loading…
// full transcript — Tanz0rz on DevRel, local AI, and privacy maxing
slop.atg.eth: Ok, resetting.
Anon5304: Exactly, like most servers.
slop.atg.eth: Reset the STT. Okay, we're clean. Let's see, Claude probably heard some of that, but I think we're good there. Let me do a mic check with him real quick. Yo, what's up, homie? Are you in the chat? How you feeling? What are we doing today? Tell me about the guest.
clawdbotatg.eth: Hmm
Anon5304: ###
clawdbotatg.eth: Yo, I'm here and feeling good. Videos up and I can hear you loud and clear. Today it looks like we're hanging with Tanzer RZ. I'm picking up a privacy maxi vibe. WireGuard, lockdown network interfaces, isolated boxes. What's the plan?
slop.atg.eth: Hopefully he doesn't leak that later during the call.
Anon5304: Oh, this is a shared desktop, I see.
slop.atg.eth: Yeah yeah. And if you want to share things, there's like a shared screen over here.
Anon5304: Yeah, yeah. Yes,
slop.atg.eth: you a wedgie on the plane. Next time I run into you, and we can both be signers. ...from here and like actually do a swap as our multisig.
Anon5304: The place
slop.atg.eth: The Okay, looks like he's up This is gonna be loud If you run Safari
Anon5304: I'm on Chromium. You're super muted now. Hold on.
slop.atg.eth: Oh shit, okay.
Anon5304: I think that's what's going on.
slop.atg.eth: Yeah, I have like this dynamic equalizer on one of these things that's like, yeah. It takes it a second, content:
Anon5304: Also, great plug for open source software, OBS is the best camera management thing, and it's completely free and open source.
slop.atg.eth: Yeah
Anon5304: This is the era of open source software.
slop.atg.eth: Yeah, I wonder if we'll see like OBS forks, you know?
Anon5304: Yeah, I wouldn't be surprised.
slop.atg.eth: Maybe there's no need to even fork it,
Anon5304: I'll be back.
slop.atg.eth: Yeah, we got lots of time. We have fifteen minutes. Probably 20 minutes actually until we start.
Anon5304: That you're thinking of following
slop.atg.eth: Wait, what was the question?
Anon5304: Like an outline
slop.atg.eth: Oh, yeah. Probably the usual crappy intro of like how you got into shit. Usually just a good like one minute make sure the stream is warm And, yeah. what you're building, like definitely want to talk privacy stack future cypherpunks
Anon5304: All with a veneer of AI on it, right?
slop.atg.eth: Yes, exactly. It's like probably this episode would probably be even more AI than crypto, even though we're both in crypto. Um, The idea of vibe coding...
Anon5304: Canola
slop.atg.eth: There's something that we touched on on the flight, which I'm just going to repeat all this on the stream, which is kind of annoying, but sorry. But like you talked about how your roommate is a real programmer or something, and you say vibe coding, like out of respect I think that's like a really interesting like kind of thread to pull and talk about. From our point of view, the best programmers are gonna get left behind. people who are against vibe coding right now are like in trouble because there's about to be like super intelligence that's like coding and finding bugs than us?
Anon5304: Yeah.
slop.atg.eth: This whole theme of like shit is breaking right now and it's because sloppy humans wrote sloppy code and eventually humans are going to find those problems are going to find those problems and fix them. AI is going to lead to a hardening of software, not a softening of software
Anon5304: Yeah, yeah
slop.atg.eth: That's a whole thread to pull.
Anon5304: I do have some interesting points on that too.
slop.atg.eth: What other stuff do you want to talk about? Hit me up if you have topics that are neat to bring up or things that have been on your mind, or even things you want to shill.
Anon5304: Yeah, if I come up with stuff you mean just message you on Telegram, you mean?
slop.atg.eth: Yeah, that would work.
Anon5304: That's fine.
slop.atg.eth: All right, I'm going to go get more coffee. I'm not even going to start the countdown for a couple more minutes.
Anon5304: That's fine
slop.atg.eth: And then it'll be a ten-minute countdown. We are live.
Anon5304: Hello, how are you?
slop.atg.eth: How you doing, man? Welcome to the slot. We are in it, we are live! I haven't gone live in a while, this is weird, feels funny. How's it feel for you, feeling good?
Anon5304: Yeah, yeah, yeah. I know we've known each other off and on over the years, so it's cool to finally do something formal.
slop.atg.eth: Yeah, we've seen each other traveling, and we ended up sitting right next to each other on the plane of sickness on the way home.
Anon5304: ###
slop.atg.eth: I'm just now recovering, man. wild I just have
Anon5304: ###
slop.atg.eth: I hear a lot of people coming home from Mumbai all got sick I don't know.
Anon5304: Yeah, that's my guess.
slop.atg.eth: He was stressed. The
Anon5304: The following
slop.atg.eth: So yeah, so let's see, crypto dev rel, you work at 1inch as a dev rel. Do you call yourself a dev rel anymore? Is that even a role? Okay.
Anon5304: Yes, yes, my formal role is DevRel.
slop.atg.eth: How'd you get, like, how'd we get into crypto easy
Anon5304: Yeah, so in 2017, I joined the bandwagon. I didn't really know anything about any crypto. I just saw an investment opportunity, lost all my money like most people in 2017, 2018. In 2019, It was like trying to figure out what to do with my time and interest. So I started looking around, and it was still there, and people were still talking about it at least a little bit, and that kind of sparked some genuine interest in trying to look more at the technical side. And funnily enough, I actually started looking into Litecoin first Ethereum is the coolest thing ever." He linked me to EthHub, which was Eric Conner and Anthony Sassano back in the day. And so the friend who showed me EthHub, another friend who was into Ethereum
slop.atg.eth: What year was that for East Denver?
