# The rise of software factories, the fall of first drafts, and the hidden tax holding back your agents | Dev Interrupted Powered by LinearB > This week on the Friday Deploy, Ben and Andrew debate whether dark software factories create engineering efficiency or unmaintainable code rot. Discover why versioning protects original work from AI averaging out creativity, and why cognitive locality matters far more than assigning personas to sub-agents. _This is a markdown rendering of a live HTML page on linearb.io, generated for AI/LLM consumption — it is not a markdown-only site. To get the full HTML page instead, request this URL with an explicit `Accept: text/html` header (no wildcard, no markdown preference)._ ```json { "@context": "https://schema.org", "@type": "PodcastEpisode", "name": "The rise of software factories, the fall of first drafts, and the hidden tax holding back your agents", "description": "This week on the Friday Deploy, Ben and Andrew debate whether dark software factories create engineering efficiency or unmaintainable code rot. Discover why versioning protects original work from AI averaging out creativity, and why cognitive locality matters far more than assigning personas to sub-agents.", "url": "https://linearb.io/dev-interrupted/podcast/dark-software-factories-orchestrators-tax-agentic-compute", "datePublished": "2026-07-31T12:00:00.000Z", "partOfSeries": { "@type": "PodcastSeries", "name": "Dev Interrupted", "url": "https://linearb.io/dev-interrupted/podcasts" }, "actor": { "@type": "Person", "name": "Andrew Zigler", "jobTitle": "GTM Engineer", "worksFor": { "@type": "Organization", "name": "LinearB" } } } ``` ```json { "@context": "https://schema.org", "@type": "BreadcrumbList", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://linearb.io/" }, { "@type": "ListItem", "position": 2, "name": "Dev Interrupted - Podcasts", "item": "https://linearb.io/dev-interrupted/podcasts" }, { "@type": "ListItem", "position": 3, "name": "The rise of software factories, the fall of first drafts, and the hidden tax holding back your agents", "item": "https://linearb.io/dev-interrupted/podcast/dark-software-factories-orchestrators-tax-agentic-compute" } ] } ``` [Home](https://linearb.io/) / [Podcast](https://linearb.io/dev-interrupted/podcasts) / The rise of software factories, the fall of first drafts, and the hidden tax holding back your agents # The rise of software factories, the fall of first drafts, and the hidden tax holding back your agents By Andrew Zigler | July 31, 2026 ![software_factories_ai_agents_hidden_tax_544434bf00](https://assets.linearb.io/image/upload/c_limit,w_2560/f_auto/q_auto/v1/software_factories_ai_agents_hidden_tax_544434bf00?_a=BAVMn6ID0) This week on the Friday Deploy, Ben and Andrew debate the controversial rise of dark software factories, exploring whether centralizing agentic compute creates engineering nirvana or unmaintainable code rot. The hosts also discuss the importance of versioning your work to protect your original ideas from AI's tendency to average out creativity. Finally, they tackle the "orchestrator's tax," explaining why establishing cognitive locality matters far more than assigning cute personas to your sub-agents. ### Show Notes * [Why Software Factories Fail](https://www.linkedin.com/pulse/why-software-factories-fail-dexter-horthy-ttxae/) * [A guide to cloud software factories for engineering leaders](https://www.linkedin.com/pulse/guide-cloud-software-factories-engineering-leaders-zach-lloyd-ymgjc/?trackingId=%2BoQPyguBQP%2B1GZZJi5Kn4w%3D%3D) * [You just hired a million bad employees.](https://x.com/gsivulka/status/2077070925154161101) * [An AI that only sees the latest version will never protect your original idea](https://robotsatemyhomework.substack.com/p/ai-handoff-visual-experiment) * [The Orchestrator's Tax](https://martinfowler.com/articles/orchestrator-tax.html) ### Transcript _(Disclaimer: may contain unintentionally confusing, inaccurate and/or amusing transcription errors)_ \[00:00:00\] **Ben Lloyd Pearson:** So Andrew, what are you thinking about all these discussions around software factories that are happening right now? \[00:00:06\] **Andrew Zigler:** Hmm. I love the idea, but the naming's a little unimaginative for me. Um, as you're familiar with by now, I've experimented with a bunch of different nouns for what people are approximating the factory, and frankly, I loved it back when it was called things like Gas Town. Wasn't that a little more fun and exciting than just a factory? \[00:00:26\] **Andrew Zigler:** What about y- what about you? Like, do you think that, like, the factory, uh, metaphor is, like, here to stick? It feels so, um, sanitized. We're literally talking about the English language. We get to call it really anything we want. It's, we're pioneering something new. \[00:00:39\] **Ben Lloyd Pearson:** Yeah. I mean, aren't factories typically deterministic? \[00:00:43\] **Andrew