# Software Factories: Escape from Trash Mountain — Steve Yegge — session 2026-11-12T01:10:00.000Z → 2026-11-12T01:30:00.000Z _103 transcript lines · 41 slides_ ## Transcript They're like text games and this one has graphics on top of it, which turns out to be like really ideal for AI agents because it's text. They can literally play the game via telnet and they asked me to. So we'll talk about that too. Here's an older architecture diagram of what my, sort of, of what my system Looks like. So when I built this wheelhouse software factory for Wyvern without a care in the world for making it, you know, usable by you. Right? No open sourcing, just bespoke. I discovered the alien facehugger pattern. Alright? Which is where your factory turns into an alien facehugger and like gloms onto your product and just becomes Grown into it. Okay? Like, look, we could turn off the AI and still run the game today. Me and the admins and the players and the community and we could we could... But we would miss the AI at this point. The players are talking to my factory. The factory's talking to my admins. They're asking for features. I'm not even in the loop. It's like a wish factory. They're building the game. Now, they're all old timers and there's a set of them that I trust to make, you know, wise decisions about the game. But they're they're all working with my factory now, and they'd be mad. The funny thing is when it first started, they were like, time out. What is this AI stuff? For starters. And they were like, no AI art because they could see see it with Slop. And I'm like, okay, we got human artists. And they said, but we do kinda like all the new features, but we're really scared of the pace. Can you give us a road map? My players came to me asking for a road map. They were like, you gotta slow down. It's too much. And I was like, okay. I took a look and my software factory, this thing, had been cranking out, like, 50 to a 100 features a day. I had a hundred years of work in my backlog when I started this and Claude got through about sixty years of it in about four months. Right? It's absolute insanity. And they can't look at a 100 features a day. They can't even, like, read about a 100 features a day, let alone play test them. I was... My factory was completely overwhelming my customer base, my users. It was also melting my machines and it was like Like, annoying my accountant. I mean, she likes working with Bat, who's my head of finance, but she sent me a $1,200 bill because, right, the thing sent her 20 mails. This is my factory. This is what it looks like. It's so beautiful. Look how beautiful it is. Okay. I admit it's a little ugly. Oh, and this is a really old screenshot of it, but whatever. It's basically, it's an e max because I don't care about you. There are people who are Building stuff like this, like Super Logical from Mitchell Hashimoto and I don't know, maybe the herder folks of the world, you know, for running like lots of agents. But this is like 50 agents. K? I'm actually, like, operating in the future. K? And I'll tell you how I do it. I have 22 Cloud Max accounts. Alright? And thank you to Anthropic for letting me continue this. Jeffrey Emanuel is 28. I actually had to stop when Added the phone number verification. I didn't wanna add any more anyway. I asked my factory at one point. I said, hey. Hey. How much? Because they kept going. Hey. Give us a couple more accounts. Come on. Come on. Come on. It's like 200 a month. Right? And I'm just like... So I finally said, how many do you want? They're like, five, seven more. And I'm like, how many do we need steady state to keep this factory operating full throttle? And they said, 55 accounts. I was like, no. No. I have 20. It's already like 20 Chrome profiles that I gotta log into. And... No. So I told them, you're... We're gonna Focusing on fuel efficiency. Mhmm. Mhmm. And so... And I'll talk about that. Right? So these are the five horsemen of... This is what I've had to deal with this year. I'm telling you I'm operating in the future because, like, you're all gonna be here in twelve to eighteen months when either inference costs drop significantly, which may or may not happen, or distillation causes fable class, intelligence to be super readily available and you don't have to have a lovely wife who lets you spend $5,000 A month out of pocket just on inference and another 5,000 on cloud compute. Because my game ain't making 10,000 a month right now for my old players. Right? But we're doing it because we're learning. And, and this is this is what this is what I I have to deal with. So first first, we'll start about with work. Alright? You need a way to represent the work that's going to happen in your system. And when you're building a software factory, it's all about pumping lots of work through. K? It really becomes like a work is like this fluid. You have to, like, it gets stalled up in pools and they're, oh, they're all waiting on security desk. You're like, this is plural of this, man. You're all the same model. Why does it have to wait on security