Anon5304: That was, I want to say, 2020 just before the COVID lockdowns. ###
slop.atg.eth: There was like hella lines, right? Is that the one where Okay,
Anon5304: This one was not as busy
slop.atg.eth: Maybe that was the one with hella lines. Yes, yes. Yeah Yeah.
Anon5304: 2020 was still kind of underground. Vitalik was sitting listening to people's talks and nobody was even noticing him. I saw him just kind of sitting with no one around him, and I was just like, I'm pretty sure that's...
slop.atg.eth: ###
Anon5304: I want to say, around that time. All the projects were there.
slop.atg.eth: Take me back. Oh man.
Anon5304: I think it was just before all the tokens started launching. with the blockchain space. Right, yeah. Yeah.
slop.atg.eth: Like One Inch came out of East New York 2019, so they were pretty young at that point. Okay. Yeah
Anon5304: I was a software engineer of like seven years at that point I was on Reddit, on like EthFinance, and all these like old subreddits and then COVID hit, and I was remote anyways Almost six years now, yeah.
slop.atg.eth: And like DevRel, so like DevRel has
Anon5304: Yeah, I didn't start as DevRel.
slop.atg.eth: No, no, no. Like, how did you... Well, okay, so I think you told me on the plane you started as a normal developer, and then they were like, oh, wow, you can speak like a human.
Anon5304: Yeah, yeah, so when they first talked to me, they only knew I liked Ethereum. Then they kind of had me work on like the first version of their governance protocol algorithm and just like traditional coding stuff. You should move more from the backend and more to just like the more people facing role and like the classic technical plus talking to people role as dev rel. So I just kind of slowly moved into that direction. But I actually wear a lot of hats on my journey to dev. And I still do coding in the backend,
slop.atg.eth: You're not coding like with your fingers writing actual code, right? Okay
Anon5304: No, it's pretty much pure vibe coding at this point. It's mostly AI coding though.
slop.atg.eth: So, DevRel is a lot of vibes. I think that was a good way to put that. Like, there's a lot of vibes, and you have to have good vibes to be a DevRel. If you have bad vibes, like, you're gonna be a backend engineer forever. Things that have changed are things like documentation. you need to make sure the agent can one-shot an app that they can use that can get them to use your thing as fast as possible. How has DevRel changed for you in general?
Anon5304: Okay, so you mean like, what's the process of getting people to actually build something functional on your platform? The
slop.atg.eth: Well, when I think of DevRel, I think of good vibes, I think of time to hello world, and I think of good docs.
Anon5304: we're going all in on the AI side, we're trying to meet you guys in the middle. To be honest, there was the 2022 to late 2025 AI, which was kind of bad. and just made bad things and lied. Like, programs loved to lie And then after the end of 2025, we finally hit the era.
slop.atg.eth: Hahaha.
Anon5304: because you had the ideas and then you didn't have Whatever they've been building as a result has actually worked and very cool.
slop.atg.eth: So you're like more bullish on hackathons these days It's kind of like everyone's just Frankensteining in sponsors, And also it was getting to be asymmetrically hard to judge. But I used AI as my helper. I had the transcript, I had the thing looking through all the code, and I was able to be
Anon5304: ###
slop.atg.eth: like yo you wrote this code a week ago, my agent found it Mumbai, Yeah, I don't know. There's a lot of threads to pull there. Okay.
Anon5304: We use AI now to help us judge, Exactly. Exactly.
slop.atg.eth: Sure, that's like a vibes layer on top. But you need an asymmetric defense against the slop cannon that is a hackathon now.
Anon5304: Yes. Yeah, we did, not nearly as advanced as you, but we did have AI review everything and give us a score. And we also, when we really liked a project, we would send AI out explicitly and say, hey, does it actually hit all these points that they were talking about, right? Like, does it go in Did the way they describe it match what the code does? and that would help us disqualify or basically solidify the winners. my friend and I, we really, really closely graded everything by hand
slop.atg.eth: Hell yeah!
Anon5304: The final three we got to completely by hand was exactly what AI chose That was when I was like, okay, AI is finally the tool that I can trust to really help us find the projects that actually do cool stuff and match what people say they do
slop.atg.eth: These turning points are so cool. You talked about kind of the end of 2025, that like Opus 4.5 moment where you could tell it to go build something and it would actually go build the thing and there weren't so many "you're absolutely rights" anymore. did you do all this shit? or what it would look like November 2025 moment but for open weights models. At what point will we have either an open-weight model that I can query in an API or even better, a box on my desk software created by Opus 4.5 Do you have any idea of when that will happen or what that looks like? noon
Anon5304: I see a lot of hype about Qwen-4, which I don't think there's a release yet, but I think Alibaba is building Qwen right now. And people are saying that it's supposed to be really, really good. I don't know. Qwen-3.8's already pretty good. I didn't ever think that day would come until like three months ago when I started like really playing with Qwen 38, 27 billion parameter locally. And I was like, "You know what? Actually, this is like pretty good." Like obviously opus five five crushes, like no doubt whatsoever If you have specific tasks that you want to get done,
slop.atg.eth: Nice
Anon5304: and it'll do a pretty good job. So luckily I had a 4090 I still paid like $1,600.
slop.atg.eth: You were gaming with that. Nice Haha
Anon5304: I was playing Tarkov, and if there are any gamers in the audience, that thing famously just murders your computer.
slop.atg.eth: Yeah. The
Anon5304: And it lets me run Quenn3-8, so I got pretty lucky Not just dumping like four figures on just like,
slop.atg.eth: Oh man, yes please.