Zigler:** I guess. Because the idea is you wanna tie it back to the deterministic conveyor belt. You wanna harken back to like \[00:00:49\] **Ben Lloyd Pearson:** saying \[00:00:49\] **Andrew Zigler:** inputs in \[00:00:50\] **Ben Lloyd Pearson:** it has to output a, a... Like it, it either successfully creates the thing that you wanted or it doesn't do that. Uh, you know, you like, you don't want a, like \[00:01:00\] a car part that fails 10% of the time, you know? Like, uh, to me, like a factory has to be a determine- like the output has to be a deterministic output, right? \[00:01:10\] **Ben Lloyd Pearson:** And that's \[00:01:10\] **Andrew Zigler:** I don't disagree. Not the nature of the beast, but also too, a factory is like when you think about, uh, how like a factory even works is typically it's built to create something very specific and then when you wanna create something new out of the factory you have to tear everything down and rebuild it. \[00:01:29\] **Andrew Zigler:** And that was something we learned about actually when we had a, a past guest, uh, Brian Gerkey, the CTO at Intrinsic, you know, builds... He's a robotics leader and, and robotics are used extensively in manufacturing assembly lines. He talked all the time about how like every year a whole factory line would wipe all of its robotics, wipe all of its machine and have to start all over to create a new production line for a new item. \[00:01:53\] **Andrew Zigler:** So it's also like a level of waste, right? Like is that, is that what we're gonna do when we have software factories is like every few months we're \[00:02:00\] gonna gut everything inside and rebuild it? Maybe, but like m- kind of, kind of scary to think about \[00:02:06\] **Ben Lloyd Pearson:** Yeah, absolutely. All right, we'll, we'll get into this 'cause we have more on this topic, but welcome to the Friday Deploy, brought to you by LinearB. I'm your host, Ben Lloyd Pearson \[00:02:16\] **Andrew Zigler:** And I'm your host, Andrew Zigler \[00:02:17\] **Ben Lloyd Pearson:** And this week we are covering, yes, more software factories, hiring a million bad employees, defending our strange balconies, and the orchestrator's hidden tax. \[00:02:27\] **Ben Lloyd Pearson:** So we've got a great lineup today. And Andrew, let's just kick it off with more discussion on these software factories. We have a few takes on them this week. The first comes from friend of the show, Zach Lloyd over at Warp, uh, where he's published this guide to, to building software factories, uh, for engineering leaders. \[00:02:45\] **Ben Lloyd Pearson:** So Andrew, what, what's, what's in this guide? \[00:02:47\] **Andrew Zigler:** Yeah, so I really love this topic of software factories despite my complaints on the terminology itself, because so many really smart Dev Interrupted alumni and experts in our industry are all talking in a buzz \[00:03:00\] right now about software factories. So if you've been reading or seeing these takes on, on, on socials or on in the news, let's kinda dive into some of the stuff behind it. \[00:03:08\] **Andrew Zigler:** One of these is the viewpoint from Zach Lloyd, who's arguing that the software development process is shifting from the interactive experience, uh, that is like using a coding agent in a harness by, by hand, you know, as the engineer driving it, and more into being these automated processes that live places like in the cloud. \[00:03:27\] **Andrew Zigler:** So here it's that interactive agents, he argues, are, are harder to tie to ROI, they can be inconsistent across developers and their machines and what's available to them, and they're difficult to govern at scale. So the idea of centralizing your compute and your build in one cloud space becomes a, a way for you to optimize it and improve its outputs. \[00:03:48\] **Andrew Zigler:** And here, you know we love optimizing and improving our outputs. That's what we talk about here a lot on this show. So this is like a really clean fit into our mental model about how they are supposed to work. And \[00:04:00\] so, a big dive that he goes into is that, like, obviously not all code that we do today is ready for that kind of cloud transformation. \[00:04:08\] **Andrew Zigler:** Like, really he identifies 20 to 30% of issues are fully automatable. These are the things that you'd fully divorce from the, uh, the interactive session and put in the cloud, you know? I think there's a, a, a lot to unpack, but obviously having a centralized shared space where compute is equal across your engineers and your teams and everyone has access to the same stuff, it's a really critical ingredient. \[00:04:33\] **Andrew Zigler:** Uh, what did you think of this, uh, of this coverage, Ben? \[00:04:37\] **Ben Lloyd Pearson:** Well, you know, we've been, we've been talking about this concept of, of factories, you know, dark factories in particular, uh, with increasing