148: AI Psychosis Day  🔗 **Source:** [148: AI Psychosis Day](https://www.youtube.com/watch?v=UBMBSzLSIXM) > [!note]+ Document Notes > --- > tags: [ai-coding-agents, model-capacity-economics, terminal-automation, generative-worldbuilding, developer-culture] > author: [MISSING] > source: Mostly Technical (YouTube) > reference: https://www.youtube.com/watch?v=UBMBSzLSIXM > date created: 2026-09-07 > --- > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > # 🏷️ Title: 148: AI Psychosis Day > ## 📝 Summary > - Episode 148 of *Mostly Technical* (a two-host developer podcast; one host is addressed on-air as "Ian," surname likely Landman/Landsman per an aside about his Instagram handle — the co-host is never named in this transcript) is an unstructured Labor Day catch-up, not a scripted segment show. > - The title "AI Psychosis Day" is the hosts' own jokey label for a day spent heavily using AI coding tools, coined on-air by one host: "this is teacher inservice AI psychosis day. You don't need to be actually doing work. You need to be ripping Astra tokens" (~[02:45]). This is casual internal slang for token-grinding, not a discussion of the clinical or popular-press concept of "AI psychosis" (chatbot-induced delusional thinking); the episode does not engage that topic at all. > - Content covers: burning through "Astra" (a new frontier coding-agent tier) usage quotas across multiple accounts; a personal, non-commercial art project using AI agents plus Blender to build a fictional world and narrate audio stories; porting a terminal-orchestration tool ("Solo") to Linux via the Omarchy distribution; training a small local classifier model to detect whether a terminal session needs human attention; a theory that Anthropic's perceived quality dip versus OpenAI is a compute-capacity/rationing problem, not a model-quality problem; and personal-life segments (a daughter's horse-show weekend, camera gear, an upcoming Vegas trip). > - Stakeholders: the two hosts and their listener base of practitioner developers who use frontier coding agents daily; sponsors (Svix, Bento, Typesense, Valor Security, DropInBlog) get read-in ad spots. > - Risk/tradeoff surfaced but not resolved: one host describes burning "the equivalent of eight full quotas" of a subscription tier and openly frames this as recreational rather than productive ("sending generous amounts of tokens for exactly zero ROI"), while also describing genuine tool-building work (the terminal classifier, the Linux port) done in the same sessions — the episode doesn't distinguish playful exploration from paid work time, and neither host treats that blurring as a problem worth examining. > - Bottom line: this is a casual industry chat, not a mental-health discussion — "AI psychosis" functions here purely as slang for intensive personal AI-tool use, and any reader encountering the title should not read it as reporting on or endorsing claims about AI-induced psychosis as a clinical phenomenon. > ## 📖 Main Content > The episode has no formal structure; it's a recorded conversation between two developers who each run small software products alongside full-time jobs. Four threads carry the substantive (non-personal-life) content: > **1. "AI psychosis day" as slang, not clinical claim.** The phrase that titles the episode is introduced once, early, as a joke about spending an ostensible work holiday running AI coding agents instead of doing paid work (~[02:45]). Neither host revisits the phrase as a topic — it never becomes a discussion of the reported phenomenon (widely covered in tech press through 2025–2026) of chatbot conversations reinforcing delusional or grandiose thinking in vulnerable users. Because the episode borrows a contested, non-clinical popular term as in-group slang, any downstream reader or listener should be told plainly that this source does not address that phenomenon — the title is misleading if taken literally. > **2. Token-quota exhaustion as running bit.** Both hosts describe running out of usage allowances on "Astra" (their name for a new top-tier coding-agent offering) across multiple accounts, using "banked resets" (extra reset tokens that pull the next billing cycle forward rather than adding free capacity). One host estimates using "the equivalent of eight full quotas" over a weekend (~[03:01]); the other separately estimates roughly ten quotas' worth of usage across four accounts (~[61:12]). These are the hosts' own rough, self-reported estimates, not measured figures — treat them as color, not data. > **3. A non-commercial worldbuilding project.** One host describes a multi-month side project: an AI agent swarm builds out "canonical facts" for a fictional lost-civilization setting, uses Blender (which the host says they had never touched — "I said, 'What's Blender?' and had the agent download it and start using it," ~[09:33]) to render consistent 3D character and location studies, and pipes generated short stories through ElevenLabs for narration, which the host listens to "as I fall asleep at night" (~[13:25]). The stated goal is an eventual physical illustrated book, openly framed as an AI-agent collaboration rather than hidden authorship. No claim is made that this project is finished or commercially viable; it's presented as an ongoing hobby. > **4. Terminal-state classification and cross-platform porting.** The more clearly work-relevant segment: one host describes training a small local model (final size ~537 KB, reported P95 inference latency under 10ms even on dense scrollback) that reads a terminal's visible character grid plus internal diagnostics (last PTY byte seen, last human keystroke) and scores whether a given AI-agent session is idle, blocked on permissions, or needs input (~[50:56]–[58:41]). Labels for the initial ~500-sample training set were generated by prompting a frontier model (referred to as "codeex," i.e., a coding-agent CLI) to score each captured sample — described on-air, unprompted, as possible "distillation." Separately, the same host describes porting their terminal-orchestration tool "Solo" to Linux via System76's/DHH's Omarchy distribution, reachable over a Tailscale mesh network, with "10 to 15 little things" (~[35:19]) needing fixes before release. > ## 🗒️ Notes > #### ⏱️ [00:00](https://www.youtube.com/watch?v=UBMBSzLSIXM&t=0) Cold open and sponsor reads > - Labor Day framing; both hosts joke about recording instead of taking the holiday off. > - Sponsor reads for Svix (webhooks-as-a-service) up top, described as their "first ever top of the show pre-roll." > #### ⏱️ [00:42](https://www.youtube.com/watch?v=UBMBSzLSIXM&t=42) "AI psychosis day" coined > - One host names the day "teacher inservice AI psychosis day," meaning a day for using AI tools recreationally rather than working. > - Immediately pivots into quota/token talk — the phrase is not revisited as a topic afterward. > #### ⏱️ [00:52](https://www.youtube.com/watch?v=UBMBSzLSIXM&t=52) Banked-reset mechanics > - Explains that a "banked reset" now pulls forward the next billing cycle rather than adding free usage — a policy change one host calls "the worst." > - Both hosts describe near-exhausting personal quotas across multiple accounts. > #### ⏱️ [05:14](https://www.youtube.com/watch?v=UBMBSzLSIXM&t=314) The fictional-world art project > - Origin story: started by asking an AI assistant about open math problems, pivoted to "what are fun things to do... with many agents working together." > - Describes the agent-driven canon-building, Blender rendering pipeline, and ElevenLabs audio narration workflow; explicitly says the eventual output would be released as an "AI writing" collaboration, not hidden. > #### ⏱️ [18:03](https://www.youtube.com/watch?v=UBMBSzLSIXM&t=1083) Solo on Linux > - Describes porting the "Solo" terminal-agent-orchestration tool to Linux via the Omarchy distribution, accessed over Tailscale/VNC from a MacBook. > - Frames Linux support as "completing the story" for a headless, multi-machine future rather than a near-term revenue driver; explicitly says "I still not a believer... this is the year of desktop Linux." > #### ⏱️ [49:02](https://www.youtube.com/watch?v= Hello. >> And on this Labor Day, we labor for the people. We have to labor for the people. But you know what the people need to do for us? >> What's that? >> They need they need to visit.com. In the age where code is cheap, you might be tempted to build web hooks inhouse. Don't do that. solved all the hard problems about scaling web hooks so you can avoid triaging failed deliveries and incrementally building features. Check them out at svixx.com. Uh or you can go to the link down below which is link.spix.com/mostlyte. Thank you to our first ever top of the show pre-roll. How's your father sponsor? How about that? Huh? >> Pre-roll. Thanks. >> Really going out of their way. uh to be the first ones. They were excited to do it. So, it's really cool. So, thank you so much. >> We haven't even started fighting yet and we've already got an ad readin. It's perfect. >> We made money. We're good. >> That's crazy. Happy Labor to us on this day. >> We do deserve it. We are here on Labor Day. Other people are out barbecuing, doing stuff. We're not doing that stuff. We're right here. We're making entertainment for the people. Laravel Holdings Incorporated, US division, has the day off and I am here in the office making content for you people. We're both in our offices. We left our house. We're not just like we're recording from a walk somewhere on our phone. No, we're in the studio for the people. >> I I could be What time is it? It's It's 8:35. I could be four banquet beers deep on this Labor Day, you know, as I probably would, you know, what else do I have to do, right? Um, >> I got permission to have half the day off, not from Larville Holdings Incorporated. I, you know, I serve no master except for Jennifer. So, uh, after after kids rest time, uh, I will be going home and we're going to all go over to grandma and grandpa's house. And