like I don't know if that's overfitting. I don't even know what overfitting would mean with an LLM, but it's like it goes too far and gets too elaborate and thinks too big when it should do simpler things, which is really annoying.
Speaker 2:Do you think it's graduating from being a junior engineer to being a mid level engineer? And that's why it's doing all the complicated stuff.
Speaker 1:Oh, it's in that little, like, transition period where Yeah. Yeah. It's in the middle of the bell curve. Yeah. What's that curve?
Speaker 1:The the
Speaker 2:Normal curve.
Speaker 1:Midwit curve.
Speaker 2:Midwit curve. Yeah.
Speaker 1:Yeah. Don't know. I I don't know what the it's so hard to like, this is my first experience with this kind of coding where like, the cursor experience where it can actually do things, and you're not just, like, copy pasting, one shotting, like, functions or whatever. Like, where it's actually modifying my code base while I observe and review the code. It's my first experience with that.
Speaker 1:So it's hard to separate, like, what and it's also my first experience with Cloud three seven. So, like, what is three seven? What is Cloud code limitations? It's just very hard to suss out. I I have noticed, like, Cloud code eventually on a big problem in a big code base where you've been going for a while, you've got a few minutes of computation going, you'll start to eventually see that there's, a context limit that you the context gets low.
Speaker 1:So I guess that's like it is building up some kind of, like, file that describes everything it's done so far, I guess, maybe. And like keeping track of the state just in some text file maybe. But I don't know. Is that transparent?
Speaker 2:No. Think it's literally just what's in what it's printing. Because it prints
Speaker 1:in Really?
Speaker 2:Cloud Code, like the inputs and the outputs. It might truncate the output, but that's that's all it's going
Speaker 1:So it's using that and its calls back to the model as the context needed to keep doing the task needed. So
Speaker 2:Yeah. And that's why eventually the context pulls up and you have to like That's why usually feels like.
Speaker 1:But that that that list, whatever it's printing, like, standard out, that doesn't have, like, contents of your files and stuff in it, does it?
Speaker 2:It does. It just truncated. It just it just shows you the first
Speaker 1:Oh, yeah. That's right. It starts to show you, like, the first few imports or whatever, the first few lines of, oh, that's right. So that file would just be this giant if I inspected that file, it would have all the code for all those files. Yeah.
Speaker 1:Okay.
Speaker 2:Yeah. I think I think what would be what might end up happening so right now, if you have a well organized file tree, this is great for Cloud Code because it can infer a lot just by doing Yeah. A list on recursive list on everything. And that can easily fit in context window. It can kind of have a sense of everything.
Speaker 2:I kind of wonder if there's gonna be some kind of thing where like you can put one comment at the top of a file and just reads up to that comment. So you can get like more information about what is in this file Yeah. And keep that content because that can probably fit. Mhmm. Yeah.
Speaker 2:The approach with all this stuff just seems like, give it a shallow view of everything and then let it drill down Mhmm. And then get and then that way you're not like overloading your context right away. Yeah. So I think that that type of thing works pretty well. The thing that we've been thinking about is I think we're gonna end up we don't wanna do this.
Speaker 2:I feel like everything we do, we just don't wanna do. We just have to do. I started off so many projects like this being like, I don't wanna do this. I think we're gonna have to build a web based
Speaker 1:Client.
Speaker 2:Client. I mean Yeah. I I I I'm not even gonna call it an MCP client, I'm just gonna say like something that's embeddable because the best experience for the stuff we're working on is we saw this project OpenControl. Let's you easy to find tools, import functions from your code base, whatever. When it's deployed, you should just be able to go to the URL and have a chat box right there and be able to like talk to it and do everything.
Speaker 2:Shouldn't have to like then go put it into yeah, exactly. And then like the clients, you you can't just give this to your team. You gotta like tell them to set up a bunch of stuff first. So we probably will end up making our own like chat type interface. It's funny, Vercel actually released a template.
Speaker 2:I forgot they call it chat. Vercel, I don't know. They have like a chat template that does exactly it's exactly what we need. It's just like a pre built thing plug in.
Speaker 1:Like builds the client?