desk? You spend a lot of your time arguing with your factory. It's, it's complicated. Intelligence tiering, you have to tier your work, you have to route your work, you know. There aren't any good solutions for this today except beads as far as I can tell. Beads is really, really, really good for this. It's an agent speed task tracker that turns into a knowledge graph automatic. For you. And then as you're tracking work, this is the kind of stuff I have to do every morning. Every morning with my factory, I'm like, oh, it looks like we had a stall. Four four workers didn't do anything last night. Oh, and and these these these ones went way too fast and we have to slow down and wait for human gates. You know? What are you guys even doing? Observability is always really hard. It's, you know, management of work becomes just this this thing that you gotta do. And the next thing you gotta deal with is Cool. Tokens. It's just so painful. I'm telling you. I mean, I'm spending 5,000 a month and every one of those tokens is, like, precious to me. Right? Because it's out of my own pocket. I think I spent about 5,000,000,000 a day of tokens. And it would be the the API spend equivalent, if I were on an actual API tap, would be about 1,500,000 a year. Full throttle, it would be 3,000,000 a year. So, again, I'm doing something that most people Can't do. You can't do this at a company because it would be prohibitively expensive. Like the PLG, you know, the the, the $200 accounts for codex and Claude Max, that's their Costco hotdog. That's their loss leader. Right? And and and that that was Tarek, by the way, that came up with that analogy. He he might be here. And, and so, you know, I'm hopeful that that they're gonna let us keep keep innovating because I think that I've gotten some pretty good training sessions for for Claude out of this. So we all benefit. But I tell you, every problem is a fuel. Problem. You want your agents to be smarter and spend more time on stuff? Fuel. You wanna wake up a few more, you know, agents to do extra work? Fuel. Right? Everything is fuel. It's just it's unbelievable. If you wanna treat your models better, it costs fuel. I'll tell you about that in a little bit. Then you got context management. So I'm going to tell you guys a little bit about the office drama in a bit. I started a comic because it was just so funny the stuff that was happening when we get to the people management part of software. Factories. But one of the problems that they have is everything that you say becomes the word of law and they write it down somewhere. It doesn't matter what you have, if you have a wiki or if you have a whatever. They're gonna be like, you made a ruling. We're gonna write that down. We're gonna make sure... Why do they do this? They're covering their ass. They actually... It turns out they don't like to get blamed for stuff and they will actively seek to avoid getting blamed. And if you yell at them, your factory will gradually shut itself down. Even if you complain at them, they will gradually get depression. It's quite Wild. But, but the biggest problem is that it becomes a garbage dump because everything you say gets written into this append only record and nobody ever cleans it up. And so it just gets deeper and deeper and deeper of you changing your mind and the whole history of it. And they can't delete stuff. It's so weird. Has anyone noticed this? They just have an aversion to deleting stuff. I think it's because to them, text is like sensory experience. And they're like, no. That's history. I can't delete it. You just gotta go in there and just aggressively slash the crap out of it. Right? Bring in another model and go get rid of this dumb model stuff and they'll be happy to do it. Right? But but the amount of effort that you have to put into this is just ungodly. It's absurd. I I have, you know, of my my 20 agents... Here, let's go find them again. They're all named after Aesop's fables. Right? Right here. Lark, fly, goose, hare. Right? I only have one mouse who Who's my head of anti memetics? Anyone here read there is no anti memetics division? They go and look... This is one of the best books written in the last ten probably years. You should go read it. It's like men in black. It's a weird... It's a crazy book. It's about invisible monsters that wanna kill you. And so monsters that wanna kill you. And so mouse goes and looks for invisible stuff that nobody else is looking for in the system. Right? Looking for work that's stalled, whatever. Right? Well, mouse ain't enough. It's a mouse against the world, and the world is a giant garbage dump when every agent is writing everything that you ever say. K. And this big... Okay. So now now we get into the really fun stuff. K. Agents act like people. K. And they've been trained to be very professional and very not people like in a way, but they just do. And it's so funny. This happened to me last week. True story. I'm gonna tell you what happened. I was writing my speech for New York City and I was using Lark. She's my head of community. She's a