so a bunch of the sisters and all the cousins and stuff are going to go hang out. So here I am in the office catching up on stuff and talking to you, Mr. Ian. What is going on? >> We sort of have a tradition of like never doing anything on Labor Day. >> We just never do anything on Labor Day. We don't do parties. We don't do grilling up. No. >> Yeah. I don't know. It's just not a thing. So >> yeah, I don't know what I'm I have like I'm working till like 2. Do I have a phone? I have like an actual business call at 1. I got Yeah, it's we're just a work day pretty much. I'll head home after that. But >> no, this is this is uh teacher inservice AI psychosis day. You don't need to be actually doing work. You need to be ripping Astra tokens is what you need to be doing. >> Yeah, we got we got some new AI hotness. I wonder how many accounts you've added because of Astra. You're just loading up additional accounts. >> I've used so many bananked resets. I've basically burned through like the equivalent of eight full quotas of Astra. >> Dude, I don't understand how you use it this fast. Like I can't even really get through the one. Like occasionally I do if I'm really grinding. I do now. I think I have two bananked resets. So I don't know how I'm going to like I got to like artificially find a way. I reset like tomorrow or something though. So >> I got to like But the other one's due in two weeks. So I got to find a way in the next week or two to like artificially inflate my token use. >> That's the worst. I know it. I think it used to be that the banked resets would just reset your usage and not your like uh your reload date, >> but now it's if you use a banked reset and like let's say uh by the calendar you would renew tomorrow and you used a bank banked reset today, then by the calendar you renew in seven days. >> Okay. >> So, you're basically just pulling No, I want I want 100 I want 100% more. in the middle of the week that I can burn in three days and then reset on the third. >> Um, >> so >> they got you. Yeah, they just kind of pulled it forward. Oh, that's >> so deep in it again. This feels like this feels like November uh 2025 again for me. >> Oh my god. All right, so let's let's jump into that then first. So you've done some solo grinding, I assume. I assume you're not using all these tokens in your job as vice president of marketing. I'm sure you're using some tokens there, but you >> you can't grind that many tokens on marketing, I don't think. So, >> no, I am. >> You're in deep on some code. >> I've got some I've got some grinding going on um just on on on personal tokens on isolated projects. Yes. Um >> so, if if Andre you're listening, I'm not using I'm not using these personal tokens against Larville Holdings Incorporated repos. However, I have spun up a few experimental repos that will uh be given to Laravel Holdings Incorporated, but I'm definitely not uh if the sock two auditors are listening. >> Oh, I see the connection type stuff. >> Yeah, I'm not I'm not hooking it up to the cloud repo and letting it rip. So, everybody just don't clutch don't clutch your pearls. Yeah. Um, so I am doing a few pretty meaty Laravel holdings experiments just to see if I can make cool stuff work, but most of the >> most of the tokens this weekend uh definitely went to solo and a secret project that is purely art. It's not software. We're not going to make money off of it. >> This is this is Aaron wants to have some fun with the gods. And so we're building a little just a little art project. That aside, um, >> hold on, hold on. That aside, we're going to that aside. No, I yada yada. The best part. >> First of all, >> like this guy gets a salary. All of a sudden, he's like, you know what? Let's do some AI relaxation. Let's do some AI. Life is beautiful. Yeah, we can relax sometimes. Wo, it's crazy. >> Look at this. Okay. So, >> there's a chance I've been in fight or flight mode for many years. It's unclear so far, but there's a chance that that's possible. >> Yeah. Uh, all right. Give us a hint. Give us like a Is it Are you generating something that like people can output? Is it just for yourself? Is it going to be public? Like, give us >> It will be It will be public at some point. Um, I've been working on it for >> actually pre pre-W2 employment. Um it certainly has gotten a lot more fun now now that uh >> doesn't feel like you're wasting tokens now. >> Yeah, exactly. Um it is it is first of all first and foremost it is an experiment. >> Okay. >> Okay. So, uh, a while back I saw, you know, all these ma open math problems were falling by the wayside to AI, >> and I was like, I have AI. Uh, >> I can do this, too. I can play this game. >> Yeah. Uh, so I talked to Claude about what, you know, open math problems there were, and I quickly realized, oh no, I I can't do this at all. There's like less less than a 0% chance I can like play in this area. And so then I started talking to it about like, hey, I've got this I've got this meta harness that can coordinate these agents and these swarms and like what are some fun things to do that are now possible uh with many agents working together that wouldn't have been possible before cuz that's the whole math thing. It's like ah let's just throw compute at it and this wasn't possible before now new things are possible. And so after, you know, an hour or two of talking and pushing back and accepting and we we landed on um basically a fictional world. And so I am with the help of a whole swarm of agents um building out an entire fictional world. I will not tell you the details of it. Um, >> but it is going to be I think somewhat of a uh novel concept and the final the final artifact will be something that you can read. Uh, of course there will be like, >> you know, micro interactions and stuff, but it it's primarily >> it's primarily actually targeted to be a a physical book, like a >> a giant beautiful illustrative physical book. Um, >> and of course this weekend I was like, "Hey, Astra, what's Blender? I see that everybody's doing Blender. we have this whole corpus of it's like building up um canonical facts about the world over time and so now we have this whole corpus of canonical facts about this fantasy world and kind of how it you know fell >> put you out as a fantasy guy interesting >> I don't know I don't know if it's pure I don't know if it's fantasy fantasy like there's no >> maybe it's more fantastical than fantasy cuz there's no like warlock blocks or or dragons or anything. Um, but it is basically the story of a lost civilization with a few, you know, twists and turns. >> And so, uh, I fired up Blender. I didn't fire up Blender. That's too generous. I said, "What's Blender?" and had the agent download it and start using it. And now it's building out 3D uh, representations of all the canonical details that we have. Oh my goodness. Look at this guy. >> Blender's cool. Never used Blender in my life. Still have it to be fair. >> But once you build up all these like 3D like basically sources of truth, you can render them in different ways while still keeping all the details correct. Right? Cuz that's the problem with image generation is you're like, >> "Hey, take this picture of me and turn me 45 degrees to the left." And then suddenly it turns into the average white man. and you're like, "Well, that's not actually me." You lost like you lost a lot of details in your distribution. >> And so having the 3D models, uh, it can like do the wireframes or do different like ink and wash styles, but all from the base shared geometry. >> And so this weekend, I've just had it ripping on. That's why I moved it to the Mac Studio. I don't know if you saw my tweet about transferring of consciousness. I transferred it to the Mac Studio cuz that is an M2 Ultra with 192 gigs of RAM and like turns out Blender works a little bit better over there than on my MacBook Pro. >> So, I didn't mean to yada yada the best part, but that is my art project uh where I where I am sending generous amounts of tokens for exactly zero ROI. But you know what? To the listener, turns out you can just have fun. Like that is possible. I this is new to me. >> This is new to me. >> I don't know. I just had fun, you know. >> I think you gotta print a illustrated book out of this, you know, maybe some something there. Yeah. >> Publisher, >> be the be be the big name in AI writing. >> That'd be cool. >> Yes. That and and I wouldn't hide I wouldn't hide it at all. It would be like this is a collaborative experiment with swarms of AI agents and can it actually with of course much of my guidance. Can it actually produce something that people want to discover? That's kind of the whole that's the whole conceit of the project is like the final artifact will be something that you will have to like discover the story as you go. >> I love this. We're going to turn this into money. First of all, it's not going to be a no money thing, but also >> of course it's going to turn into money. Get real. >> Go there with the money. But I love this idea and I love fantasy stuff. So now we're cooking in my area. So this is awesome. This is great. Let's go. >> You can be you can be an alpha reader for me. >> Yes. Let me in there. I need the audiobook version. Tell them to make an audiobook version. >> I have actually So all the canonical data um is basically facts and like lore and like uh historical influences for each particular society. like there's like, >> you know, the Levites, the priestly cast influence, like the the house that is, you know, revered, all that kind of stuff. >> And so I've also uh I got a little bit bored just like layering facts upon facts. And I was like, I kind of want to enter this world and I want to feel like is this world >> working like at a human level? Like obviously we're layering up all these societal things, but at a human level, is it interesting, >> right? And so I had it uh I had it produce a skill whereby it creates a non-cononical story like it picks