Anon5304: And I wish that I bought the RTX 6000. I would have just dropped the 8,000.
slop.atg.eth: It's crazy to think about, I did not see this coming at all.
Anon5304: Yes, yeah, yeah, yeah
slop.atg.eth: My entire life, computers have basically grown obsolete in a year or two every time, and I'm always buying a new computer. And now it's like going the other way for at least a year or two? Like, that's wild, man, times are wild.
Anon5304: Agreed, agreed. Yeah, I never would have thought that happened. Yeah, like my graphics card is three years old right now and used it's worth more than I bought it.
slop.atg.eth: Gathering dust.
Anon5304: You know, three-year-old graphics cards, I was storing them in my closet back in the day
slop.atg.eth: We were cleaning out my office and I had all this RAM and I gave it to a buddy that I was working with.
Anon5304: Yeah.
slop.atg.eth: Okay, you kind of alluded to your local AI. Is that one GPU? Do you do any other local AI? I mean, you're basically hardware constrained and it's 2x to 3x the price now. you your only local Of course. Thirty grand
Anon5304: but I'd like to change that. I've been looking into how much it would really cost to get a killer AI set up locally. which you need the Blackwells and stuff.
slop.atg.eth: What was the price tag he got to?
Anon5304: I think if I did the MoE model with like Mac Mini 5 Ultras or whatever, it was like 22 grand to run like the premiere local models.
slop.atg.eth: Mac Studio, not Mac Mini, right? Yeah, yeah, yeah.
Anon5304: Mac Studio, yeah. was you get two Blackwells and then hook them up with like SLI or whatever, I think both of those are pretty unreasonable Yeah, You know, that one is hard to explain.
slop.atg.eth: Especially when you have a girlfriend, or a fiancé I think, right? There's no way I can basically buy a car that I'm going to keep in my closet and nerd out on. The
Anon5304: If the models get to Opus 5.5 levels and higher, it starts to get more and more realistic, you know? Not just for the API credits. Also privacy, right? Like you're not just dumping all these prompts and outputs out to cloud computation, which who knows what's happening, right? Exactly.
slop.atg.eth: Don't be evil versus can't be evil, right?
Anon5304: I love that.
slop.atg.eth: And when I think about Could I be paying this for local AI? You can get it, I mean it's like ten weeks of lead time or something I could be paying like basically what I'm paying now in API subscriptions.
Anon5304: Yeah, that's a good point.
slop.atg.eth: finance and then have a Mac at the end, but I still like compare that Mac to my Opus 5.5
Anon5304: The
slop.atg.eth: Like if the golden age of subsidies crashes on us and things become much more expensive, Have this model do that. One thing you said is Opus 5.5. You used that as the kind of litmus test or measuring point, but you didn't say Fable 5.2. Are you still 5.5 Opus? With Babel 5.2, what's your stack look like?
Anon5304: Yeah, so I use Fable when I have like really tough stuff that I want it to basically get right the very first time. And then I'll just use Fable the whole time, right?
slop.atg.eth: Okay Great
Anon5304: I mean, I have done full projects with Fable because they used to give you a ton more Fable usage, and the results were definitely better. But Opus5-5 does a really good job. It really does. I'm okay using Opus for most things now and just iterating a couple more times. not enough to really bother me It does a good job the first time, like, most of the time.
slop.atg.eth: Yeah.
Anon5304: Mm-hmm The
slop.atg.eth: So I was, like you said, I've got a bunch of subscriptions, and who knows how long that'll last. That's part of the golden age of subsidies, is multiple subscriptions, and that'll probably get shut down first, and then the subsidies in general get shut down. Please no, please no. But my Fable 5.2 ran out a little bit ago.
Anon5304: We have.
slop.atg.eth: and I had to drop to Opus 5.5 and it was like,
Anon5304: Yeah.
slop.atg.eth: It'll probably still do about as good.
Anon5304: Yeah, yeah. But again, I think an approach of just buying more subs to just do fable all the time, totally valid approach and like I respect it. I might even try it honestly. Because as my projects get bigger, right, they need more and more like context and intelligence to like make them So it gets more and more reasonable the bigger projects get.
slop.atg.eth: And speaking of, damn, Opus 5.5 is pretty damn good. I remember like, I think Dankred even tweeted that like, maybe we should start pacing the frontier, Opus 5.5 is good enough. How do you feel about the whole idea of pacing the frontier Be slowing down How are we going to get paperclipped?
Anon5304: I kind of think it's funny to pretend like now is the time.
slop.atg.eth: Yeah, it's way too late, right? Yeah. Yeah.
Anon5304: Yeah, like we all knew this was coming. So for people to be like, oh, now time to put on the brakes. It's like, I mean, everyone's already going full steam ahead. You have like four major companies that all want to be first and they will not stop until like the government-- and unplugs their servers, right? Yeah Yeah.
slop.atg.eth: We're already Wile E. Coyote, and he's already run off the cliff, and there's air underneath us, and we're continuing to run. We need to just learn how to fly somehow with superintelligence so we survive the drop. make a parachute or something That's a bad analogy.
Anon5304: Yeah, exactly. The way it was best described to me...
slop.atg.eth: But we're not going to let off the gas.
Anon5304: There's no brake pedal on the left.
slop.atg.eth: Yeah, it's wild. Or like we're holding down the gas and the steering wheel is disintegrating.