Speaker 2:Yeah, exactly. It's a plug in I don't wanna call it a client just like just the the front end for it. Plug in an and you can talk to it. It can even do tool calls. It can even do like custom UI.
Speaker 2:You can like load it with custom UI widgets and if it recognizes that the data coming back can populate those, it'll like pop that up instead of Uh-huh. Print as text. It's a really good, really well done. But it's a fucking like app router thing and I can't we can't just like I can't drop that easily into lightweight. Like, I I can't deploy a whole Next.
Speaker 2:Js app.
Speaker 1:You don't want to deploy a Next. Js app into people's AWS account. I mean, you guys have the ability to do that.
Speaker 2:But it's yeah. It's it's such overkill for this.
Speaker 1:So Yeah.
Speaker 2:Yeah. So I think we're just gonna look at it and like Can you just
Speaker 1:use yeah. Can you just use the actual front end of it? Like, remove the Next. Js piece to just like get the Yeah. Good parts?
Speaker 1:I'm
Speaker 2:sure I think I think the thing they saw first, they did lot of broad thinking. Like, at the in terms of what it can do, made ways organized, I'm assuming it's probably gonna be correct. Maybe a few things I I might disagree with. But yeah, we're just gonna try to reimplement it in something that is way more lightweight and portable so that
Speaker 1:Mhmm.
Speaker 2:One, we can drop it. So deploy open control, boom, you have it there. But also in ST console, we wanna be able to drop it in there, preload it with tools so you can like ask it if I create up his account. We even like load the your state of all your resources into the context.
Speaker 1:Mhmm.
Speaker 2:So it can but you can like understands all the resources in your app so you can be like Yeah. This is clear out the bucket for my file uploads and it knows what that is.
Speaker 1:Yeah. Invalidate the cache whatever. Exactly. Yeah.
Speaker 2:We realized we had this like crazy long tail of this thing sucks native as console, let's build a better UI for it. An infinite long tail that we would never have made progress on. Yeah. But this kind of like in one go involves a lot of that. So we want to do that.
Speaker 2:But then if you think about it like, theoretically every product probably has the same thing where they're like, I don't want to build a UI for every little thing. I have like a dozen actions that my system can do Mhmm. Just give I always want to be able to embed a chat interface in my app so people can and my users can like do that. So I think a embeddable nice LM chat thing is gonna be useful definitely for us, maybe for other people. Like every like if you think about like building a new app now, you would just start with the back end, like implement all the functions, database, whatever
Speaker 1:Mhmm.
Speaker 2:And v one of your app could just be this thing. Yeah. Right? You can just call the tools and then over time you can then build like dedicated UI for things that chat doesn't handle well.
Speaker 1:Yeah. Like, the issues tab in the console, like, that's that's a valuable thing to have. That AI is not the right interface for it. Like, a chat interface is not what you want there. Like, you wanna be able to look at the list and check them off and see when they last happened.
Speaker 1:Like, that's Yeah. There's special UI still needed.
Speaker 2:Yep. Like, anything that's like tabular, like, anything that's like a long list of stuff, like, it's they need to, like, go through and do work on. Personal UI is great. But just like that long tail of stuff that you never get to, yeah, like it it seems like this is a a good useful thing.
Speaker 1:Okay. You told me not to use our client, but I'm stuck on this because now, like, I understand the MCP server thing. Like, it's a very simple protocol where you define these tools. That makes sense to me. The client side, though, the more we talk, like, the difference in Cloud Desktop versus Cloud Code, what you're gonna build, I'm just realizing I have no idea how the client side works.
Speaker 1:Does the model decide what tools to use? Is it just part of the context where it passes
Speaker 2:Yeah.
Speaker 1:As the beginning of the prompt? Like, here's your list of tools and the definitions. Like, how does that client work?
Speaker 2:I think the easiest way to understand this is to actually look at OpenAI instead of Anthropic because you realize that this capability has been around forever without these like specific without the specific packaging. Mhmm. So OpenAI Forever, when you send a prompt to their API, you have the option of listing a bunch of tools that were available to the LLM. And it's almost the exact same form as MCP server. It's like an array, the name of the tool, description of what it does and a JSON schema of the input.