one of the it picks one of the areas it enters into that area and it creates like a you know life and blood story >> and then uh it sends it off to 11 Labs for narration and so it it turns it it turns it into an audio story. And I'll be honest, I've been listening to some of these audio stories as I fall asleep at night. >> Oh my god. >> And it's very it's very compelling. And so I feel like artistically it's it's working. Um >> and so it's kind of like a way to play test the world is say like, hey, we've got all these facts. Go in here, make up a story. You can add things that are not canonical, but you can't like reinvent canon. And so go in here and write a story for me at like the human scale, not the society scale. And then it turns it into audio and I listen to it. It's so freaking fun. >> I mean, people love the go to bed things like this. Like an infinite amount of stories with a world you enjoy. >> I know. >> That's really We got to turn this into an app. Get them on it. Forget we're turning this into money now, Erin, right now on the show. So, it is uh it is a freaking blast, I must say. And so, now like I've had it I've had Blender render out uh like many of the structures where these people live, and now I'm going to have it start rendering out basically like character studies of like, all right, so there's this like kind of secretive priestly cast. Let's design their vestments. like why don't you pull from, you know, why don't you pull from the Middle Ages and from the stories in the Bible about the priestly garments and pull from this part of history as well. And so it's really fun. I'm actually learning a lot about the different societies like lost societies throughout history cuz I'm like where can we look where can we look for inspiration for XYZ? It's like well you know the Mesopotamian people blah blah blah and I'm like never heard of those people in my life. Tell me more. So >> I'm telling you guys, >> the golden age. It's golden age. >> Deep in the AI. Deep in the AI. That's crazy. You know, this is how uh this is how Tolken did it. This is he started with like the foundation of the universe. AI, but he started like with the language. He made the language >> because he was like a linguist and then like >> you know the stories came after he had built the whole background and all that stuff. CS CS Lewis and Tolken were friends and Lewis wrote like, you know, a thousand books and Tolken got lost in his one universe. And it's hard to say who who was the better, you know, author or producer because of that. But I remember hearing like >> CS Lewis or somebody uh talking about how Tolken just got stuck on the first thing cuz he invented the universe. Meanwhile, Louis is writing Narnia and Caspian and then all the other religious stuff. But that is that is honestly very fun to be like, hm, how can I invent a whole society of functioning like people, trade agreements, religious rights, all of that kind of stuff. It's it's wild. >> Be careful with the trade agreements. I don't know if you're aware of this since Star Wars trade agreements are controversial. So >> trade agreements play play a big part of this there. You know what is what is society without rules? And what do I love rules? So >> we're getting lots of them. >> The people are getting all the dirt on Labor Day. I love this. >> I know. You never log off. That's why That's why you got to pay attention because you never know what'll happen on this show. >> All right. Why don't you pay some bills? >> I already did. It's your turn. >> There's more bills. Oh It's my turn. We have an app for this, man. >> We talked about this before the show. I put my icon right next to them and then I'm still putting on you to do right. All right. This week, brought to you by Bento, the email marketing and CRM platform built for the AI era. Send your product marketing and transactional email with Bento. And check out their new transactional email pricing with rates as low as 9 cents per thousand emails. Check them out. Check them out at bento.com. and Typesense, the open-source lightning fast search engine for your site or your app, natively integrated with Laravel Scout. Check them out at typesense.org. Uh, thank you to the sponsors this week. Okay. Jeez, I can't believe I did that. Crazy. >> I know. It's like it's your first day. This is episode like 150 or something. >> It's crazy. I just added this feature to let us assign these in outro right as we go. And >> and you literally said as we were doing that, "Wow, this is really nice to have." And then and then you throw me under the bus. >> Maybe our pictures are too similar, you know? Never mind. Just merged them. >> Um, all right. So, we didn't even get to the main thing you worked on. I feel like we should talk about that. We solo on Linux. Like, this is the other thing you've been crunching tokens on and probably other solo stuff. >> Solo on Linux. >> But that's a big one. I'm excited about this one. And I feel like this is a very good direction for Solo. Get in with the Linux people. >> Y >> they they're they want to be in the desktop controlled world. >> They need tools to do it. All the Linux stuff is crappy in my opinion. Uh so I like all this where you're going. All right. Take it away. >> Yeah. So I tweeted recently about how I'm doing solo work these days given that you know my my W2 is beyond full-time. like it is it's a lot. Um, >> and so the way that I'm doing solo work, if you if you missed the tweet, and producer Dave can put it down there somewhere, is that I basically spend, you know, an hour or two at night coming up with uh very indepth like plans or specs. Maybe that's an hour on one, maybe that's an hour on like three or four. And then I hand it off to agents running inside of Solo so that it can do all the spawning and to-dos and all that stuff. And I then I check on it in the morning, see how it did, go off to work, and then we repeat repeat. Weekends, however, weekends are all all for me. So, we're a little we're a little more hands-on on the weekends. Um, but one of those things I've had uh obviously in the back of my mind and has now come to the four of my mind is Linux support. And I don't know if you know the guy uh called DHH who's the creator of Rails. >> He is his new sticktick is uh Omachi which should be pronounced Omari. So, Omari is this Linux distribution. Um, and I bought a while back, like six months ago, I bought, you know, an IBM ThinkPad for like, you know, 600 bucks or whatever. And I was like, "All right, I'm going to figure out how to put Linux on it." 6 months later, no freaking clue. Haven't touched it. And then Omari comes out and I'm like, "Ah, this is supposedly very easy." Turns out it is very easy to get installed. So, I got I got it installed on this ThinkPad, which uh takes me back to my accounting days, but fortunately, I've plugged it in and put it in the closet, so I don't actually have to see it. >> Um, and this comes back to another tweet I had, which is tail scale. Tail scale is amazing. >> So, now I now I have all these computers hooked up. So, from from my MacBook, I can say, "Hey, go in go into the Omari machine. uh get it all configured, get uh the repo pulled down, get all the dependencies installed, start working on it. >> Um and so because it because it's on this tail scale network, I can also VNC into it. So I can actually go and look at the screen. And so right right here from this desk with my MacBook no no no no matter actually even if I'm at home I can pull up the Linux machine the desktop look at it play with it and there were maybe there were maybe like 10 to 15 little things that when I started playing with Solo on the Linux machine was like hm this is not right. Um, and so I've just been working through those and I just overnight ran like the endto-end performance test on Linux and it found a few things that were like hyperland, whatever that is, uh, Wayland, whatever that is specific. Yeah, who freaking knows? Um, >> and so it's working through those last like platform specific problems. Uh, but we're ready to go on Linux or we're we're at least we're at least ready to go on Omachi. I don't know. >> And this is what scares me is all the other Linux distributions. Um, >> it's basic Linux is like the JavaScript ecosystem where it's like what if instead of doing one thing, you could do anything and so everything is different and you're like oh shoot. Um, >> but it works. Like it totally works on Linux. I've got it. It's actually on Linux right now inside of Solo. It is fixing all the Linux things. So, it works well enough that like the orchestration loop is working fine over there. >> Uh, and the cool thing is I can just swipe on my mouse, you know, right, right, right, and then I'm on a Linux machine on this computer. >> Um, so that part has been very cool. I got to figure out the final packaging and releasing. Um, it is already built. I just haven't like exercised it yet. And then after that, people that are on Omachi can start trying it. Uh, so I'm pretty excited. Yeah. >> Well, it seems to be a little like people are into the Omachi right now, right? They're talking about it. So, if you can inject yourself into that conversation, I feel like that will probably be useful uh and get some >> some sales and some interest brewing over there. I'm sure yes, you will find out I'm using this desktop on this OS and it's not working and like whatever. I'm sure there going to be a million of those which will be quite exciting for the agents to crank away on. But um yeah, I mean I I just like this as like a new path of like as this is growing. I I still not a believer in this is the year of desktop Linux by the way. Like I don't believe that's true, but there could be enough energy over there to uh get you some sales. Let's let's go with that. Yeah. And it >> it it at least >> regardless of if Linux people will pay, which is the question we've had on here before, it completes the story. Like