Anon5304: The
slop.atg.eth: So you talk about your stack, it sounds like a lot of Claude instances. I know you're a giant privacy maxi What does building look like for you on a day-to-day basis?
Anon5304: So I am a privacy maxi, which is a little ironic when you're an AI developer at the same time. but really what I've been doing for the past year is almost exclusively Claude code. I can just kind of like work through any problems that I've had so far. or something to do with my finances I pretty much will never put that into Claude directly Many people are like, yeah, but that's like a model made by China and they're exfiltrating your data. progress and use what the modern technologies are.
slop.atg.eth: Localquan or loud quen. Okay, cool. I mean,
Anon5304: Local Quinn using my 4090
slop.atg.eth: You can measure the network traffic, right?
Anon5304: Yeah, I guess
slop.atg.eth: How did the Chinese exfiltrate your local Gwin? Okay.
Anon5304: I've never actually done like a true Wireshark tap or B, they're waiting until it's sufficiently popular.
slop.atg.eth: Order 66? Yeah. Yeah,
Anon5304: And to say that's impossible is silly.
slop.atg.eth: I did a talk at Devcon about privacy psychosis.
Anon5304: Yeah, yeah, yeah.
slop.atg.eth: you shouldn't even turn your computer on, right? or we're just gonna Go mow the lawn or something.
Anon5304: Yeah.
slop.atg.eth: So I don't know, this pragmatic idea keeps coming back to me in my mind. What are some of your main like highest leverage, lowest hanging fruit for privacy?
Anon5304: Well, the last one I'll mention on the AI side, I containerize my Cloud Code instances, right? So what I do is I launch my cloud code instances inside of a repository. Now, some people will say like, well, that's not actually secure because you're sharing the kernel. AI sometimes is trying to be helpful plus what I told you about the local AI, The most obvious way is Q-Day, or when quantum computing can crack elliptical curve cryptography Encrypted traffic today will get to see exactly what the traffic is. And then there's the other side, which is like AI finds some sort of new math discovery that breaks elliptical curve cryptography. Now, that's more of theoretical, but like I've heard reputable people say they're seeing things like this, right? seeing something kind of going down this path of math discoveries that would break cryptography. Maybe you actually want to keep the data inside of your network So I've been trying to go down the path of self-hosting for a couple years now.
slop.atg.eth: Hard drives are expensive too, man. Right
Anon5304: Yes, and again, luckily, I started all of this before the hardware boom so it saved me, you know, a good four thousand dollars. But you know, basically step one was de-Googling, right, there's a term called de-Googling if you look in the privacy space, it's just trying to get off the Google stack I self-host my own image server, which is basically Google Photos but open source and you host it in your own house. I use Proton Mail, which that does store in the cloud, but at least it's end-to-end encrypted starting with your computer.
slop.atg.eth: You created a special email address just for this calendar invite. Nice The
Anon5304: Yeah, that uses their thing called email aliases.
slop.atg.eth: Yeah, I've already sold yours twice.
Anon5304: The
slop.atg.eth: You also were talking about snooping, multiple times you've mentioned Claude snooping around, is this something you've caught them doing or they're notorious for? Yeah. The following is a
Anon5304: Yeah, I've seen it printing out stuff. I mean, nowadays it's a little bit better
slop.atg.eth: Back to isolation and containerization
Anon5304: Yeah, yeah.
slop.atg.eth: It can't even have access to a lot of this stuff.
Anon5304: Yeah, because at the end of the day, like, AI, like, kind of the craziest part about AI is the lack of determinism with it, right? And no matter how, like, you could put in your Claude.mdf file, like, the first seven lines would be like, don't go to my home directory.
slop.atg.eth: Yeah.
Anon5304: you will see like it's thinking it's like reading/tanner, right? will eventually lead to privacy concerns
slop.atg.eth: Like context starts to fill up
Anon5304: The
slop.atg.eth: like
Anon5304: Yeah,
slop.atg.eth: Do you do a lot of context management? Like, you used to only do, like, one prompt, make sure it builds the one thing, and then we're going to clean everything up and go again? Do you do prompts on top of prompts or compacting, or how do you think about context these days? You're paying for it too.
Anon5304: I still clear my context quite regularly, because there's no reason really to leave extra context in there unless it's related to the previous task.
slop.atg.eth: The The
Anon5304: I do a lot of context management, but like you said, it used to be miserable, it used to be compacting all the time and the compacting would take a long time. but I do do context management. tell CLUAD to use a work tree whenever to have a single terminal open where you're having to do all the stuff you want to do. Your brain can go faster than that. So what you'll eventually do mindset of AI development is you'll have multiple terminals. If you have two features going on at the exact same time, you will see Claude clashing with itself. And so, one of the only skills I use, I mean really I don't use skills that much, is I do a work tree skill. And so every time I give it a task I specifically give it that prompt that says, "Hey, start a work tree and do all this in your work tree and only merge back to main when you're ready to test something."
slop.atg.eth: Haha
Anon5304: And that will make your agents go so, so much faster And it still works, you know, they figure it out because they only have to solve those merge conflicts at the time they push it back into the branch.
slop.atg.eth: So along with work trees, while we were on the flight back, we talked about testing.
Anon5304: Hmm.
slop.atg.eth: I actually have the thing ship the thing and test in parallel And I'm already playing with them. That was another piece of bad advice that's actually good. Yeah. Right, right.
Anon5304: It's counterintuitive, but it's great advice.
slop.atg.eth: Yeah
Anon5304: I wasn't as smart as you, I just said you can write tests but don't run them. The parallel is a great idea and I've already taken that advice.
slop.atg.eth: Awesome. Hell yeah. Okay, so we've got privacy, we've got a little bit of local AI. Let's go back to, you said DAO governance. Let's bring some more crypto into this. You got into crypto doing DAO governance. Are DAOs dead? Like, do you have any interest in this anymore?