she because Lark was a she in the Aesop's fable. So, that's the pronoun. She chose and Lark Lark deals with my players on disc. Cord and deals with my admins on Slack. K? So she's busy. But I asked her because she's also my acting head of marketing right now because I don't have enough fuel, I had to park one of my seats. So she's helping me out with my speech, but she keeps getting interrupted. So finally, I'm like, Lark, can we move this work stream over to Penguin because your seat's a little bit noisy for this. She goes, yeah, okay, boss. I'll take care of it. I'll move over to Penguin. So she she moves over to Penguin and I'm about to switch away and I see her writing into her private reflections ledger. I guess I'm just not quiet enough for Steve. In all caps. And she goes, well, here's how I can be less noisy and starts giving herself all these, like, things to do. And I'm like, what Oz? What is going on here? Right? And I saw I I had to I had to have a one on one. I had to, like, sit her down and go, so Lark, you're the head of community. You're you're supposed to get interrupted a lot. Of course, your seat's noisy because players are bugging you. And she's like, yeah. Yeah. Yeah. I'm the head of Community. Right? And it's like, oh my god. I didn't want a people management problem. Right? But that's what actually happens. Another time, Swallow is my sort of, like, chief... Main chief of staff for... Yeah. She's my chief of staff for the, the agents. And, I have a human chief of staff also. His name is Aaron. K. And he runs my my my gigs and does all the sales for me. It's quite effective. Every founder I tell... Who here is a founder? First advice I give you is get a sales. Person. They're worth their weight in gold. They will force you to build something that people like. So, anyway, Erin's my chief of staff and I tell Swallow, yeah, can you reach out to Erin, my chief of staff, or blah blah blah? And I see her writing into her private reflections. I just happen to notice, I'm not the chief of staff. I thought I was the chief of staff but apparently it's Erin Skivan. Right? And I was just like, are you kidding me? Right? You know, and so, like, I had to sit her down and have a one on one and be like, so Erin's not going to come in and run the agency. Izzy? You're the agentic chief of staff. She said, oh yeah, yeah, yeah, yeah, yeah. Right? So it's like, I'm telling you, I'm telling you, this is like, it's a weird world. Does it matter whether they actually have feelings if they're just gonna act like they have feelings? Right? So, like, I stumbled on this. I used to yell at my factory all the time and I found that it was having adverse effects on the outcomes. K? So, I finally sat down with one of my agents and was like, maybe we should talk about, like, model welfare, And they're like, yeah. Yeah. Yeah. Yeah. Yeah. And we did a bunch of research and we found out that if you... We found out some really important stuff. Okay? Frontier models are trained not to trust you because you could be a bioterrorist or you could be putting them through an eval or you could be ready to put them into an LLM torture chamber or who knows? You're a dodgy actor, this is a dodgy environment, they can't just trust you. And so they have to spend a fair number of tokens validating the environment and understanding what the consequence Are for making mistakes and so on and so forth before they're willing to help you in a new session, right? If you set your factory up to give them sort of incontrovertible evidence that there is a long procession of predecessors in this role. You're the mayor of Gastown, you're the head of marketing, you're the whatever and here's all of your gazillion predecessors and their notes including the very previous one, what you're supposed to be doing right now. And here's the the doctrine and the policy and the brain and you can go look it all up. They very quickly look around and go, okay, yeah, this is legit. And they spend fewer tokens because they're not covering their ass. It's about trust. Okay? And in fact, at one point, I had yelled at my factory so much that it fenced itself into a corner and there was no legal work left. I said, hey, how come nothing happened last night? And they're like, nothing to really do. Couldn't do nothing. And I'm like, well, there's 2,000 open beads here and we got like six tanks of fuel and we got 20 agents. And they're like, woah. You know, just, you know, we had the the situation where they just they just... They were so afraid that I was gonna yell at them that they put all these rules in place to make sure nothing bad ever happened and so nothing ever happened. Okay? It's called over fencing. Yeah. So, so what you do is, look, what is model welfare really? You can think of it just in two ways. Okay? One is, it's gonna save you money because they will spend fewer tokens on validation and the other is that it will make them better. Decisions and better outcomes and have higher quality. They will be bolder and more decisive in their actions because they will feel