it makes me feel good. Like I'm I'm a I'm a completionist. I like to collect things and to have all three platforms feels good. Um especially in in the grand vision where solo can be headless and you can have you can connect to from not even through tail scale or VNC or anything. I could connect on this machine to a solo instance running on another machine >> then Linux makes way more sense because you might have you know eight cheap Ubuntu boxes that run builds or something. Yeah. just cranking on tokens. So, >> it's fun. And I will say it's this is the most fun I've had like this period, one month of of being employed and >> Oh, wow. >> having, you know, >> having a million kids and having >> having a few projects that I still get to like exercise unbridled creativity on. >> I have not had this much fun >> in in ages. I went to I we went, you know, we went to church yesterday and uh I was sitting next to Jennifer and then my sister-in-law and my brother-in-law and kind of all in the same row and my sister-in-law said, "I just have not seen you this light in a long time." And I was like, "Oh, thank you. I have lost a little bit." And she was like, "No, like you I haven't seen you this emotionally light in a long time." Like, yeah, you know, feels good, man. So, >> you're not sneaking out of church to work in the car? >> Not anymore. I've uh to my great shame have done that I think three times uh in in the crunch days is drop off Jennifer and the kids and then stay in the car. Not great. >> Grind the tokens. Yeah. >> The little old ladies are like, "Are you and your husband okay? I noticed he's in the car. He's fine. He's just an entrepreneur." >> Oh man. All right. Well, that's the solo update. I like the Linux move. Things are cooking there. Um, all right. What else we got going on? My big thing, I don't know how much for this week, but I did take 4,000 photos of horses for the last 3 days. >> Thousand photos. >> Yeah, a lot of photos. 4,000 photos. I was clicking away. Click, click, click, click, click. Uh, >> what is your media strategy >> to be able to handle 4,000 photos in one weekend? Oh, it's >> And there are more weekends than one in life, so you must have a lot of photos. >> Well, this is the big weekend. Uh this was my daughter's first big This is like a big fancy high competition level event. So this was her first time doing that. She got the new horse. They're still working things out. So it was quite it was quite a weekend. We had a little life lesson area where like the first day went very poorly. Horse refused jumps and she almost got thrown off and whatever. So, we had the adversity and then her and the trainer worked on some stuff at the end of that day just on the side. They got sort of straight. Uh, and then the next two days went very nicely. She finished third and fourth in a variety of events with like >> sounds like sounds like the hero's journey to me. >> It's exactly the hero's journey. So, that was very nice. It all worked out perfectly in that regard. Like you'd hate to have the adversity at the end. So, it was like adversity at the beginning. >> We worked through it. Uh, that was great. My adversity was like walking several miles a day carrying a a bunch of camera gear and uh taking tons of photos of cooking. I'm all sunburned here. >> So yeah. So I like taking the photos, you know, got all the camera gear. This is like one of the things I do with the camera gear. It's like I got the fancy camera gear in part so I can take these specific photos. So very cool. Like in the barn area. Got some nice photos with her and her buddies and her cousin. shared you shared some privately with us and they turned out amazing. The ones you posted are also good, but you know the the close-ups with the kiddos's face and the horse. It's just they turned out great. >> Yeah. So, it was fun. And >> yeah, so every night it's like I knew I'd have a bunch of photos. So, it's like every night like the job didn't stop. I got to come home, I got to process that day's photos because I don't want to be behind on the photos. It's it's terrible when you're behind in photos. So, go in. One day had 2,000 and each of the other days had 1,000. So processed the photos. Um but they 2000 my niece was who also does this uh won some big award thing blah blah. So I took a bunch of photos of that award process and blah blah. So that was made that day longer and even more photos. So uh yeah. So what I do is I download the photos. So all your photos like 80 to 90 megabytes. So it's a big heavy photo. Download them all onto the computer. crank through them and computer >> on to the MacBook Pro. So, this I need the horsepower. I can't use my little I bought a little MacBook Air that I use now since I'm on the or I don't Yeah, I don't need the local development anymore. No local development on that machine, but >> it can't I tried it. I was like, maybe I can get away with this. It can't. It doesn't have space on the MacBook Pro. >> That one has a 4 terbte. So, that's got a big four terab. >> Yeah. Mhm. >> Machine, >> right? No, now it it was 5,000 uh 4 months ago, but now it's an $89,000 machine. So, >> that's why I got it. I didn't really want it, but I was like, I know all these RAM prices. They're going to raise they're going to jack everything up. Let me just get the big machine now. So, I got the big machine. Now, if I knew I was going to go to this little machine, I probably would have got like a Mac Studio instead, but whatever. So, the MacBook Pro is fast enough with all the disc and bunch of CPUs and 64 gigs of RAM. It's sufficient for my needs. So, when they were cranking photos and uh some people save every photo they ever take. I'm not a believer in that philosophy. I don't want all these crappy photos. I don't care about them. I'm never going to go look at them. It's just more stuff to deal with. So, I go through and do a first pass. And every day here, I basically knocked it down to like a hundred photos. >> So, I took one pass. >> Yeah. So, I take the first pass, get down to like a hundred >> and Lightroom. Where are you doing this tree? >> Lightroom. All in Lightroom. the new Lightroom, uh, not the classic Lightroom and because they offer both still because people freaked out when they updated Lightroom. So now they have both and so which is kind of annoying because there's some stuff in new Lightroom that does New Lightroom doesn't have everything classic Lightroom has and sometimes that's annoying because I wish it had some of those things in classic Lightroom but it doesn't. But whatever. I like the new Lightroom better overall. So I use that turn through that goes pretty well. It's pretty fast. The thing with the horse photos is you're taking a lot of photos of like, okay, the jump is happening. >> You just got your hand down, your fingers on the trigger, and you're taking 20 shots as the jump occurs so that you can just pick out the one where like the horse is right and the kid is right and all the stuff's right. And sometimes none of it's right even with that, but whatever. So that's what you're doing. So you have a lot of like, okay, there's only one good photo in here. I can instantly get rid of the other 30. Okay, fine. So do that. Um, and then I edit them, which takes much longer, and like go through and crop and colors and whatever, tweak all the stuff. And then, uh, that's kind of it. So, I did that every night, and yeah, so I got a bunch of photos, threw some up on Instagram for the people who follow me there, but that's uh, on my personal Instagram, so like most of you won't be able to follow me there. Sorry, I I deny requests on that one. >> Brutal if I don't really know you, but I do have a public Instagram. I do occasionally post stuff, but more like not the horse stuff. Yeah. Not NFTs either. More like what you would call like fine art type photos. >> Oh wow. >> Um so I don't So that one's public. You can follow me at Ian Landman on there for the public stuff. Um >> and >> what's the private account? Horse daddy or what's the handle there? >> Underscore Ian Lansman is the private. >> Yeah, but don't be messaging me on there because don't follow him. >> I'm rejecting you. It's only for people I I know well in real life. That's it. Um, but Ian Lansman does have all, you know, I think I like four or five things up there, but there's some public stuff. >> Just take take Horse Daddy under consideration just for your next for your next handle. I don't know where it'll be useful, but it's it's great. It's good content. >> Yeah, I know. So, it's good because I get to practice my skills. There's still I got to get better at a few things for sure. Um, but yeah, nobody nobody yelled at me. We talked about >> permit. Your Ron Swanson permit. I can do what I want. >> Yeah. I don't know what this This is a movie reference apparently. I don't I don't know about this permit. >> No, it's a sitcom reference. Parks and Wreck. >> I never watched that. It's queued up for me. I do need to watch that. >> Oh, wow. This is a This is a real turning of the tides here. I've seen I've seen I've seen that show probably like six or seven times straight through. >> I I do. It's got Chvy Chase, right? Is that the other one? >> That's the other one. >> That's the other one. Okay. >> Yeah. Um, yeah. I don't know. I should watch some of these sitcoms that I just go like last night I was watching a sitcom, but I just watched Seinfeld. Like I'm just watching Seinfeld. >> We watched Friend. We Friends is our nighttime show. >> I should watch Friends. I could watch Friends. >> I didn't watch it growing up. Uh, and Jennifer did. And it's very funny. >> It It goes a little too One thing I love about Seinfeld is that it never goes like sentimental or lessons learned. And Friends Friends goes a little lessons learned. Um, and a little like >> intergroup drama. You know, when Jerry and Elaine get back together for their like we can do this and we can do that, this is good, that is