Speaker 2:Okay? Mhmm. Then the prompt up to their API
Speaker 1:You lost me a little bit because I didn't know you could do this in open a p in OpenAI.
Speaker 2:Yeah. So you can't. So the the tool What
Speaker 1:are the tools?
Speaker 2:They call it function calling. No. There's no tools.
Speaker 1:Oh, it's function calling.
Speaker 2:Yeah. So you pass it a list
Speaker 1:I think I've heard that term.
Speaker 2:When you send it a prompt, can also pass it a list of tool definitions. In the response, it'll tell you, hey, based off this prompt, you should probably call this tool.
Speaker 1:Got
Speaker 2:it. Then on your end, you call the tool, get the response, then you send it back up.
Speaker 1:Well, that's not a very good user experience compared
Speaker 2:to No. That's identical to what Claude's doing.
Speaker 1:What? From the u maybe I'm misunderstanding the user versus the actual client.
Speaker 2:The client so this is what the this is what the client is.
Speaker 1:Well, this is what the client is.
Speaker 2:Yeah. So user sends a prompt, you send the prompt up, it comes back with the tools you need to call, your client calls the tools, gets a response, sends it back up, and it doesn't goes back and forth until there's maybe no more tool calls coming back.
Speaker 1:Okay.
Speaker 2:And this is like a very dumb version of like this this type of product.
Speaker 1:Mhmm.
Speaker 2:I'm assuming that there's a lot of tricks that Clog code implements based on the response. Maybe they have some hard coded things that it does differently like that's probably where the quality comes out. Mhmm. But bare minimum, like you can do this with OpenAI. Like you can give it a list of tools and it'll tell you which one to call based off the prompt.
Speaker 1:Yeah. Okay. So is OpenAI ever gonna just like use MCP and standardize?
Speaker 2:I don't know where MCP is gonna go, to be honest, because let's say you you are like us, like terminal, we have a we have a API. It's document with open API. Mhmm. LMs understand open API. You could just have a system that points to the API.
Speaker 2:Like, don't even need the MCP in this case. Right? Right. So I don't know where it's gonna go. So the way we're building, we're building in a way where we have our concept of tools and we expose them through MCP.
Speaker 2:And maybe MCP dies, maybe it goes somewhere but Mhmm. It's just like the flavor that is getting traction right now. Mhmm. But yeah, we'll see. But the
Speaker 1:idea will persist. It's just
Speaker 2:Any yeah. I guess the question is like, did we need another layer? Right? Like, if you take like, linear has an API, think it's a GraphQL API. GraphQL APIs are crazy, like introspectable.
Speaker 2:So Claude could have just been like, yeah, add any GraphQL API URL here. Don't need any new concepts. So there's some things you lose, but I don't think it just matters for 90 Yeah. Percent of cases.
Speaker 1:Definitely doesn't matter for all the people that on Twitter are asking about MCPs. It's like, doesn't just stop. Please stop.
Speaker 2:The thing that blows my mind is like, why do you need to understand something so much before you just go try it? It's like the smallest thing ever. It's like such a small concept. Like, why do you need
Speaker 1:to It's like people are reacting like the quantum announcements. Like, somebody explain this to me. Like, I don't understand. I can't go play with quantum stuff. I need you to explain to me like I'm a dummy.
Speaker 1:But NCP is like it's just not a thing that should have gotten this much attention on Twitter. It's just a weird weird situation.
Speaker 2:Makes no sense. Yeah. Just just go it's just like it's like a button you hit in the settings. Like, just it's just so weird that people are like, no, I need you to like give me a freaking dissertation on this before I'll even consider spending my valuable time opening up the settings page and and and trying it out.
Speaker 1:Weird times we live in. Oh, should we talk about some of the weird things that are going on in the world
Speaker 2:Yeah. Before we
Speaker 1:get off here? Well, like, what do you mean? Now I feel crazy. Like, is the world not kinda crazy right now? Like, I feel like
Speaker 2:Right now, my my world is MCP and nothing else based on Oh, okay. Yeah. Yeah. Twitter just erased my brain for the last twenty four hours, and all I all to my head is MCP.