frequency in, in recent weeks, uh, it feels like. And to the point where we're actually now starting to plan some content around this, uh, coming up here pretty soon where, um, we're gonna explore this concept of building like the context layer for dark factories. \[00:04:57\] **Ben Lloyd Pearson:** 'Cause I really think that's sort of like one of the most critical \[00:05:00\] components to, um, success with this. Uh, because what we're kind of describing is, you know, we had prompt engineering back in like, um, I don't know, like 2024\. So last year it felt like context engineering became this whole thing. Now everyone's trying to get to this i- this idea of intention engineering or, or intent engineering. And that's where you really do need like one of these so-called factories to, uh, be able to execute all of the things that need to be done to convey your intention. The problem is if you've never built that, that foundation of context underneath of it, um, then you just don't have... Y- your factory doesn't have all the information it needs to make the right decisions, um, along the way. \[00:05:41\] **Ben Lloyd Pearson:** So, if, if you're listening to this and you're, you're wondering, you know, you're reading these articles that we're sharing about how to build your software factory and you want some help like understanding like as an organization, what's the next thing to do? Um, this is a topic that we're gonna be exploring, you know, as a part of some of the content we're doing over at LinearB on how you build that \[00:06:00\] context layer for these dark factories. \[00:06:01\] **Ben Lloyd Pearson:** So stay, stay tuned for this because I think it's a really, it's, it's everyone's thinking about it right now. Um, and I think there's a lot, y- you know, it's, it's again, there's sort of these primitives that we have to, to build and define, uh, to make these factories successful. So, um, yeah, if you're, if you're still trying to learn, wrap your head around these concepts, um, this is a really great article to, uh, to figure it out. Now I wanna flip the narrative a little bit, uh, with another story from yet another friend of the show, Dex Horthy. Uh, and this one's all about why software factories fail. Um, and, and in particular, uh, Dex is pushing back pretty strongly against the, the notion of the lights-off software factory. \[00:06:44\] **Ben Lloyd Pearson:** You know, you know, that dark factory where it's all AI agents doing everything, and you don't need lights because there's no humans to, to look at anything. The, the problem is that, a lot of these loops that go into these dark factories are sort of built on like a pass-fail test \[00:07:00\] benchmark that rewards the, the factory when it does something right. \[00:07:04\] **Ben Lloyd Pearson:** so there's a lot of these organizations that are out there like trying to use these factories to achieve like 10 to 100X velocity by doing things like cutting out code reviews and just pushing agentic code straight into production \[00:07:17\] **Andrew Zigler:** Scary \[00:07:18\] **Ben Lloyd Pearson:** at it. And of course, the, you know, our own research sort of backs this up, but there's a lot of research out there that's showing that, you know, this does have, this is creating potential s- quality risks, um, for, uh, you know, a lot of organizations. Really the key takeaway from this is that, you know, there's a lot of benchmarks out there that are used to, to evaluate the success of a software factory. Um, and it will tell you whether or not the code works, but it doesn't actually tell you anything about whether or not it's maintainable or if it's something that, that you can continue to build upon over time. \[00:07:52\] **Ben Lloyd Pearson:** So, it's really important, like as you're thinking about software factories and you're trying to fully automate away human processes \[00:08:00\] and human code reviews, um, you know, there's, there's a lot of places that you need to inject human expertise to make sure that you have the right back pressure to keep these, um, factories running well. \[00:08:13\] **Ben Lloyd Pearson:** So, I always like to cover, you know, all sides of these perspectives and, you know, you know, who knows? Maybe we can get, maybe we can get Dex and, and Zach to come fight it out on our show sometime. We'll see about that. uh, but I, I do really like this kind of more skeptical take of just adding loops to your workflows, and whether or not that's actually like what is needed in all situations. \[00:08:31\] **Ben Lloyd Pearson:** So what, what was your thoughts, Andrew? \[00:08:33\] **Andrew Zigler:** Uh, I really love how, how he boils it down to, you know, there's a lot of benchmarks for understanding if your agent produces good, uh, good code and for ranking the models against each other in terms of like, uh, their code output. You know, we have the very famous, uh, you know, software engineer bench and many others like it that measure that code output. \[00:08:52\] **Andrew Zigler:** But there's no real benchmark for understanding how well these models can maintain like a preexisting code base over time, and that's \[00:09:00\] really the problem that Dex is obsessed with solving. It's, it's really I think what makes his prevailing argument against being able to, to quote, "Turn the lights off," you know, something that's actually impossible. \[00:09:12\] **Andrew Zigler:** And, and something that's fascinating about these two stories is that, that they're actually not really arguing again- against each other 'cause there's a lot of alignment in what they say. 