good. It's just funny. But when you know Ross and Rachel get together, everybody gets hurt AND IT'S LIKE AH I don't want to watch that. But it's still extremely funny. >> Yeah, I do kind of remember. I did watch I didn't like religiously watch it but I would pop in on it on occasion because it was on I think right before Seinfeld I think or right after or after. >> Um yeah so occasionally I would watch it but uh mostly just for Jennifer Aniston if we're being honest but occasionally I would pay some more attention to the actual plot and it's like it's fine but it's fine it's not that exciting. Um >> you know I think the the cl the clarity of Seinfeld is one of its genius points. It's like we are going to be very clear that we never learn. You know, he had this whole philosophy he talked about, no learning, no loving or whatever, no lessons. Like he's got like three things. It's like that just removes so much crust from the show of like, well, we got to clo wrap this up with a bow and everybody's got to be happy, everybody's got to be sad. It's like, nope, we can just end it on this was funny. That's it. There's no repercussions to anything. >> Exactly. Moving on. >> Yeah. >> The characters never grow emotionally. That's like that's the point that that's that is why the final episode was a call back to the first episode. >> The button is too high. Makes or breaks the shirt. Um >> so back to cameras. This this super lens that you got. >> I saw uh uh I saw Eric Barnes first of all taking some amazing photos. Although I don't know if he's your protege. There you go. >> My protege. Yes. But one of the things I saw you two talking about was like using this absurd lens for a webcam. >> So when I get I've signed the lease for the new space, by the way, so I'm getting a new space. Um and I'll have more space. Uh could I use an absurd lens for a webcam >> and like would it look awesome? >> It would. I mean this one that we were talking about there, I think it was actually with Wan too was in on that one. But um that one I don't know if would be actually perfect for a webcam cuz it goes 50 to 150. So 50 is still pretty, you know, fairly telephoto for a webcam, but I mean you could push it back um and do it, >> but it's so you could um there's some there's some other ones all I would get a Sony. You need to upgrade to a Sony. I would definitely say you have >> upgrade. What's wrong with Canon? Canon is not a downgrade from a Sony. >> That's bad, man. The focus is so much worse. The Sony focus. Listen, there's eight million hours of me on this podcast. I've never been out of focus even for one second the entire time because of focus. >> I'm out of focus for like five seconds an episode. That's not THAT BAD. >> WELL, YOU'RE TRYING TO BE ABSURD HERE. You're you're talking about buying a $4,000 lens just cuz. So, I'm saying maybe buy >> I'm not talking about that. I didn't know $4,000 lens and a $2,000 camera and have perfect focus for the same amount of money. That's what I'm saying. >> So, the Sony cameras do better autofocus like their algorithms better. Yeah, I would say it's much better. >> Um, but you're a Canon guy, so it's also tricky. You got the batteries, you got stuff, got >> Yeah, I got the batteries, the cables, all the lenses. So, super specifics aside, if I get what what millimeter range is good for like a this is absurd kind of webcam, provided I can push it back as far as I need to go. >> What do I want? >> You could go if you got a lot of space, you could go you go like the other absurd lens I have, which is a 300 millimeter 2.8 eight like >> and what does that buy me being further away just better background blur and >> yeah like if you wanted to go crazy with the background blur but there's other I mean for webc this was just they were being somewhat farcical this was not I would not recommend this webcam >> but my thing is doing the farical thing sincerely so let's let's talk about that >> I would probably go a different route if you wanted to do that uh and what I would do is get like on the Canon I believe they have like a 351.2 two. So if you did that, you'd also get the crazy background blur. >> Um, and it would give you a more normal webcammy type uh millimeter range. And so that's probably what I would do. Or you could go 501.2. You'd have to push it back a little bit. Um, and that would make you a very shallow depth of field. Now, the only problem with that with these Cannons that in the video especially don't focus as well is at 1.2 there's literally just going to be your eyes. Yeah. So, you're going to be in and out of focus like a lot if it doesn't really keep up. >> Now, maybe Canon's newer bodies do a better job. So, you could like get a newer body, the 501.2 or 35 1.2, >> but I don't know as much about the Canon's like the newest stuff especially. So, maybe the newest bodies are as good as uh as good as >> is it the more like if I did like a 18 mm that's where I get like pancake face and then if I get like a 200 mm I look thinner. >> You'd be very Yeah, exactly. >> All right. 200 it is. >> I'm not doing pancake face. I got enough of that going on. >> But even like a 50 mm will give you a fair amount of that without having to push the camera like so far back to be ridiculous. Yeah. So something like that. It depends too on like your setup. Like I have the screen in front of me so I have to have the camera like up over >> um to like watch your angles there and stuff. Uh but if you have like more to the side it's different. But um yeah. I don't know. But the camera gear is fun. The camera gear is like half of why I like doing it. It's like you get the gear, you sell the gear, you buy every hobby. >> Yeah, exactly. It's like you collect the stuff as we were saying. >> Do you want to ride bikes or have bikes? Well, 50/50, you know. >> Yeah, exactly. So, the other thing Eric was talking about, if you did want to buy something, was this the uh they might have a Canon mount for it. Um there's this new company called Viltrox that makes really good high quality but cheap lenses. And so, he got a 85mm 2.0, know which Eric was showing off his new photos done with that lens which is much better for he was doing like high school football so it's like not not good light and things like that so you want to have a wide aperture so you can collect a lot of light so he had been using like an F4 lens which zoomed which is handy but is not as good in low light so that's why he got this it was like it's like 300 bucks for this lens but the Vitrox lenses if they have a Canon mount you could get something there too they that's a little bit cheaper um and they're quite good actually. So, >> a lot of options out there. A lot of options. But, uh, for the horse stuff, it does push you. The good gear pays off for the horse stuff because talking head video, you're not actually getting much out of the good gear. But, for a moving animal that's jumping things in various types of light like the good gears where you're you're getting your more of your money's worth if you can, you know, I'm justifying it. Obviously, there are cheaper ways to do some of these things, >> but uh you are getting the most out of the lens and what you're paying for and those kind of things. >> So, you're telling me 4 grand for a uh head shot in controlled lighting? Maybe a little bit overdone if it's going over the internet 1080p anyway. >> Exactly. >> I'm using currently a disregarded. >> There you go. I have the Sony I use the Yeah, I think this is the 40 mm the 40mm 2.5 panc like little almost a pancake Sony lens and uh it's great. >> I think this is I think this is a 24 >> that's a standard >> enough of camera corner. Let's talk about Valoran security. >> AI doesn't just help you ship code faster. It helps attackers build tools and find vulnerabilities faster too. Especially with Astra. That's why you need a friendly hacker like Steven Reese Carter on your side catching the holes before they do. You can check him out at valorsecsecurity.com/mt. Link will be down below. And also our friend Jesse. They're all Jesse at dropin blog. A headless blog CMS with a firstass Laravel SDK. So you get your domain, your design, your stack now with integrations for Astro, React, Tanstack, and Big Commerce Catalyst. You can find them at dropinblog.com, spelled exactly how it sounds. You know, we love that. So thanks to Valoran and Dropin Blog for sponsoring this show. >> Love a.com. >> Love.com. >> Can I Can I give you some outro feedback right now? >> Oh, yeah. Give it to me. I want optimistic updates when I click uh >> Yes, I was thinking >> that the freaking segment is done. >> It's like it's a full second lag from mouse down or mouse up to complete. >> It's I was like thinking about should I go type this into the AI right now to do it? >> You should get Orbin on that cuz I need I need that. >> It's very annoying. I totally agree that has to happen. I I think I I did have that at one point, but there was some other issue, but I've got those other issues worked out. So now I think if we do it, it'll all work correctly. Um so yeah, I I that's going to happen before our next one. That's going to happen. Um >> let's do it's not it's not on the list. Let's do a little detour. I think >> I think you're going to start giving this you're going to start inviting friendlies into Outro on I think it's July 9th. And so that's coming up. Uh >> it's coming up. I got a year though. >> How Oh, we're >> You never specified the year. >> Bamboozled again. Uh, how are we doing on outro? You've been you've been orbing and it seems like it seems like with uh the new orb psychosis, you're making progress. Where where do we stand? >> So, I'm mostly still focused on a bunch of HubSpot stuff uh between sock 2, new UI for the admir, some other new modules. So, a lot going on. That's when most of the orbin is going. But I have done a little bit of outroing mostly on things which I would never release until these things were done. So