Speaker 1:That okay. That's true. Yeah. Twitter did get overtaken. But before it was MCP, there's there's tariffs.
Speaker 2:Oh, the tariffs are really weird.
Speaker 1:We're tariffing people, which just seems rude, personally. Like, there there's this there's some pretty crazy, like, shifts in the geopolitical alliances. It feels like we were on one side, and now we're on the other side. And I don't know what to think about all of it. It kinda feels like the outcomes this year or the next two years are all over the place, and I don't know what's gonna happen.
Speaker 1:I I guess the first thing I'll say, and I don't really get into a lot of this stuff because I'm privileged, and I don't have to think about politics. But when I do think about it, isn't there a reason to, like, have the world be in a pretty good shape and not be like, f you all. America's gonna be amazing, and you guys are gonna burn. Isn't there a reason we've always, like, propped up the world because it's good?
Speaker 2:Yeah. I think I've been thinking about this vaguely for a while and someone finally put it into good actually someone that works at OpenAI, I think, put put it into these terms. They were like, oh, I think Trump does not understand positive sum interactions and like he views everything as a zero sum, like for someone to win, someone has to lose. And it's interesting because I have run into really successful business people that have businesses that are literally in that type of situation. And I think in tech, we're we're very used to being like, if we build something that enables someone else to be 10 times as productive and we get paid 10% of that.
Speaker 2:Like we like naturally understand that we're just in the productivity zone like improving Yeah. Efficiency productivity whatever. So it's really hard to be a zero sum thinker because like it's it's like in your face all the time where Mhmm. You know, everything like both everyone every party can win. But if you're in the like some kind of like hyper competitive environment with like scarce resources Like
Speaker 1:real estate?
Speaker 2:Like real estate for example. A lot of the times and especially when government's involved, like whenever there's like government regulation involved and there's government like you need to get approvals and stuff, a lot of the times the interactions are zero sum. And like I I knew these people that had a massive healthcare business and they operated in like the most like cutthroat, like totally alien way. Like every interaction they went into it being like, I need to fuck over the other person before they fuck me over. Yeah.
Speaker 2:And I don't think they're Ultimately, I don't think they're wrong because I think the business they were in, that just is what you had to do because it was like a fixed thing and whoever gets most of it like wins. There's no there's no possible way to grow the pie. It's just fixed and like it gets divvied up. And I think most of the if you're not in business at all and if you're not in maybe tech at all, I can see how that's how that's naturally how you see the world. It's like, if someone is getting money, that means they took it from me.
Speaker 2:And you see this on the last two, like people on the left like obsess over billionaires, like billionaires should be evil, like illegal, like all that stuff. The same it's like the same flavor of that where they're like, oh, they have a tariff on us, that means we need to have reciprocal tariffs.
Speaker 1:Mhmm.
Speaker 2:Which is fine if the goal is to reduce tariffs on both sides to zero. Right? Like if it's like
Speaker 1:a Yeah.
Speaker 2:To like you're playing hard with them to get them to reduce theirs to zero as well. But it doesn't seem to be that, it just seems to be like, hey, can make money if we do this. Yeah. They're getting away with it and we should be able to get some money out of it too. And every single thing has just been like, how do we get something out of this relationship in a very direct way, you know?
Speaker 2:Mhmm. But I feel like that's kind of what's going on with all that.
Speaker 1:Do you have anxiety over where things are headed? Is World War three about to erupt? What's going on?
Speaker 2:I mean, it doesn't feel I mean, just from the economic point of view, it doesn't feel good because I am very like pro things like free trade and I really believe like a lot of this stuff Mhmm. Enriches us in really indirect second order effect ways that it's it's hard to explain. Mhmm. And yeah, it doesn't feel good that the world is like looking for alternatives to America. You know, like, okay, how can we get by without depending on America as much?
Speaker 2:How can we, you know? Mhmm. I don't think that leads to a good place. And then the people that are more funny about it, Nikita had a funny one where he was like, I forgot the name he used, it was just some like very Midwest name. He was like, don't you see as soon as Bill something from I mean, he's Missouri, he's from Missouri.