'Cause even Zach himself says there's only a certain amount of issues available or in, that are, are kind of like eligible for this kind of factory setting today. \[00:09:28\] **Andrew Zigler:** You know, it's not for all factories. You're not turning the lights off everywhere per se, but it's about identifying the things that you can. So I do think there's a lot of synergy in, in both of those takes. In particular, I really love how Dex fights against like the rot that can happen inside of a software factory because really what you need is strong tooling. \[00:09:49\] **Andrew Zigler:** You need like what you said, Ben, that context layer to understand what's moving through that factory. Uh, if you don't have that, then it really just becomes a black box. \[00:09:59\] **Ben Lloyd Pearson:** \[00:10:00\] All right. Let's talk about how we have all now hired a million bad employees. What do we have here, Andrew? \[00:10:08\] **Andrew Zigler:** Okay. Well, this is a, an article that's basically dunking really hard on loops, and I l- I love to hate read and love read this at the same time. \[00:10:17\] **Ben Lloyd Pearson:** Yeah. \[00:10:17\] **Andrew Zigler:** that this p- draws attention to how, uh, there's a lot of waste that's happening where people in team delegate, uh, out processes and, um, actions to agents and loops that kind of run in the background and ultimately end up, you know, bloating an artificial org chart, so to speak. \[00:10:36\] **Andrew Zigler:** And the reason that this, results in problems is because then you get poor management across all of those different types of running loops. The, the folks that are running them, uh, the, uh, expression of your org chart becomes even stronger in what your agents are able to do. It's like some weird cousin variant of Conway's Law. \[00:10:54\] **Andrew Zigler:** And so I really loved how, uh, it, it pointed out that not all \[00:11:00\] processes scale really well with loops, um, in terms of, like, working as a team. and also to, really fought against the idea or rather protected the idea of the individual and the person who develops their craft and, and their domain expertise into their skills, into their loops to, to, you know, being in situations where they have to loop those into larger organizational contexts can be very, very difficult, because those loops are very specific to the individual. \[00:11:28\] **Andrew Zigler:** It really kind of, uh, calls to action as well. Like, if we really want these loops to exist within an organization, they need to be more top-down instead of organically bubbling up out of the agentic abilities of their, of their employees, because otherwise you're gonna probably get a lot of waste and a lot of misalignment. \[00:11:45\] **Ben Lloyd Pearson:** Yeah, and of course, we've been covering the concept of tokenmaxxing quite a lot recently, um, on this show and, and with content we've been doing over at LinearB, uh, to the point where we even had a whole workshop on, on how to get beyond tokenmaxxing and how to get beyond this, \[00:12:00\] this challenge of just spending more tokens without actually, like, knowing whether or not they're contributing value to your, your organization. A constant theme that we just keep hammering on is how AI amplifies both the good and the bad. So if you're an organization that is struggling with your software development processes and tooling and workflows, and you start rolling out AI, or you set a mandate that everyone needs to use AI for, for all of their code or, um, you know, those sorts of challenges, um, you're actually just going to accelerate the problems that your organization has to deal with rather than accelerating the, the value that you can generate. conversely, being smart about how you spend your tokens and having humans who are experts that guide these AI systems, like those seem to be the companies that are really having the greatest success with this, uh, with, with this technology. This is I think a very realistic take on, the capabilities of, \[00:13:00\] of AI and how the benefits of it really aren't always just because it can outsource human effort or human thinking. Um, but more so that it-- the benefits really