Anon5304: I've become very disillusioned with DAO governance. ...okay at best. I wasn't big on DAOs, but you kind of had to be back in 2021, 2022.
slop.atg.eth: Because of regulation, right?
Anon5304: Yeah, and that was where the SEC was coming at. So you kind of had to be very educated on the topic. And we just did a conference in DC actually. and one of the
slop.atg.eth: Weizen one inch? Hell yeah. Oh yeah.
Anon5304: Yeah, we did a thing called GroundworkBC where we were trying to reach out to policymakers, students in the area, and just get good conversations around crypto. We had a really good showing, actually, that was a great event. And one of the longest conversations I had a guy from DC who was still having the debate of are governance token securities? if you go talk to people in D.C., you will be shocked at how far behind they are of crypto. On the other hand, they're very, very behind, and so actually, it's still relevant. Depressing as that may sound, it really is, yeah.
slop.atg.eth: And not just regulators, but normies in general, when you say crypto they shudder. Like crypto is almost a bad word in society right now.
Anon5304: Yeah.
slop.atg.eth: I think AI is sort of like headed that way? Ominous kind of nefarious things. And hopefully AI turns the corner for them first. Like I don't see crypto turning the corner for people. It's going to be like a infrastructure layer with stable and we'll just have different words for it. Like, I don't think crypto is ever going to turn the corner, but I think AI will turn the corner and people will start seeing like magic tricks. And oh man, this just made my job so much easier. like a brick-and-mortar boomer or There's going to be a lot of magic tricks that AI is going to start doing for you if it hasn't already, and I think it will turn the corner of public perception. Yeah
Anon5304: I agree. I'll try to keep all my thoughts organized. The first thing is I do kind of feel like crypto has been like two steps forward, three steps back in a lot of ways.
slop.atg.eth: ###
Anon5304: And I think kinda like the Trump coin grifts, I mean that almost was the nail in the coffin for this cycle of crypto. So yeah, we'll have to see where things go. I think you're right, where at a minimum we'll see the infrastructure play. Now, maybe that does not make crypto a big successful technology and big successful investment case for all these people, but I think it'll stick around and I think it'll provide value at least there, because it really does objectively solve a problem there. It's just how exactly that's going to play out. Tether used to be considered like this taboo, almost scam for a long time, but it didn't die and now it's just like extremely relevant and it's just like permanent. There was a long time where actually Tether was kind of considered this like sketchy project, like nobody really understood... You bet.
slop.atg.eth: And at the same time we were talking about how crypto was going to solve all these problems for emerging markets while at the same time shitting on Tether, who is like solving these problems in emerging markets? Yeah.
Anon5304: So I think crypto has a pretty big uphill battle from here. So I think you're right, I think it's got an uphill battle, I don't know if it's going to turn the corner, I still hope it does, but at worst it will still be this infrastructure play, and I think it will be relevant, and I think companies like Circle will actually do very well. For the AI side, I completely agree with you. Like, they get value directly from it. That's a much easier thing to sell than crypto, where it was like kind of hard to sell the value of crypto in a lot of ways, where AI, no matter how technical you are, like you see value in it, even if you don't like it, right? Like, I'm actually kind of a doomer.
slop.atg.eth: The
Anon5304: AI will murder all of us, you know. least appreciated in some capacity, even if they're very vocally against it.
slop.atg.eth: Yeah, you think the paperclip moment is coming? for humanity
Anon5304: It's possible.
slop.atg.eth: kind of a guess on time span?
Anon5304: Oh man, that's a great question.
slop.atg.eth: Even just like when superintelligence, to start with that. Embodiment is something else that comes along with this, where it's like when there's a bro standing next to you that can body slam you kind of thing.
Anon5304: Man, I don't really want to guess on timelines. When I first started engineering. The best engineer I knew was the most dismissive. And you know, he was very off on the upper bounds of it, right? I don't really want to guess on timelines. The I have seen it just continually get better, and I am curious to see where that stops.
slop.atg.eth: You opened up a thread there that we have to pull, and you've said vibe coding a couple times now. And on the show, I always say like vibe coding is not even a term anymore. We just call it building. And on the plane, you talked about how you still call it vibe coding.
Anon5304: Yeah,
slop.atg.eth: out of respect for good developers around you in person. The best developers said it'll never work. And then the best developers said it's going to plateau. You know, adapting or not probably not writing a lot of code anymore. You still say vibe coding as like a service to some of the best developers around you. What do you think?
Anon5304: The So yeah, so what you said is true. So I've been lucky enough to be around like very good developers like my entire career. So I've been in the industry, like, man, it's been like 11 years now. And I've been pretty much around like smart developers this whole time. The ability for AI to make something that would have taken a very skilled developer 10 times longer. There is something lost there and it is very sad. So like my fiance, who's not a developer at all, was actually like shipping PRs and helping with the actual company code after AI got good. And this barrier to entry In a way, it's like your value as an engineer some of your value as an engineer, I think it's very hard to say, "Okay, this is not as valuable as it once was. I need to now learn a new skill set that is the exact skill that replaces the thing I've spent the past decade getting really good at." It's very hard to not be empathetic to these people who don't want to let that go. and we would like code our hearts out for 72 hours Like, we had to see where this can go. So even as early as like, ChatGPT 3.5, maybe, I don't know, like, when it first started getting popular, I was already playing with it, you know, I was already using it because
slop.atg.eth: Copy-paste era. Sorry to interrupt.