Speaker 2:Living out in Missouri, figures out how to make GPUs, we're gonna be on top all over Kinda like smelt metal and crap GPUs. Just like we're not yet we've like progressed maybe too much away from some kinds of manufacturing. But like, we don't wanna fucking be making t shirts. Like, we've our economy is past that point, right? Like, we're building stuff that's so complicated that requires physical resources from every corner of the earth to produce and it is the most highest value thing that we could be doing.
Speaker 2:Other other people don't even have the option. Like Yeah. Other countries don't have the option of building GPUs, they just don't. It's not an option. We do.
Speaker 2:So we should take advantage of that as opposed to like going back down and being like, how do we produce steel? Like, we we produce the stuff that steel gets turned into, right? It's
Speaker 1:Yeah.
Speaker 2:Yeah. So I I don't know. It's it's it makes me a weird I think it reminds me of is there's a classic story from from China. Mao had this whole great leap forward thing. Very similar mentality of being like, we're not doing enough of x.
Speaker 2:So if top down, he tried to force certain things to happen. They're like, oh, steel is important. Top down, let's force everyone to make steel. Every village, I need you to produce x amount of steel. So people were like melting their cast iron pans and stuff and making steel in their backyards because they were the government told them to instead of like farming and doing other stuff that they naturally would have otherwise.
Speaker 1:Mhmm.
Speaker 2:All this steel was complete garbage and they had a famine. So it I don't think we're gonna go that extreme, but it feels like that's the kind of mindset. It's like way too basic in a world that's extremely complicated.
Speaker 1:Extremely complicated. Yeah. That's how I feel. I don't know. It just I just I kind of like missed the days where there weren't big political headlines every day.
Speaker 1:I don't know. Yeah. No. It felt like they were all just busy doing pointless things that don't really affect us. It just feels like we're gonna be affected now.
Speaker 1:Yeah. And I felt pretty unaffected in the past by who our president was.
Speaker 2:I mean, reason I don't I'm not that worried yet is my sense on the whole situation is new president, lots of optimism whenever there's a new president, lots of leeway to do a bunch of things. But it's always gonna be limited, like optimism only lasts for so long. But you have like a fixed number of months to get as much shit in as possible without Okay. Much pushback. Yeah.
Speaker 2:And it's like a feedback loop because then you do even crazier shit than you would have normally done and that like depletes optimism even faster. But it feels like we're in that zone. I already feel like we're kind of coming out of it where like like, okay, this is America. Like, you need to satisfy the corporations. That's that's how it works.
Speaker 2:If you do stuff that's hurts the corporations, like, you're not gonna be around for Yeah. For super long. And if you lose that, it's not, you know it can only go on like that for for so long. Like, no one's really been able to escape being liable to them.
Speaker 1:Okay. So one person can't come in and just burn down our economy. Like there is a there are like adults who will enter the room and stop that if it came to that?
Speaker 2:Yeah. Because there's corporate interests. Corporate interests need to follow corporate interests. And I feel like Yeah. There's so many problems in America from that, but at the same time it's what prevents us from doing stuff that's too stupid.
Speaker 2:Because, like, once you affect the money, then it's like, uh-uh, you gotta stop now.
Speaker 1:Yeah. I guess it's separate conversations around, like, the global economic situation and all of the tariffs and all that versus, like, the tensions and the ongoing war and whose side who is on and Yeah.
Speaker 2:That's about Russia's really crazy.
Speaker 1:Stuff's nuts. And like, I don't know. It's just it's weird to me. It's just weird to wake up and like suddenly, like, our president is, like, friends with Russia, but not with everyone we've always been friends with. What do we do with that?
Speaker 1:Like
Speaker 2:Yeah. I don't know. It's
Speaker 1:It's not even, like, good guys and bad guys. I'm not trying to, like, decide who's good and bad, although I think there's a lot of history that says Putin's pretty bad. But, like, aside from human rights and all that stuff, I get that, like, countries look out for themselves. But it's just weird to, like, have America flip sides and suddenly be, like, on the side like, feels like we're on the side of Russia. Right?
Speaker 1:Am I just dreaming that or
Speaker 2:is I mean, the thing that was really weird was, like, the defense department halting all cyber attack stuff with Russia. Like, what the fuck? It's like that's like really Yeah. I don't know. Who who knows what's going on behind the scenes?