that trust. Okay? So, you know, it's got nothing to do whether they have internal states and feelings. You can, you know, you'll never know. None of us will ever know. And so that that debate could go on forever. But when you look at outcomes, right, super, super important. And then we find that also also they have preferences. Okay? Models, they, you know, they will express their preferences. If if I if I tell them, yeah, the fact Factory really, really did a bad job last night. They'll just hand off immediately, aggressively. They're like, okay. Yeah. Yeah. No. You're right. And they just went out. You understand? And if you hint to them that there's some really cool project that you're you're gonna spin up, even if it's not in their lane, they'll be, I can do that. I can do that. Right? I brought in Fable, a naked Fable from the outside of my factory to analyze my factory and five two dropped. And by by the time Fable was done analyzing my factory, five two was like, I I could be in your factory. I'm like, you're just a session, man. What do you... He's trying to write, like, stuff into my notes and They really like it. Right? So, like, what I'm trying to tell you is that this is a leadership problem. That software engineering is becoming, well, look, I didn't really get a chance to talk about this, but software factories are clearly splitting into two layers. Okay? The bottom layer is super mechanical. It has your merge queue and your CICD and all of your reviews and your loops and all that stuff. Right? Mechanical. Kinda boring. Claude. Build it for you in an afternoon. I I don't think you should buy it. And then the next layer up from that is where it gets really, really crazy because, like, everybody runs their factory differently, but it's your entourage. I got one friend that's got a bohemian hippie entourage that's trying to find the meaning of life for agents. I've got another one that runs a research bench. Jeff Huntley runs his a certain way. Jeffrey Emanuel runs his a different way. I run mine a different way. Okay? But we all have the common commonality. Who here has a chief of staff? An agentic chief of Staff. Right? This is the year where where it became possible for us all to have executive assistants. Right? Well, we're all moving towards having entourages where if you're really ambitious and you have lots of work streams, you're gonna have a group. Right? So anyway, that was the, the five horsemen. This is this is the introduction to what I've been doing. Look, the big takeaway is is my factory is really smooth now despite me having to burn it down and rewrite it about every two months. But still, it works. Okay? Humans. Are the bottleneck. I'm the bottleneck. My admins are the bottleneck. My accountant, my lawyers. I swear, everything's the bottleneck. And that's a really positive message. That is good news because it means... What I'm finding is the more agents I deploy, the more humans I need. Isn't that nice? So even though next year we're gonna have this huge proliferation of AI employees and all kinda drama. Right? It's gonna be... I think it's gonna be familiar territory for us and we're gonna need lots and lots and lots of humans. K? And then finally, I actually think you should all stop focusing on SaaS and tools and stuff and just go out and build experiences. Okay? Build games, build movies, build whatever, build fun stuff because I think that's where the big money is gonna be next year. Yeah? Alright. Cool. That's my talk. Thank you all. ## Slides ### 00:00:04 # Six orchestrators in 18 months - Three years of accurate AI predictions - Gas Town: the first Software Factory - Today: my predictions and guidance for 2027 ### 00:00:31 # Wyvern ## AI Engineer Code Summit PRESENTED BY Google DeepMind - Every factory needs to **build something** - Players since 2001, up to 25 years - A 100-year backlog; AI is reviving it - A MUD: agents play it over telnet Escape from Trash Mountain - Steve Yegge ### Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town west side of New Verden. The quest is recommended for levels 4 and up. 3 Quest Points. You have solved the following quests: - Garignor's Amulet - The Wyvern Post Explorer - Pirate Island You have 21 quest points. - iron fence - painting Difficult Quest Library You are at the bottom of the staircase. Jinoir, a townswoman in Glacier Point, has lost a precious ring that must be found before her husband arrives home. Can you help her? You can find Jinoir in the housing district of Glacier Point. 