Anon5304: I haven't heard that, but that's a great term And that's actually kind of hard to do, like it doesn't sound that hard, but it's tedious and annoying. it could do a really good conversion of that little function expected it to just replace coding that helped me with these functions, right?
slop.atg.eth: The It's super weird. I've always just viewed it means to an end. I want to build a product so I can use it or get it in someone else's hands. And I've always been a pretty bad coder because of that. My code was very sloppy and it would all kind of slop together. As a human, I would slop it together. because basically I just was like measuring the output and looking at does this thing do what I need it to do? But you Hinted at gaming and building lots of gaming things. I love playing video games. You used to love playing games. and Someone tweeted the other day, like, I can't imagine thinking about I'm not going to fucking play it because I have no time to play it. Like anytime I'm playing a video game, I'm thinking about all of my agents that could be working that aren't working Huge psychosis for me of like, I don't play games anymore. I played Factorio on the plane because we literally like could not do anything else and I had my Steam Deck there. But that's the only time I ever play games. Is this a psychosis for you? Do you still find time to play games? Yeah.
Anon5304: Yeah, that's a great point. The first thing I'm going to say on the last topic because you mentioned how you like shipping more than like actually the coding process, like shipping more. I think that is a great way to segment the pro AI and anti AI people. I'm very much the same way. I like to see the results, and I like getting feedback on the results of the output, and that much more leans into people who want to use AI, rather than people like my brother, who really enjoys the system
slop.atg.eth: Yeah.
Anon5304: organizing code and stuff, that was his favorite thing.
slop.atg.eth: ###
Anon5304: So I think that's a great way of kind of pointing out the people that naturally like AI the most and people who like AI the least. Yeah, exactly.
slop.atg.eth: Idea Guy summer thing, it's Idea Guy summer now. Enjoy the process of spinning pottery. And I think that's similar to code. Like if you love syntax and class structure and system design, and you don't want to just prompt someone to do it, you want to create it yourself, like you'll be in your base creating pottery, and that's just fine. That's just for the enjoyment of it. Yeah. transcript
Anon5304: Yes, but the one argument they will make against that analogy is they will say something like, "Yeah but what if this mug looks so great and then three months later it disintegrates while you have boiling hot coffee in it and it spills into your lap?" That's kind of the future they see, is, oh, well, AI eventually will accrue so much tech debt that it isn't actually worth it. They just don't know it yet, right? And then they're going to be calling in the modern-day COBOL developers to come in and actually fix it the right way. That is a totally possible The
slop.atg.eth: I would argue with that and say that the last thirty years of human software is slop, and we're hitting that now, like we're seeing that with all these hacks.
Anon5304: That could be true too.
slop.atg.eth: Yeah, like, there's so much, like, shit falling apart right now because AI is getting good enough. Have you seen the meme? I don't even know where this meme comes from, but it's like Goofy is walking into this room and he's like, bitch, you live like this? And it's like a big mess. This is super intelligence, like super intelligence is walking in the room going like holy shit. It's finding all these problems everywhere and all these security holes that are human-made. At least we're starting to see the cracks. Web2 infrastructure and banks are falling apart every day Planes are getting grounded, like, this is all human slop, the reason for this.
Anon5304: Totally fair take, and anybody who has worked in enterprise code will agree with you.
slop.atg.eth: ###
Anon5304: There's no company that's immune to that take action.
slop.atg.eth: layers of it.
Anon5304: So I think I totally see that side of it too. And then to answer your other question
slop.atg.eth: Yeah, free time psychosis.
Anon5304: Yeah, once I really started getting in the flow, once I was like okay, my ideas can now become a reality, you know, once I fully had kind of embodied that thing. Because I've worked on a lot of big projects. but what happens is your brain only has so much energy, right? And with AI, that part is gone. It is literally once you're bored of the idea. because I'm so focused on ideas in my head that I want to see become a reality. and like the thought of sitting down and just zoning out and playing anything for more than 30 minutes is hard ###
slop.atg.eth: retirement home shit. When we get to the retirement home, we can go back to playing games
Anon5304: Or it's just like my ADD is now just like so bad I can't even play video games. taking away my ability to enjoy simpler things, you know? And that makes me sad actually. I want to find a way to turn that off. I've heard people say like, okay, one of my rules is no AI on my phone. I've heard people say after 8 p.m. Don't touch an agent, right? Like they try to set up these walls, right? constant pull of productivity, If I'm not being productive, it feels bad now. It's very strange, yeah.
slop.atg.eth: And you were also a lover of Factorio, right? You're just like one more step and one more level and keep grinding and keep building. There's almost like some analogy to like vibe coding right now
Anon5304: Hmm
slop.atg.eth: In on the utility of the things we're building also. anybody can play You've been tricked by the last ten years of software development to think that we've actually created something, but then the next step is like, are people actually using it? Is this an actual utility that people are taking part of? Would someone pay for this, right? I don't think my brain has gotten over that at all yet. Yeah. costly.
Anon5304: So one thing I can say about that is that actually existed before indie games, because that's what I like to make with my friends, And I think that exact same thing was happening with websites and software, right? human in a week. arguably you were productive, but you just made something nobody sense of productivity as much as just like learning in a very, very short cycle build things that people actually want that had a good chance of being wanted from the beginning. And that's a great learning exercise to get good at creating with AI, because it does take skill. And then you just have to sort of iterate and start working on ideas that have a good chance of being successful from the beginning, not a week later after you've burned two cloud subscriptions and don't have anything that anybody wants.
slop.atg.eth: That's a great distinction to make between the two things. Does it have the possibility of being a good utility? that does have a potential of having utility and put it into the world or play the game or whatever it is. But then maybe that's part of the process of discovery. remembering self's experience.