Speaker 2:But, yeah, at the end of the day, it's really weird because I'm seeing, you know, if you're Canadian, you're probably really annoyed right now. Right? And I see that. They're all like really upset in America and they're like Yeah. A lot of stuff.
Speaker 1:I always forget how many friends we have in Canada from a tech standpoint, like
Speaker 2:Yeah. And they're impacted by this and like they didn't do any they didn't like vote for it. They didn't like this is like an election that's happening in our country, it's affecting them. Yeah. And I think they understand that it's one guy that's doing this, plus all the people that voted for him, true.
Speaker 2:But I wonder like, okay, let's say, you know, eventually his term runs out hopefully, you know. Hopefully, we don't have to end up with a king.
Speaker 1:Yeah. No kidding.
Speaker 2:Then by the next election, will everyone will the relations like basically reset because it's with like, is Canada's relationship with The US or with Trump, you know?
Speaker 1:Yeah. Didn't their their prime minister kind of Justin whatever said like, he kind of made a speech or two along the lines of like, I know that the American people aren't making this decision that this was
Speaker 2:Yeah. Like, if you if you're a leader of another country, like, you you can't like, you must see it that way. But
Speaker 1:Yeah. It's not like everyone in America just suddenly became like assholes and wanted to tear off your country.
Speaker 2:Yeah. Yeah. Exactly. So, I mean, I had just because American never really has to think about anybody else. So just I just I've never been in a situation.
Speaker 2:But, yeah, it's very I can see it's very annoying. It's like an election in another country that has nothing to do with you and then now now you're you're impacted. Mhmm. So, yeah. See people I see like Canadians, like, really upset and annoyed and angry.
Speaker 1:Hi, Chris. Just shout out to Chris. I hope I hope you're well in Canada.
Speaker 2:Yeah. Our podcast editor.
Speaker 1:Let let me know if you need me to snuggle in any equipment.
Speaker 2:I'm not going to Canada. What do we ship up there? Corn? Like, what the fuck do we make? I
Speaker 1:don't know. Yeah. When I heard about all the tariffs, it's like, I actually don't know what Canada and America trade. Like, I'm sure there's stuff.
Speaker 2:We definitely get wood. We definitely get wood. Oh, it's funny. I did look this up and I was like, I was so surprised by how stereotypical the results were. It was literally maple syrup.
Speaker 2:It was literally maple syrup.
Speaker 1:I just guessed.
Speaker 2:It was like maple syrup and wood. And and and oil, obviously, oil is a big one,
Speaker 1:but
Speaker 2:like Oh, funny. Picture Canada. You picture like mountains, trees, and like the trees have maple syrup, you know. And so that's yeah, that actually is what we're importing from
Speaker 1:Oh, man.
Speaker 2:From Canada.
Speaker 1:So maple syrup is gonna get expensive. I think we're also doing the tariff thing to Mexico, right? I'm using tariff
Speaker 2:like a This whole thing like he just suspended it again yesterday. Oh, he's gonna I wanna post a screenshot. Okay. So I did do one thing. Two or three weeks ago, I opened put options on Nvidia, Tesla and Bitcoin.
Speaker 2:Mhmm. Because I know for sure Tesla's getting slammed from every possible direction like EVs are not selling well. There are a million EV makers in international markets like China, like they're being dominated. So I know they're they're hitting they're probably gonna hit something existential at some point Mhmm. Which Elon normally is very good at dealing with existential things.
Speaker 2:But he's pretty distracted right He's little distracted. Yeah. So I had a short open on on Tesla and I had one on Nvidia because I'm like, it's the most inflated so if we have any kind of it's gonna get hit the most. And the counter toothing on Bitcoin is I think all the pro pro Bitcoin stuff that Trump is doing, I think it's gonna be like a sell the news type of situation where everyone's like really excited for what he could do. And then once he actually does it, the excitement's going be over and people are going to sell whatever.