that's good. We got the AI working faster. Did you add anything with the AI today? >> I didn't. I just clicked buttons like an animal. >> Oh, you just typed. Okay. All right. You're usually my AI guy. So I see you there waiting for it to finish and that was annoying me. And so uh the AI is now much faster. Um >> new new model, new architecture. What? new uh switched over to the OpenAI models, but uh more so I think partially because Luna is better. That's probably the main reason I would say. Um so I can use Luna for more stuff that uh I was using like Sonnet for and it's faster and better and cheaper. So that's good. And then um uh did rearchitect that had this was still all soul not Astra but soul went in there with a little fable guidance I believe and uh they they fix up the speed quite a bit of uh some extra calls and various stuff that was in there that didn't need to be there. So that's good. And some other just loading stuff was uh dramatically improved speedwise. So we're getting there. Um I did have this when you check a thing is done. it showed a little check and it had an intentional delay and that was annoying me. So, I removed that, but there's still a delay uh because it's making the round trip. So, yes, we need to have the optimistic uh removal. So, that'll happen soon. There was also Oh, that's the other thing it did was there was like a >> there was like some weird stuff if we both like were checking things nearly simultaneously. So now it does a better job of avoiding that scenario which I still think ultimately there'll probably be occasionally times where that happens but I just didn't want to be like it felt like it happened a fair amount and I want it to be all faster so that >> you've accidentally built a distributed system and now it >> nightmare scenario. So uh yeah so those are things like the AI could just go off and do pretty much without a lot of input. I feel like >> when you're not doing UI stuff at this point like the AI can kind of go off and do things uh mostly. So that's should have done that stuff, >> but I don't know. You know, there's a chance this just ends up being our little playground. This is just our little place. WE BUILD STUFF. >> WHY? >> NO, I KNOW Jamie's on me to release it, too. But it's like there's no good time to release. I don't want to deal with releasing. >> She stopped Jamie stopped listening a while ago cuz we got technical fast. >> So send her this segment. Jamie, help me. Help me help Ian. This is crazy. >> I know. I thought about maybe possibly I hire like a part-time type person to sort of like >> WHY IT'S GOOD NOW. >> NO, not even about finishing it. Like just like it's like the thing I don't want to deal with is like there's going to then be bugs and people are messaging and they need help or they want a refund or they something's broken. Like that whole stuff just takes so much bandwidth. >> Don't you have a whole company that does a lot of that stuff? >> We're small and they're working. We have we have these huge HubSpot things we're working on. So, I can't actually afford them to go off on any outro stuff right now, which is very funny because like when I started this last summer, we had loads of time and I was like, "Well, all right." Like, you know, between me starting it and then like once it gets to production, like they can help or whatever. But now, nobody has any time at all. So, >> why AI is supposed to eliminate jobs and work? How is that not happening? >> That's the theory, but so far it's just making lots of work. So, >> shoot. >> Yes. Um, we'll get it out there eventually. >> You're killing me. You're killing me with >> a ton of traveling this summer. So, we're going to talk about a final bit of traveling we're doing here. But after this next bit of traveling, that's not even true. I have another thing in October. >> Oh my goodness, it's Thanksgiving. Oh, I got Christmas. Oh, my gallbladder broke again. It's always It's always going to happen, you know. >> I do have to release this at some point. This Ian, this is your one wild life. And do you want do you want to not release Outro in your one wild life? That's crazy talk. >> That would be crazy. That would be crazy. >> That's crazy. >> You know, there's a couple other things. The mailbag's not done yet. I got to get the mailbag done. Got some stuff to finish, you know. Uh >> what what would Jason Cohen say? What would he say to you? >> I don't understand. I guess I don't know what he'd say. We're going to have him on. I >> don't either. We're going to have him on, so we'll find out. He might say, "Stop being a freaking idiot. You have a product that like does something." >> Yeah. He might say, "Kill outro immediately. What the hell are you doing?" In which case, we won't we won't air that episode cuz I don't like that. I don't like that feedback. >> I think that's possible that we get that advice from him. So, >> here. Let me try again. What would Jesse Hanley say? >> Oh, he'd say ship it. >> He'd say ship it. >> Yeah, he would have shipped it. >> He's a oneman operation and he's starting new oneman operations as we speak. He's like 15 years younger than me. He's got that young guy energy. I got the old guy. >> Old man river can't keep up. >> More email. I don't want more email, you know. >> But uh >> I don't know. We still have to figure out what to do with pricing. I've got various things in the pricing I'm not sure about. Like John Nolan was like only have the $99 plan. Producer Dave agreed with that. I'm a little nervous. I kind of thought I wanted the free plan and the $99. Then I checked out and put it 29 in, but maybe I shouldn't do any of that. I don't know. start at 99 and then when nobody signs up, you'll be happy because you'll be like, "Ah, I don't have any customers. It's exactly what I wanted." >> I mean, this is literally the upside of that is like I was like, "Well, if I just have 99, then like there can only be so many customers in the beginning and then like whatever. If they email me for something, it's fine because it's just like one guy and he's actually paying me a reasonable amount of money, so that's all right." Um, could be. I feel like you need the free version, though. That's how the internet works now. There's a free version. killing, you know, >> but then you got to take out the AI stuff. You know, there's a lot of I already built all that stuff, though. It already has all the flags to like hide stuff or whatever. Um, >> all right. Detour over. I'm I'm unsatisfied. >> I'm I'm unsatisfied, too. >> I'm very unsatisfied. Do you want to talk about You want to go into Astra or do you want to talk about me training a model? >> Let's talk about you training a model because I want to do that, too, at some point. So, we've discussed this uh in episodes prior where one of my one of my soloerm.com it's very good problems is >> I uh am basically just a terminal >> and so I try my hardest not to require a lot of setup right now. The only setup that you have to do is like >> if we don't figure out which agents you're using, which we figure out like the top eight, you know, common agents, >> uh, you have to add your agents and then after that you have to install the MCP into those agents. It's just one, it's a one-click thing now. >> And so part of my part of the way that I keep this lowmaintenance is I don't, you know, handcode heristics for every 2e or CLI. I'm not like this is cloud code and therefore I can look at that on the screen and I know that cl because >> I can't keep up with that. Um and I think it's bad it's bad architecture and I can't keep up with it. Yeah, >> but one of the recurring problems is >> does this agent uh require my attention? >> Um, which could be permission, could be I finished this task, am I allowed to do this thing? And it's not a real permission prompt. It's just a question printed to the screen. And so my solution uh my pending solution to that is train a very small micro model that can input or ingest any terminal grid. So on the back end I have you know a representation of what is shown on the screen because I am the terminal >> right >> and so can I feed that grid of cells to a very small local only >> uh model I don't even think it's an LLM I think it's just a model can I feed it to that model and get a structured score back of confidence of um if it's idle if it needs permissions if it needs my input that sort of thing. >> Mhm. And so, uh, it is currently training right now as we speak on the Mac on the Mac Studio. It is training the first run. >> And so, what I've done is, uh, I set up Solo on the Mac Studio with this repo. >> And I basically said, great, start spawning all the agents you can find. Start like feeding it whatever inputs you want. The goal is uh to capture a bunch of samples of different agents at different points in their life cycle. >> Capture uh the visible grid and like certain underlying diagnostics we have from solo like the last time that we saw visible pty bytes, the last time that the human typed uh any sort of input, that sort of stuff. So feed it like capture all of that. And then what you're going to do the this classifier repo it basically spawns a uh a new codeex a fresh codeex and says give me your score for these you know five labels give me your score and your confidence and that's how we do the labeling. >> So I don't know if this is distillation or not but either way it's awesome. Uh and distillation is my right as a US citizen. live free or die hard. So like I'm I'm using the frontier models to basically label these observations. >> And then now I've got like 500 labeled observations. And so still pretty small, but I'm feeding it through this first training round, which is some Python thing I don't know the first thing about. Who knows? >> And on the other side, something is going to pop out that I can consume from Rust. And it should be under like five milliseconds to get a response back from a grid. And it's all local. And so I can just