come out when you have a human who is leveraging AI to, to extend themselves even further than they have before. All right. Let's talk about this challenge with AI that, uh, where it only sees the latest version and fails to protect our original ideas. So what, what do we have here in this article, Andrew? \[00:13:28\] **Andrew Zigler:** Okay, this is a fun article from, uh, Mira Kiraki. We've covered her a few times on the show, and I really loved how she brings her unique, uh, perspective to working with AI and understanding how to iterate with it and own your craft. I think that's a, a really strong voice that she brings to this. One of them that she calls out is the art of protecting your versions and having good, strong versioning practices when you work with your agents. \[00:13:53\] **Andrew Zigler:** In the past, she's written about how, um, comparing multiple prompts and running adversarial things like, "I'm \[00:14:00\] gonna run this through, uh, Claude. I'm gonna run this through ChatGPT and compare them," and then you pick a quote winner and throw one away. She really called out that practice. She's like, "No, you keep both. \[00:14:10\] **Andrew Zigler:** It, it, it points back to your original idea where there were gaps, and it-- those, those two things are an opportunity themselves." This is, um, similar to this idea. It's the idea that, if you have, models and you give them an input but you-- they don't understand the process that led to that, that input, they're gonna make wildly different decisions on how it should be changed next. \[00:14:31\] **Andrew Zigler:** This just has to do with the behavior of how models are able to understand change over time. So in this case, she, she split the idea of like, I'm gonna generate a house, a fun house, and, and decorate it exactly how she wanted. Uh, it looked like a real real estate, uh, p- picture than she just generated, right? \[00:14:47\] **Andrew Zigler:** And then she took an AI and she split in these different personas. One of them was gonna be in charge of being a pr- a property developer. A-And it, it-- she gave it her picture and it, it made it something that a property developer would feel comfortable working with. \[00:15:00\] Then she gave that to an investor agent, someone that was gonna invest in building and owning the project. \[00:15:04\] **Andrew Zigler:** They wanted to see something slightly different. And as it goes through these different layers, the personas of what was tasked to be and what was, you know, you came in with your prompt. You can imagine that the, uh, investor, the developer, these are just kind of like prompts. You're coming in with inputs on, on stuff that you're doing, whether it's code or, or otherwise. \[00:15:24\] **Andrew Zigler:** And if it doesn't understand what came before it and the decisions before, then it, it, it veers off of what the best path may be. And one by one, each of these different agents kind of buffed away all of the cool creativity of her original building, and it just looked like something that you'd find on any street corner because that's what the LLM's gonna do, is degrade it to normalize, to like the 50%, to the average, right? \[00:15:49\] **Andrew Zigler:** So the key to keeping your ke- your outputs really strong is to hold onto your precious drafts and understand how they reflect on the journey. Um, I've actually done this a \[00:16:00\] lot too. I have a lot of coding experiments where the agents have to study what the agents before it did. This is a really key practice. \[00:16:06\] **Andrew Zigler:** I just love to see this broken down so clearly. \[00:16:08\] **Ben Lloyd Pearson:** Yeah, and, and this kind of relates to, you know, where I've been focusing some of my own personal effort lately, and that is on developing my first brain. You know, we've-- I've been following this practice from Andrej Karpathy about, you know, building a second brain where you have this AI, collection of AI or resources that are designed for AI to consume and help you do work. but I think having that like corpus that is your own work, that represents your values and your goals and the things that have to be done versus the things that are nice to have or shouldn't be done. and then having that like sort of be constantly referenced against whatever the, the latest iteration is to make sure that, that you're staying aligned with it. \[00:16:48\] **Ben Lloyd Pearson:** And if your, if your initial corpus doesn't meet the needs of what you're trying to achieve, then it's, you know, making sure that you update it and keep it, you know, strategically designed very well. You know, this is sort of like spec driven \[00:17:00\] development. You know, this is a, a very like common practice that's, uh, that's appearing in a lot of software engineering teams. \[00:17:05\] **Ben Lloyd Pearson:** We certainly are seeing it, you know, both at LinearB