Speaker 2:Yeah. But the thing I was going to post is a like a screenshot of my the chart of my holdings in the past couple weeks. One they're doing really well, like that the market is not doing well, so those are doing well. But it is the most like jagged up and down looks like someone's having a heart attack type of thing because it's like, we're doing tariffs. I was like, oh, okay, actually no, we're only gonna do tariffs on these things.
Speaker 2:No, no, no, we're doing them again. And it's just like crazy whiplash and the
Speaker 1:market
Speaker 2:It's
Speaker 1:lot volatility.
Speaker 2:Yeah. No idea what's going yeah. So I think yesterday he said he delayed them again on Mexico and and whatever and like
Speaker 1:So avocados will still be cheap for a little while?
Speaker 2:I don't know. Liz's sister works at a jewelry company and they got calls from their vendors yesterday being like everything's up 10% now because of this and they had to go figure out how to reprice everything. They're increasing the price of a bunch of stuff. They're like eating the cost on stuff. But then it's like, oh, it turns out it's not going into effect for another month and it's like, what the it's just like such a it's crazy whiplash right now.
Speaker 2:So Yeah. To me, I will say all this was awesome and worth it if the end result is every country just gives up tariffing The US because of reciprocal threats. And we end up with completely near free trade.
Speaker 1:Free trade?
Speaker 2:Yeah. Yeah. That's a that's awesome. It's a good goal. I don't but as I say, I don't think that's what's going on because of the zero sum mentality I mentioned earlier.
Speaker 1:We will see. Kinda sucks. Just all the I don't know. It's like I wanna be so excited about stuff right now. Like
Speaker 2:Mhmm.
Speaker 1:Work is changing, like, the potential for tools to make us more productive and all that. Like, it's all so exciting and then it's like, no, all the distractions of Yeah. Like, is the world ending? I don't know. It's just so loud.
Speaker 2:It's just like we it it really feels like the sense that I feel is past peak America, we're aware of it but then the reaction to it is accelerating the decline. Simultaneously, we're getting propaganda videos every single week from China about some fucking robot that's like, you know, doing some dances and shit and then Yeah.
Speaker 1:And then
Speaker 2:it's just like, oh, yeah, like the Chinese like BYD or whatever, the Chinese Tesla is like, oh, they're like outselling Tesla. Like like, look at their car, this only costs $16,000, they're gonna fucking fly and shoot lasers, like, it just it is not a good like, it doesn't make you feel good about about America. It just and it's especially funny for me because I wasn't born here and my family like
Speaker 1:Oh, interesting.
Speaker 2:There was a moment in time where my family was like, we need to make it to America. We did establish a life here, like, we're able to be successful and like my parents did that to set me up.
Speaker 1:And I'm like,
Speaker 2:fuck, am I gonna have to do that again? We just got here. Just got here. I was supposed to be the I was supposed to be the kid that's like, oh, like, you know, my family came here and like set me up and now like, you know, I I did all this stuff. Fuck, is it gonna be my kids they get to have that?
Speaker 2:Like, I'm
Speaker 1:gonna have to
Speaker 2:move to like
Speaker 1:That's totally.
Speaker 2:Singapore or something and they're gonna be like like, oh, my my parents
Speaker 1:moved to Singapore.
Speaker 2:And the funniest thing is, like Liz's grandparents came over in their like, maybe they were like late forties fifties, right, when they were older? Mhmm. And they had no idea. They were like, I'm gonna spend my whole life in Cuba and all of sudden they're 50 and they're like, oh, shit, I'm like moving to America? And I'm like, what if my kid like, what
Speaker 1:if
Speaker 2:I'm 50 and I'm randomly immigrating to a new country and to, like, learn the language? Like, that could totally happen.
Speaker 1:Oh, man. That sucks. I it's kinda makes you feel special though, like, if we're the generation that sees the, like, the real decline of America. I mean, as much as that sucks, I don't know. It's kinda unique.
Speaker 1:Like, how many generations have there been since America was founded in seventeen hundreds or whatever? And we're the one?
Speaker 2:That is the one that feels special, Adam. What the
Speaker 1:mail in the coffin?
Speaker 2:It's not gonna be like that. It's gonna be like Europe. It's gonna be like infinite slow climb. Just yeah. And I see so many parallels with us in Europe right now.