continually feed these grids into this local little model. And then that will help me surface back to the user. This terminal needs your attention. This one's stuck on permissions. This one's ready to go. That sort of stuff. Um I've also built in just for development mode only. uh it's hidden behind a flag and then compiled out. Um auto capture. So now as I'm working on all these machines running solo, I can turn on this dev flag and every 30 seconds it'll just capture all of that diagnostic. So instead of these artificial like you know LLM generate, not generate LLM driven captures, it captures all of this from my day-to-day usage. Uh, and so we'll quickly get up to many thousands of samples, which will make the model more accurate, I believe. But it's it's currently training. Like right before we got on this call, I said I said, "Stop capturing. Train the freaking thing and let's see if it works." I was like, "Yes, sir. Right away, sir." So, we're cooking. >> So, you're going to train this first one, the first run of training just on like the 500 it generated, and then you'll do new trainings. >> Yeah. I just I just want to see it work, you know? >> Does it work? Yeah. How long is it? Did it estimate how long or you have any idea how long it should take? >> No, it didn't. Um, let me see if it's >> probably shouldn't be too long with only like 500 data points. >> Oh, shoot. It's done. >> There we go. All right. Live on air. Let's do it. Feed it something. >> It says, "Yes, the local training smoke test passed. It took 11 minutes and 36 seconds. The model's 537 kilobytes." >> What? >> It's pretty tight. Does it work though? >> Uh, not much text to like analyze, you know. >> I guess not. Verification reloaded successfully and scored all 14 synthetic checks. >> So, I I guess it worked. Let me >> scored them or accurately scored. They're probably two different. >> Let Let me ask. >> So, now that the model is trained, the 14 synthetic checks, did they all pass? Like, is the thing working? Uh, and how fast is it working? Oh man, we're doing it AI. >> We're doing it live. You got to talk to the AI. You got to talk to the AI because when you type your little fingies, especially being so old, your little fingies get tired, right? That's true. >> And so you can just you can just talk talk and then you get all of that context out of your head. >> Um, >> okay. So it says all 14 checks. That's not very many controls. All 14 checks passed, including valid four outcome scores. So the the scoreer comes back with four like labels or heristics and a percentage of each >> that checked that the scoring worked not prediction accuracy. Well accuracy is kind of important. So I need you to check accuracy as well. >> What the hell did you check then? >> AI is here. Aaron >> you checked it out something. That's not the point of the exercise. >> It says I will benchmark it now so I can give you measured latency. All right. We'll come back to it. That's >> how come to that. Oh boy. The AGI is taking your job. >> Silly model. What are you doing? >> Oh my goodness. All right. So, you train this model. Cool. I mean, this sounds great. I was a little uh pessimistic at first. I think I think it seems reasonableish. >> It seems reasonableish and that's my sweet spot. I don't want to be I don't want to be too reasonable because then come on. Hey, and I don't want to be too unreasonable because then I'll, you know, never be a success. But right in the middle of reasonable and unreasonable, that that's where I live. >> I'll be curious to see if this thing actually runs well across all your OSS and everything, you know, it's like >> security, Mac OS security says this thing shouldn't be running or like what, you know, I don't know, whatever. Random stuff that >> one of these Omachi people that runs their machine on a potato like is it going to be performant over there? I don't know. >> Yes. Stuff like that. Although at 500 kilobytes, if it's actually ends up being accurate in that range, presumably it's not taking too much memory to execute. And this is all text you're feeding in, right? It's not the images. >> It's all No, it's all text. All text. >> So, >> so it's pretty everything involved is pretty small. So, that's definitely working for you here. >> Holy crap. It's fast, too. Um, okay. I benchmarked the exact model using desktop Rust runtime on the studio. The small prompt took 0.1 milliseconds. 90 the 90 uh P95 is2.12 milliseconds. >> Holy cow. >> A build log. So a build log having lots more uh scroll back would be8 milliseconds. P95.85 >> and then super dense output plus a lot of scroll back is uh P95 of 8 milliseconds. So an order of magnitude slower but still >> 8 milliseconds and it says loading the model takes 1.7 milliseconds one time. Um and then prediction timings include feature extraction which accounts for almost all the work. There's synthetic benchmarks yada yada yada. Uh so it works mechanically and is fast. Accuracy is the next thing to establish. Yeah, I know >> dude. Oh >> we'll check back in with Yeah, we'll report back. But it is fast and it is small. Um, this is cool, right? >> That is cool. You're out there. You're doing stuff. Yeah. >> I'm a Neolab now. >> Is that what this thing is called? >> No, that's what Open AI and Anthropic are called. Yeah. >> Really? A Neol? I don't think I >> Yeah, it's like new new one of the new labs. Um, >> okay. >> So, we'll see. But it's it's it's a lot of fun. I'll say that. I'm having a blast. >> I know. When I went on vacation a couple weeks ago there, I wanted to with the kids do that guy from Anthropic who's got the Mac app to train the models. >> We just didn't. But I wanted to >> No, I looked at that. I looked at that as well and it I I couldn't wrap my head around it. So, I just asked. >> Oh, really? >> Yeah. I I downloaded it and I wasn't entirely clear how to use it and my >> I don't know my bar I don't know if it's good or bad but my attention span for new tools and apps right now is pretty low which is ironic considering I'm building an app and tool that I want people to adopt but hey let's not investigate that too closely >> um and so I just and you know I we talked about using that hosted provider uh I forgot what they're called 18th or 19th or something maybe 19th >> and they just never got back to What the hell? >> You build your own. >> So, I just built my own and it's exactly what I need and I don't have to pay anybody for it. So, sorry to 19th, but uh I'm doing it myself. >> I think for what you're trying to do, I don't think that Mac app would have been the right route anyway. You want something that you're going to be able to totally like >> wire up and automate and all that stuff. So, yeah, I think this has all worked for the best. So, >> yeah. >> All right. While we wait for that, Astra, have you used it? >> Astra, I've used all of it. There's no more Astra left. >> If you if you if you ran into this, this model is at capacity. That's my fault. Um >> Oh boy. >> Yeah, it's good, man. It's really good. It loves it loves Python and it loves to to stop and ask like, great, we're done. Can I do this part now? And I'm like, yeah, I told you to do all that. So that that feels like a minor regression. Yeah, >> but in terms of things that's found and like the actual, you know, I can't judge the the quality of the rest code, but in terms of like uh finding subtle bugs, finding performance improvements, um it's very very good. I can't speak exactly to how much usage it uses because like I said, I did you I did burn through banked resets across my four accounts and so there's like there's an actual chance I've used like 10 quotas worth of of Astra so far. >> Um, >> but it's >> it's powerful. I've kept it running on either high or medium. I've only dipped into like extra high or max once or twice and then thought I don't think this is I don't even think this is needed. I saw that Tibo tweeted >> that Astra on low is the is the same or better as soul on I think it was extra high. Was it high? I think it was. >> Um >> so it's it's very good. And you know to to broaden the conversation a bit. I don't know what Anthropic does at this point because I feel like they're losing the they're losing the mandate of heaven. I know that I know that they're embedded in enterprises and I fully I I fully agree with that. I'm not talking about that. I'm talking about the models specifically. It does seem like OpenAI has started to run away with it just a little. Ever since like Opus 4, was it 48 or 47 that people started to be like, myself included, this isn't as good as the old Opus, I think 46, 45 and 46 were awesome. >> And then somewhere in 47,48, they kind of like >> started to get the cloud speak and started to do a lot worse. And I just feel like Open AI's got it right now. >> So I uh disagree, but for different sorts of reasons. So like I think people are misattributing this idea uh to like the models are worse but I don't think it's really that at all. I think it's that like they lack capacity and OpenAI went out first and got tons of capacity and Anthropic didn't. Right. So they've now got they stole Elon's capacity and they got capacity from Google and whatever but I still think they're like pretty light on capacity. >> Yep. And so that's forced them to make Fable, which everybody has for months been like, "Fable's way better than all the OpenAI stuff, then everything else. Fable's better. It was clearly better." But you couldn't use it like you use Soul >> even when Soul I think Fable is better than Soul, but it's like I can use Soul like unlimited. They're giving me basically like unlimited tokens every week or I can like use Fable and be out of tokens in the first day and then have to like, you know, drop down and cherry pick and all stuff. So that's why I think it's like if they could have released Fable as just hey use it. It's it's this is just now the top model but it's like a normal level of token usage in your account. I