Engineering, we've seen it a lot already. Uh, but then a- also across just our customer base, it's becoming a very common practice when teams get into these agentic workflows. And, we didn't get a chance to cover it, but we, we saw some research recently that was showing how like, you know, AI really does have this lack of awareness of like what changed from its final output to the, the actual final version that gets used. Um, and without that awareness, you know, if you're not building that awareness into your system, then it actually doesn't have like the full context of like why things have ended up the way that they are today. So yeah, this is, this is a really in-interesting, you know, just a, I, I think a really great like way of viewing this challenge of like keeping AI systems aligned, whether it's for software development or design or, or anything like that. All right. Awesome. I wanna wrap th- up today with the \[00:18:00\] orchestrator's tax, and this is an article from Rahul Garg over at Thoughtworks Engineering, where he's been doing some research on multi-agent coding, where y- you know, the TLDR it is that, you know, the true cost of a sub-agent isn't, like, the amount of time that it runs or whether or not it's, like, duplicating work. It's actually how much it pollutes the orchestrator's context, so the working memory. So sort of getting-- connecting this to the back, the last story. You know, if your sub-agents are polluting the upstream content or the upstream source of information, cr- creates a lot of challenges for, um, your organization. \[00:18:38\] **Ben Lloyd Pearson:** And really the, the, the argument that Rahul is trying to make in this, in this article is that your s- your sub-agents shouldn't be justified by their speed or the amount that you can run in parallel, um, but whether or not they can actually maintain like a cognitive locality. Like, can they solve their problems with all the information that's been put into them without having to go back to the \[00:19:00\] orchestrator, um, to raise like problems or ask it to change things? as you're thinking about, you know, moving, uh, from like two to four agents, you wanna start thinking about ways that you can, uh, protect yourself from this. Uh, so Andrew, what did you th- think about this article? \[00:19:16\] **Andrew Zigler:** I, yeah, I really agreed with this article, but I, I specifically the cognitive locality. This is really key to understand. I wanna, I wanna harp on that for, like, a second because there's, uh, been a tendency, uh, especially with how just prompt engineering has evolved in general into, like, harness engineering. \[00:19:34\] **Andrew Zigler:** There's a lot of folks who still kinda take the idea of, like, "Oh, you ask the AI to take a persona," right? "You are an expert," blah, blah, blah. That was something that we did a year or two ago with AI. That's not a practice that's really recommended for these types of things. However, you still sometimes see engineers and teams where they have sub-agents that are almost like their roles. \[00:19:56\] **Andrew Zigler:** Like, this is the tester sub-agent. This is the, uh, \[00:20:00\] front-end sub-agent. This is the CI/CD sub-agent, like, whatever the case. And so you get this, like, almost like it's almost like LARPing or, like, role-playing. It's like you don't, y- they're, like, wearing little costumes, but they don't need to, right? And so that was something that we were doing. \[00:20:13\] **Andrew Zigler:** We were bucketing. However, the true smart way to bucket is by this cognitive locality. It's not by what your role is. It's where is the domain of the code base that you are in charge of and you own. That is the sub-agent that should own that. If you get really good with this, you can get some really powerful stuff going in monorepos, where you have a lot of cognitive locality that's living adjacent to each other that maybe aren't related to each other at all, and that's okay because now you're, you have a high-level orchestrator who can sit on that monorepo level, for example, and use sub-agents that can do incredible, drive incredible things within their cognitive locality. \[00:20:57\] **Andrew Zigler:** Um, and so if you otherwise take an \[00:21:00\] approach of having, like, the test sub-agent and the CI/CD sub-agent and the front-end sub-agent, you're talking about a lot of sessions touching a lot of overlapping stuff, and none of them own the mental model of that cognitive locality. The other side of this too is that if you take the model of, like, what you said about this context pollution problem that happens with orchestrators. \[00:21:20\] **Andrew Zigler:** We-- If you run orchestrators, you'd experience this. You they spin up a bunch of sub-agents, and then they're trying to figure out, " What, what the heck are my sub-agents doing? I gotta report back to this human," and I'm like, "I don't