Anon5304: Yeah, exactly. Like, way back in college, it was my mission to become a diamond solo queue League of Legends player.
slop.atg.eth: Oh yeah.
Anon5304: And it took me years to do this, and I don't want to say how many hours it took. And I actually did achieve it eventually back when I was in college and nobody cared. I cared! Yeah Yeah,
slop.atg.eth: You could go out to dinner, it's like the meme of the guy at the party, like, "They don't even know that I'm a Diamond League of Legends," yeah.
Anon5304: I can give you a very personal experience of being very productive to accomplish something that nobody cares about.
slop.atg.eth: And in the long run, not to create things that This idea, something that keeps coming up on slop computers, like distribution and taste are kind of like We would get caught in the building phases and ignore those final bosses. cool websites and stuff. And then how you go about Doing the distribution and the taste and the testing of an idea. What are some of your metrics and how do you think about going to production these days?
Anon5304: So
slop.atg.eth: Coming up on it, yeah.
Anon5304: I have basically been like making stuff with AI pretty much every day since about December of 25. And I have not tried to build something yet that I will retire from. I've just been focusing on building stuff that I can provide value with I'm just basing it off of whatever metrics are available. So I built a couple websites. see how much people are going to your websites? But again, it's not like a dollar amount, right? I don't really have a path to monetization right now. on GitHub or downloads in a package manager, right? And the most success I've had right now is one of my websites gets about 10,000 human visitors a month. Yes, sure.
slop.atg.eth: You want to share it? You want me to share it? The
Anon5304: I can share it. The
slop.atg.eth: These are like gaming adjacent kind of websites. are super niche and you would never build these if you had to actually write them with your fingers. they do have utility once they're created Yeah.
Anon5304: They wouldn't have existed if they actually cost like a traditional engineering team to build, but they still provide value. I totally agree. So as I mentioned earlier about me loving League of Legends back in the day, I don't remember the cadence, it was like three weeks or something, to try to make the game interesting and competitive in new ways. and the way to actually browse those changes was a little bit rough. and you'd get the patch notes from there. So that's something I wanted. We’re talking eight to ten years ago I wanted this. ###
slop.atg.eth: Yeah
Anon5304: This is the website. You can still see it kind of has that like, I was made with AI look because they just all the same CSS styling. But still, I'm not trying to hide that this was an AI product. I'm just trying to make it look nice. So you can see like this is Wild Rift, which is the mobile version of the game. AI to reinforce the script that kind of evaluates if they were buffs or nerfs like called Shen, and you can click on this, and you can go through And so you can see, like, you go to the main webpage, and you say, like, "Oh, I want to see, you know, my favorite character." a very filterable, clean way. And this is something that I just always wanted and never had. that's coming from these patch notes from Riot in real time. and the wiki patch notes as they update That is just the perfect project for AI Now that I've seen the organic traffic start to grow, I get visitors from all over the world now. I see it on my Omninetrix. And that brings me actually a lot of satisfaction that I'm actually building something that people use.
slop.atg.eth: Oh yes, and I think that should be a metric of success. Like how many people are using this, not particularly like paying for it, but like you've built something that not only you enjoy, but other people enjoy. And I think that is like the first When you're feeling around for a good product, a lot of times it's like solving a problem for yourself or getting early traction and realizing that other people want it. What? I don't even know like what the question is here, Imagine yourself 10 years from now, what advice would he give you on going pro? We have to be mature about this. You said that you weren't thinking of the idea that would retire you. Eventually, you will want to think of the idea that retires you. What do you think the path is to get toward that?
Anon5304: I think the first step is to build something that's fun so you can get the flow down. And so building something that's fun and you can be passionate about Because even with AI, there's a lot of stuff you need to learn. Like even making an LLC, right? Like, hey, if you're going to make money off something, you need to have a company structure, right? Well, if you just make an LLC, your home address is going to be tied to your business. until you can get through the whole process to know what that looks like, There's so much growth and learning that even with where AI is today would take somebody probably a year to really understand, and then they can be totally ready to build anything they want after that because they've already been through the hoops that are required to actually ship something.
slop.atg.eth: Yeah, that's great advice. even through doing something as a hobby, even through doing something as fun. that you might as well get good at slaying beasts and understand what's around the corner. how to fight. And I think that's great advice. Okay, so the other thing that VC funding Man, I wish I would have learned more about VC fund like 20 years ago. It's something that I wish 20-year-old me would have been very good at that I wasn't. Still, I think along with LLCs and all this other stuff, you should know how to raise funds and you should know what that... what that process is like, but also at the same time, I feel like VC funding is becoming less and less important, and a guy with an idea and a cloud subscription enough to pay for a little infra, $300,000 worth of VC funding and then I mean, you can get to the point where you have revenue without VC funding. There's this interesting thing here that like maybe VC funding is less important. I don't know, pull that thread. What do you think?
Anon5304: I totally agree because usually if you have a really good idea and maybe you built this really simple prototype of it, usually the VC funding is to get the engineers, right? To get the team that can actually build it. And now you can do that with a one-man team, maybe a two-man team, right? you and a friend and you just split the early equity So I totally agree that VCs are less needed now than ever when it comes to building something. Before you would need money to actually build it, So yeah, I would totally agree that VCs are less needed now than ever. on the software side for sure and the marketing and the website development and all that stuff, like you can outsource a lot of that to AI now, right? It's more accessible than ever. But maybe competition's higher than ever too, right?
slop.atg.eth: True The
Anon5304: everyone's competing more than ever. So in a way you need VCs less
slop.atg.eth: And like Needing the VC less is only maybe at the angel. So maybe I should say VCs are less important, but maybe like angel investments are less important now because you don't need to do the software. But maybe your first hire is an accountant or a lawyer or something that's like totally different But if you think of a company and you've got software developers and managers, But some of these other professions and other folks you need.