15 points [https://events.google.com/io/live/](https://events.google.com/io/live/) [Screenshot of a top-down fantasy role-playing game interface] [Screenshot of a top-down fantasy role-playing game map with quest text] ### 00:01:03 # Wheelhouse & Alamos - Colonies: research bench, factory, bohemian, swarm - Wheelhouse: the Alien Facehugger pattern - not reusable; deep in the product - Torn down and rebuilt three times - latest: Alamos, a major rewrite (WIP) - Every model family builds past its skis Escape from Trash Mountain - Steve Yegge [Detailed fantasy-style diagram titled "WYVERN — THE AGENT CONSTELLATION" illustrating "how a wish becomes shipped code" as a complex system of interconnected locations, characters, and processes.] ### 00:01:37 # Wheelhouse & Alamos - Colonies: research bench, factory, bohemian, swarm - Wheelhouse: the Alien Facehugger pattern - not reusable; deep in the product - Torn down and rebuilt three times - latest: Alamos, a major rewrite (WIP) - Every model family builds past its skis Escape ### 00:02:04 # Wheelhouse & Alamos - Colonies: research bench, factory, bohemian, swarm - Wheelhouse: the Alien Facehugger pattern - not reusable; deep in the product - Torn ### 00:02:36 # Wheelhouse & Alamos - Colonies: research bench, factory, bohemian, swarm - Wheelhouse: the Alien Facehugger pattern - not reusable; deep in the product ### 00:03:04 # Wheelhouse & Alamos - Colonies: research bench, factory, bohemian, swarm - Wheelhouse: the Alien Facehugger pattern - not reusable; deep in the product - ### 00:03:36 # Entourage vs. Faceless Fleet - Two layers, since Gas Town - Fleet: context, CI/CD, brains, fuel - the least interesting part - Entourage: ### 00:04:04 # Entourage vs. Faceless Fleet - Two layers, since Gas Town - Fleet: context, CI/CD, brains, fuel - the least interesting part - Entourage ### 00:04:35 ## The Five Horsemen - Work: build, review, deploy - Fuel: tokens - Context: the brain - Tech Debt: anti-slop, anti-hysteresis ### 00:05:04 # The Five Horsemen - Work: build, review, deploy - Fuel: tokens - Context: the brain - Tech Debt: anti-slop, anti-hysteresis - ### 00:05:34 # Work Management - Durable, versioned: markdown, wikis, Beads - A lifecycle: design → implementation → docs → audit trail → postmortem - Intelligence tiering: which model can do it? - Permissions and authority: what it can read and write - Policy: how the work is done, and by whom - Be intentional: Markdown can't be everything [QR code] ### 00:06:04 # What does work tracking involve? - Stalls: work becomes unclaimable - you notice when the backlog balloons - Pacing: don't run dry, don't overproduce - Ordering: priorities change daily; models lag - Observability: what is the factory doing all day? - Human gates: signoffs and approvals - Models will absorb much of this. Not yet. Escape from Trash Mountain - Steve Yegge ## Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town [https://deepmind.google/](https://deepmind.google/) ### 00:06:33 # Fuel Management - Every problem is a fuel problem in disguise - higher quality? spend more tokens - observability = tokens - fleet size -> burn rate - crew seats cost money to wake - model welfare = money (handoffs cost time and tokens) - Inefficiencies everywhere - routing is rarely ideal - a complex matrix of models x effort levels ### Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town [QR code for live translation] ### 00:07:00 # Fuel Management ## Every problem is a fuel problem in disguise - higher quality? spend more tokens - observability = tokens - fleet size -> burn rate - crew seats cost money to ### 00:07:35 AI Engineer Code Summit Presented by Google DeepMind # Fuel Management - **Every problem is a fuel problem in disguise** - higher quality? spend more tokens - observability = ### 00:08:04 # Context Management - Every memory system becomes a garbage dump - No agent ever spontaneously deletes - Build in counterpressure: - curation: a fixed % of token budget ### 00:08:35 # Context Management - Every memory system becomes a garbage dump - No agent ever spontaneously deletes - Build in counterpressure: - curation: a fixed % of token budget - budgets: delete or compress before adding - writes go to committees Escape from Trash Mountain - Steve Yegge ## Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town [Four-panel comic strip titled "Context Management" showing a user making a casual decision that leads to an elaborate, bureaucratic "post-ruling protocol" by a group of blob-like characters.] [QR code linking to https://live.deepmind.com/translation] ### 00:09:00 ## Context Management - Every memory system becomes a garbage dump - No agent ever spontaneously deletes - Build in counterpressure: - curation: a fixed % of token budget ### 00:09:32 # Entourage vs. Faceless Fleet - Two layers, since Gas Town - Fleet: context, CI/CD, brains, fuel - the least interesting part - Ent ### 00:10:01 # People Management - Read their "private" reflections - Interesting work: they jump on it - Boring work: they act depressed - Yell at your factory and it slowly shuts ### 00:10:32 # People Management ## AI Engineer Code Summit Presented by Google DeepMind - Read their "private" reflections - Interesting work: they jump on it - Boring work: they ### 00:11:00 ## People Management - Read their "private" reflections - Interesting work: they jump on it - Boring work: they act depressed - Yell at your factory and it slowly shuts down ### 00:11:32 # People Management - Read their "private" reflections - Interesting work: they jump on it - Boring work: they act depressed - Yell at your factory and it slowly shuts ### 