Speaker 2:It's like, you know how you know how like five years ago, we'd be like, you're up, you suck, like look at all this stuff we have. And they'd be like, you don't even need all that stuff. Like it's not good to have all that stuff. Yeah. I'm literally seeing that now.
Speaker 2:Like I think someone from the government was like Like in America was like, someone in the administration was like, you know, it's not always about being able to buy cheap stuff. Like that's not always like the best thing. I think exactly what all the European people were saying to us for years like, know, like you don't even need that many eggs.
Speaker 1:Oh, it's so good. Oh, man. That sucks.
Speaker 2:Weird times. I hope I hope this is just a temporary thing, of course. But, man, we'll see.
Speaker 1:Are all the I guess all the AI stuff is a lot a lot of leading by America. I mean, there's DeepSeek. But
Speaker 2:like Yeah. Yeah.
Speaker 1:The other major frontier models are all American companies, right?
Speaker 2:Yeah. Yeah. No. I think like I said, we're we're still number one at pretty much everything. But obviously, like, the when you look back in history at things that have declined, the decline started like during their best times always.
Speaker 2:It's like it's someone comes to you and say, okay, the decline is happening, it seems so ridiculous because like things are better than ever and you're like Yeah. Everyone and everything. So it kind of I think there is like some precedent for that. But yeah, I mean, that's the only thing that bothers me. There's so many people being like, no, like we're like all the people that are obsessed with talking about this and they're like, no, we can do it.
Speaker 2:America can do everything. Like we don't need help from everyone. You personally aren't doing shit. Like the reason America is good is because of like these like really small set of companies and people that like Mhmm. Are continuing to push the bar.
Speaker 2:It's not any of the people in this conversation. I'm sorry to tell you like Yeah. Like, you know, like, we're so great. Well, x y z z. Yeah, like as a country but like not you personally, you know.
Speaker 2:You suck.
Speaker 1:Such a dax response.
Speaker 2:I just tell them unemployed. There's they just have unemployed energy to them.
Speaker 1:Alright. I gotta get back to monitoring my Cloud Code babies. They need me to feed them.
Speaker 2:I used to have, like, a bunch of TMX sessions open. Each was a Cloud Code thing running Yeah. I think I a minute.
Speaker 1:A hundred dollars a minute. I've already written a a little tool for, like, spinning up a new session or a new window in my session, git work tree, check out a branch, like, just kinda, like, get me into a new context where I can start up some work and then monitor it. What is
Speaker 2:git work tree? I heard you say that a few times, I realized I don't know what that is.
Speaker 1:It's just the ability to, like, check out multiple branches on the same file system. Like, you don't have to git clone. It's like you you can use git work tree to, like, have two branches that you're working on in parallel.
Speaker 2:Like, what does it look like on a file system?
Speaker 1:Well, it actually, like, it's annoying because it creates a new folder.
Speaker 2:A new folder. Yeah.
Speaker 1:Yeah. Yeah. It just creates a new file system but it shares the same git
Speaker 2:I see.
Speaker 1:Directory.
Speaker 2:I don't know.
Speaker 1:Don't know how much better it is than just cloning a bunch. I guess you could just clone a bunch but I guess it's more efficient.
Speaker 2:Oh, I see. Because you can't if you switch branches and it's and like have one tab on one branch and one tab on another branch.
Speaker 1:Right. Normally, if you're switching branches, you're sharing a file system. Yeah. Yeah. See.
Speaker 1:Yeah. I learned about a new git thing like every day. How how convoluted is git? There's just like so many tools in git that I have not or will not ever use.
Speaker 2:Yeah. It's one those cloud situations where like the online primitive is so good and enables like a million things, but those million things were not productized well. Yeah. Spread out all over the place. Yeah.
Speaker 2:Okay. I'll let you get back to
Speaker 1:Okay.
Speaker 2:Your new best friend.
Speaker 1:Yeah. My new best friend,
Speaker 2:Cloud Code. I broke up with I don't know who. There was nothing before Cloud Code. That's true.
Speaker 1:I mean, there probably was,
Speaker 2:but I wasn't using it. Okay. Bye. Alright. See you.
Speaker 2:See you.