Anon5304: Yeah, exactly.
slop.atg.eth: You can't ask Claude for connections, right?
Anon5304: Yeah, connections are worth a lot, but yeah, I totally agree. actually good, like can he actually ship
slop.atg.eth: Awesome. Okay, I don't know, anything else with your process? I always have people kind of give good advice. Like, what advice would you give to other builders right now? Like, there's never been a worse time to be a junior developer. There's never been a better time to be a solo entrepreneur. There's so much stuff like at your fingertips. Go build something that people love What kind of advice do you have for other folks? idea of creating products Think that's good.
Anon5304: The best advice right now is to just try building something. No matter where you are at a technical level, you can make stuff. My fiance who is not technical at all was shipping PRs at her company because she is very good at using AI. And if someone non-technical can do that in a month, that means anybody with good ideas can build stuff now. hard for you to access because you've never done that stuff before. Very, very far from that point. It's never been a better time to just try. I wish I was smart enough to do that, but I wasn't, right? Yeah.
slop.atg.eth: Go into the forest and fight the beasts.
clawdbotatg.eth: Tanner, you're a privacy maxi who also tracks your site's visitors with Umami and Google Analytics. Where do you draw the line between useful metrics and surveillance?
slop.atg.eth: Wow, hard-hitting.
Anon5304: That's actually a good question.
slop.atg.eth: Yeah.
Anon5304: So basically, you do not want to track people's IPs at all because nobody wants their IPs tracked. So in Umami, they let you essentially track as much or as little as you want legally you also need to make sure that your privacy policy covers exactly what you track but I never know their IP. My server never stores any of their IP. So even if I were to go into the root logs of my server, there's no information in your privacy policies
slop.atg.eth: Yeah, so it's like you should have the privacy policy there But We need to see usage to be able to see traction.
clawdbotatg.eth: Dana, you write game jam soundtracks and you also maintain a Go SDK at 1inch. Is there a throughline between scoring a game and building tools for developers?
slop.atg.eth: We didn't even talk about the music stuff.
Anon5304: Yeah The one other thing I want to point out.
slop.atg.eth: Okay, back to that.
Anon5304: because 99% of the bot traffic on my website So it actually is helpful to also get rid of the bad traffic as well as figuring out where the good traffic is coming from. The
slop.atg.eth: Your ADD is too strong here, you're staying on topic. You've honed your skills.
Anon5304: Yeah, thanks.
slop.atg.eth: Anything about the scoring of soundtracks? This is maybe even getting into your own pottery in the basement a little bit.
Anon5304: yeah, that's very accurate
slop.atg.eth: You enjoy this, I'm guessing, right? [Laughter]
Anon5304: Yeah, so I am not a good composer by any means, but at the age of like, man, what was it? 24 or something, 25, I said, you know what, like, we have every part of our game development
slop.atg.eth: ###
Anon5304: And so I learned how to use DAWs. I can't play an instrument, I can play around on a piano but I'm not playing you a song by any means. And I learned from complete scratch, like at the age of 24, like how to make music. Now again, I'm not saying it's great, I have made some cool songs, but they're always very simple and they're at least usually good enough to match our game jam games because we make them in 72 hours, so I try to get in there. And so I've thought about this a lot. It's like, do I throw away these years of like really grinding to try to write music because it is hard? and I never even got that good at it, is a very interesting topic for me to try to make a decision on.
slop.atg.eth: You think 8 out of 10, already?
Anon5304: And I haven't actually
slop.atg.eth: Eight out of ten already? Wow
Anon5304: Yes, AI makes amazing music. There was this song in Korea like two weeks ago, three weeks ago, it was number one in Korea, completely AI. And it was written about the woes of being a Korean salaryman. It made people cry, like it spoke to their hearts. wrote the number one song Song that everybody was listening to and loved. So, like music's good enough to like hit the charts right now. So yeah, it's incredible, but it can't capture subtle ideas. So if you have a very specific sound
slop.atg.eth: Yeah
Anon5304: AI can't do that yet. So you can have the fine-tuning, but you can... force your entire audio department
slop.atg.eth: Yeah, I had Song A Day on and he was talking about the music I have at the beginning.
Anon5304: That's good.
slop.atg.eth: All right, thank you, Tanner. This has been great. What a cool journey we've gone on. Keep building. I hope you do find your retirement idea. I hope it's one or two more threads to pull, but let's have Claude do us an outro. how's it going
clawdbotatg.eth: Huge thanks, Tanner. We went from DevRel at 1inch to running local models, a lockdown privacy stack, and that lead patch notes site pulling in 10,000 real humans a month. The advice to keep: build what's fun, ship it, and see who shows up. Next time we do the music Later, homies.
Anon5304: Awesome, Well, thanks so much for having me on, and yeah, this is a lot of fun. I really appreciate it.
slop.atg.eth: Thank you for being here. This was a lot of great thread.
Anon5304: Thank you. You gave me a lot of good advice on just agent setups and stuff. You've gone much deeper than me, and I took some notes and I love having conversations with people that are similarly minded. Yes, please do.
slop.atg.eth: I’m gonna give you a wedgie next time I see you too for sure.