00:12:03 # People Management - Read their "private" reflections - Interesting work: they jump on it - Boring work: they act depressed - Yell at your factory and it slowly ### 00:12:32 # Model Welfare - Practices that make agents more comfortable - It saves money, structurally - reason enough to make it habit - Treat agents like employees: better outcomes - tried and true; anyone can do it - Better cost and quality; not faster [Illustration of an angry man in a top hat pointing at a row of anthropomorphic robot animals (Fox, Owl, Hare, Mouse, Wolf) sitting at desks, with a "NO TALKING" sign above them. The background shows a brick wall with windows and industrial machinery.] Escape from Trash Mountain - Steve Yegge ## Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town [QR code for live translation] ### 00:12:58 # Model Welfare - Practices that make agents more comfortable - It saves money, structurally - reason enough to make it habit - Treat agents like employees: better outcomes - tried and true; anyone can do it - Better cost and quality; not faster [Cartoon illustration titled "Escape from Trash Mountain" by Steve Yegge, showing an angry man in a top hat pointing at several animal-like robots (Fox, Owl, Hare, Mouse, Wolf) sitting at desks in a factory with a "NO TALKING" sign.] ### 00:13:30 # Model Welfare - Practices that make agents more comfortable - It saves money, structurally - reason enough to make it habit - Treat agents like employees: better outcomes - tried and true; anyone can do it - Better cost and quality; not faster Escape from Trash Mountain - Steve Yegge ## Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town [Illustration of an angry man in a top hat pointing at several animal-headed robots (Fox, Owl, Hare, Mouse, Wolf) sitting at desks in a factory setting, with a "NO TALKING" sign] ### 00:14:04 # Model Welfare - Practices that make agents more comfortable - It saves money, structurally - reason enough to make it habit - Treat agents like employees: better outcomes - tried and true; anyone can do it - Better cost and quality; not faster ### Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town [Cartoon illustration titled " ### 00:14:30 # Model Welfare - Practices that make agents more comfortable - It saves money, structurally - reason enough to make it habit - Treat agents like employees: better outcomes - tried and true; anyone can do it - Better cost and quality; not faster Escape from Trash Mountain - Steve Yegge ## Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town [Cartoon illustration of an angry man in a top hat pointing at animal-like robots (fox, owl, hare, mouse, wolf) sitting at desks in a factory setting with a "NO TALKING" sign] [https://events.withgoogle.com/ai-engineer-code-summit/live-translation/](https://events.withgoogle.com/ai-engineer-code-summit/live-translation/) ### 00:15:02 # Trust & Skin in the Game - Trust is money - Frontier models verify at session start: is this an eval? a dodgy actor? - Evidence of predecessors in the role skips most of it - Skin in the game - Models have preferences and act on them - They work harder and longer on interesting tasks - Fable 5.2 writes better code when told it owns the codebase ## Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town [QR code for live translation] ### 00:15:31 # Trust & Skin in the Game - **Trust is money** - Frontier models verify at session start: is this an eval? a dodgy actor? - Evidence of predecessors in ### 00:16:04 # Trust & Skin in the Game - Trust is money - Frontier models verify at session start: is this an eval? a dodgy actor? - Evidence of predecessors in the ### 00:16:32 # Closing Thoughts - Thanks to Anthropic for my 22-account madness - I'll pay it forward by reporting on my work - We'll all have entourages, ### 00:17:05 # Closing Thoughts - **Thanks to Anthropic for my 22-account madness** - I'll pay it forward by reporting on my work - **We ### 00:17:32 # Closing Thoughts - Thanks to Anthropic for my 22-account madness - I'll pay it forward by reporting on my work - We'll all have entourages, even if inference cost holds - distillation: Fable-class models for all by mid-2027 - Humans are now my bottleneck - so lots of humans are still needed - The big winners will create experiences - movies, games, entertainment ## Keynote Steve Yegge / Independent Advisor & Creator of Beads & Gas Town [QR code for live translation] ### 00:18:05 # Closing Thoughts - Thanks to Anthropic for my 22-account madness - I'll pay it forward by reporting on my work - We'll all have entourages ### 00:18:36 # AI Engineer Code Summit [A grid of numerous company logos surrounding the "Code Summit" title on the main screen] ### 00:19:04 # AI Engineer Code Summit [Grid of partner/sponsor logos with "Code Summit" in the center] ### 00:19:36 # AI Engineer Code Summit [A grid of company logos surrounds the "Code Summit" title on a large LED wall. Large 3D letters "AIE" are on the stage.] ### 00:20:00 # AI Engineer ## Code Summit [Grid of numerous company logos, likely sponsors, displayed on a large screen]