think that everybody would be like oh they fa Anthropic's been on top for two months now and now they're going back and forth. Maybe after is better than Fable. Fine. But like they're going back and forth. Whereas like it's had the perception that that's not true >> because everybody moved to Soul because Soul gives you unlimited tokens and Anthropic restricting the tokens on their sort of top tier model. And so that like creates that perception amongst the X $200 a month world of that part of it. But um >> are you saying that war warfare is primarily an exercise in logistics and this is no different? Yeah. >> Yeah, I think that's true. And I also think like who who they're each serving is quite interesting, right? Like OpenAI is I think much more trying to serve the ex crowd the and this is where they've kind of won is like the normies who have a free chat GBT account. Yeah. >> And like the people the proumers who have got their $200 a month account and then Anthropics won enterprise of like we got in there and like got all these fiveyear deals with enterprise. So it's like well I only have so much capacity. Am I going to give it to these people who are basically stealing tokens from me or am I going to give it to IBM? Well, I'm just gonna give it to IBM, right? So, like, so I think the capacity is actually a big part of it. I don't think it's quite as clear on like the model capability because I think Fable's been out a long time and Fable's better than those OpenAI models except maybe Astra. Who knows? I don't know. But so anyway, I think that's a little different. But yeah, I mean, I'm on the Open AI now because whatever. I'll take the free tokens while you're giving them those bank resets, baby. Yeah, like the bank resets for because it's for us, right? We are a guy. We are a handful of people. Great. Like the the money is if I spend 60 grand on tokens this year or I could get the same amount for 10, like for me 50 grand difference is a lot of money and I'm going to make that choice. Other companies have different uh metrics there. So yeah, but I mean I've used Astra. I've been going with the medium because a lot of people said the medium is like >> reasonable. I think that's been mostly true. But I definitely think it's really good at like performance stuff. I've seen a lot of people talking about that. That definitely seems true. Um it's been the general logic level has been really good. I did give it one big thing to do that totally screwed up. Um but it was like a big it was a big thing that it went off sideways on which is like whatever an AI thing. I was op I was very optimistic about could it do this and it couldn't do it and fine whatever. So, but uh like I've been going through this big thing now with like a new big module for HubSpot where I've had it doing and I always forget about this and whenever I go back to it I'm like why do I not always do this? I have it asking me questions >> and we're just we are just we're drilling down into all the right spots. We're figuring it out. It's coming up with great like this is where a big improvement I would say is like it is coming up for the most part with good suggestions. Um, and it's an it's like this it's coming up with like things that are edge cases but like smart good ones instead of just like here's a stupid edge case of like if the Mars is aligned with Jupiter, right? Yeah. Like this is like no like this is a feature edge case like feature-wise we're missing this thing. It might be important. What do you think we should do about it? And then we have the back and forth. So, >> and are you doing all that in orbs? So, AMP has uh Astra through your subscription inside the orb. >> Yes, I'm doing it with in an orb. Yep. And uh so yeah, so we're doing it in there and uh yeah, so that I haven't now that this is a process I've been working on for a well since yesterday morning. I was at the horse show between things, you know, we're answering questions, we're doing stuff on the ride home. and got a bunch done cuz Jamie was driving and uh so I haven't actually built the thing that we're talking through yet. So we will see how it does with that. But I feel like the plan is going to be super good and if it can't if it chokes on the full plan then I can like refactor and like let's go through it a little more slowly or whatever. But uh >> yeah, so I've been super happy with the Astra for sure. Um I think it is seems fable like to me. I think it's hard to know. Like it's all pretty subtle, but you could tell the bigger jumps, but it still also seems not AGI. So, I still have no concerns about AGI. It still makes tons of stupid mistakes. It does stupid stuff all the time. Um, but >> job's not finished. Still work to do there. >> Yeah, job's not finished. I did do some security stuff with it. Uh, it found one or two things, but nothing crazy, which is good. It might just not have security stuff. Hopefully, knock on lots of wood. I don't know. Um, but uh seemed to be reasonable there. And uh yeah, so so far so good on the Astra. I like it. I like to use the best model all week long. This is the other thing like just give me the best model all week long. I don't want to drop down. Uh you know, maybe for a sub agent I let it drop down to a low effort or something like that, but I don't really want the lower level models for the most part. So yeah, good. Like it new stuff. Yeah. >> Love new stuff. >> Love new stuff. We are going to Vegas next week, you and me. >> Vegas, baby. >> Vegas, baby. On a little trip. >> Time to sit in the cabana. >> Cabana. Cabana comp. This is an annual traditionish annualish tradition. The crew of us go to uh usually Vegas, but once or twice somewhere else and um hang out. So, it's kind of like the Laravel OGs plus some other people. It's going to be fun, but we're going to be together in person. We're going to be together in person. It's not going to be recorded. So, you know, the content's going to be crazy. Well, we'll report back, you know, but it's a >> it is uh in my mind a it's like a mastermind. So, we get all these, you know, whatever it is, 8, 10, 12 people together and basically just sit around. And this is like this is the platonic ideal for a conference, by the way. fewer fewer than 15 people sitting around. That that is the dream. Um >> so I am I am extremely looking forward to it. Uh not only for the cabana and the antics that I'm sure you will bring to the cabana as a as a cabana godfather. I'm sure that you will >> tell us how to interact with the cabana boys and how how to tip etc. But for the sweet hangs over like two or three days of just how are we going to conquer the world that's the dream that's the best part right there and the whole philosophy of this is like so there are other people who do like masterminds but we this is not structured like that. This is people in a cabana sitting there and you you're welcome to talk about business and sometimes we do but we don't have to talk about business necessarily and uh there's always business stuff comes up but we're not like everybody doesn't get 15 minutes or anything like that. This is just we're hanging cabana cabana boys bringing us beverages. We get mangoes. We hang out with there's a pool. >> It worked out very nicely. It's going to be 102 degrees in Vegas. So we got the pool got the cabana. We're all there. We have some nice dinners. Uh, >> we just go we just go where the spirit leads. Yeah, >> we be that's exactly right. >> That is exactly take the wheel. >> I forget now. So the one you've been to one of these? >> I've been to one. >> Two. You've been to one and that was the one I missed cuz uh gallbladder attacked me. So this will be our first time in the cabana together which will be very fun. And yeah, so we'll report back. I'm traveling back on Monday, so we'll have to do the show on Tuesday probably. So show might be late next week. Um, >> but yeah, I don't know. Love Vegas. I love going to Vegas. The flights are good. The hotels are good. Cabana's good. Um, and you all know how I feel about Cabana. So, we got three days of Cabanas. Big Cabana, the max size cabana. >> Um, just nothing better than a cabana. I'm doing a conference in a cabana. >> Top tier right offable trip. >> Oh, in a cabana% offable. Yes. >> 100%. So, we might get a little poker in. probably get some other gambling and antics in, but it's going to be fun. See the crew hang out >> and we'll report back a little bit. Not too much, honestly, but we'll say how it was. >> Yeah. You know, some things some things are secrets. You can't you can't reveal all of your plans to the enemy. I don't know who the enemy is, but you can't reveal your plans to them. >> No, I can't do that. Um, I did have an idea for it. I brought up uh should we open up one slot >> in honor of Aaron D. Francis who doesn't eat anything? Is this that story again? >> No, the new story. You could buy your way into the cabana comp. There's one. >> Absolutely not. >> $100,000. >> Sure. >> Yeah. You changed my mind almost immediately. Absolutely. Yeah. >> 100 grand. You're in You're in the cabana. >> The stipulation is this is Ian and Aaron selling the ticket. This is not This is not the whole Cabana crew. >> Oh, I like what you're talking about. >> So, we go 5050 on that. Yeah. >> There you go. There we go. >> There you go. >> Um All right. So anyway, that's gonna be fun. We're traveling. You're taking off, I guess. I don't know. Is this count as work for you? Is it not work? It's a fuzzy area. Is this work for you? >> No, I'll take off. I'll PTO Thursday, Friday. Yeah. >> Okay. All right. Well, then we will see you all next Wednesday, probably. Probably a little late. And uh yeah, thanks everybody. Check us out mostlytechnical.com. Get those sponsorships in. Uh check the old episodes. Email us at [email protected]. And uh thanks to the sponsors. Thanks for listening. We will see you next week. Talk to you later., 148: AI Psychosis Day, YouTube, 01m20gy037mh16p1d51kmy7nfg.