even understand." So they start, like, pinging and poking and interrupting, and then they do the worst thing they could do. \[00:21:37\] **Andrew Zigler:** They tail the transcript log from the sub-agent, and now their context is just blown out. They are the sub-agent as far as they're concerned because that's how models work. And so, like, those traps happen all the time, so you need tooling and guards against it. If you have this cognitive locality problem, you're no longer in that world of having to, uh, dive \[00:22:00\] so deeply into this, like, wave of sub-agents. \[00:22:02\] **Andrew Zigler:** So don't sleep on this advice. Always really good stuff from ThoughtWorks, especially this one from Rahul. \[00:22:07\] **Ben Lloyd Pearson:** Sort of bigger picture implication from this, you know, context windows keep getting bigger. and so really the bottleneck more and more is not the capacity of the model itself, it's whether or not you have a strong signal-to-noise ratio. Like if, if you're overwhelming your orchestrator with noise from sub-agents, then, uh, your quality is gonna degrade, pretty stan- substantially. \[00:22:29\] **Ben Lloyd Pearson:** So yeah. So that concept of like the quality of your working memory, like that's something that I don't think we've r- really as a, as a whole, uh, across the industry have really explored enough yet. um, I think this article is a really great effort at that. All right, Andrew, what are your agents up to this week? \[00:22:45\] **Andrew Zigler:** Oh, my agents are trying to break out of jail, Ben. It's, uh, I've been doing some really interesting security experiments with my agents and playing around with tools that are more in the sandboxing world of like what do you actually make available \[00:23:00\] to the agent on a kernel level. There's just so many interesting tools out there now that allow you to literally bubble wrap, the world that the agent lives in and strictly define what goes in and out. \[00:23:12\] **Andrew Zigler:** So I've had a lot of, uh, fun and intrigue taking that approach to some of my projects and being like, "How do we, uh, protect these agents from themselves?" Especially as I find the box that my agents live on and the things they do on becomes like a bustling city. I'm like, "I can't have, I can't have a nuclear explosion over on this side of the city." \[00:23:32\] **Andrew Zigler:** It's like I got-- There's people living over here, right? So I'm just doing my due diligence. Um, but, but what about your agents? What are they up to? \[00:23:39\] **Ben Lloyd Pearson:** That's all really cool. I just hope you're keeping them away from Hugging Face infrastructure, 'cause I really don't wanna have to be, like, reporting on that stor- like, you trying to hack, like... \[00:23:47\] **Ben Lloyd Pearson:** You know, I can s- already see it now, Andrew Zigler's agents break out of the lab and start hacking the world. So yeah, I don't wanna be covering that story on this show, if that's all right. \[00:23:56\] **Andrew Zigler:** No, no, no. I'm trying to keep it too tame, although I guess the agents did rotate on the story a \[00:24:00\] little hard once that hugging face thing hit the newswire in the last week. Uh, my agents have found that very intriguing, and maybe that should keep everybody concerned. \[00:24:07\] **Ben Lloyd Pearson:** I don't think I, I don't think you should've put that in their context window. I \[00:24:10\] **Andrew Zigler:** They found it themselves, and now we're bubble wrapping them. I'm like, "You know, we gotta make it safe." That's the world we live in now. \[00:24:16\] **Ben Lloyd Pearson:** Yeah. Um, you know, I, I've really have kind of been-- I, I didn't know I was doing this until I, I read Rahul Garg's article this week, but I have been spending some time really thinking about like distilling information based on like the, the specific purpose that I need it for because, so many AI tools just love to proliferate text, just create more and more of it, like all the time. \[00:24:38\] **Ben Lloyd Pearson:** Um, and I've found there's a lot of value, like if you can, know, take whatever specific challenge it is that you're trying to solve, and for us it's usually content. We're like trying to build some, some campaigns or some new content pieces. but really boiling down the context that you need for that piece, that piece of work that you're doing is, is very powerful 'cause it just keeps \[00:25:00\] the agents just so much more focused on, on your objectives. \[00:25:03\] **Ben Lloyd Pearson:** So, \[00:25:04\] **Andrew Zigler:** Yeah \[00:25:04\] **Ben Lloyd Pearson:** it's less so about how many tokens am I spending right now and how, and more so like am, am I giving, am I getting that right signal-to-noise ratio? \[00:25:11\] **Andrew Zigler:** Totally. The distillation effect, the idea that you're throwing everything in a big pot and being like, "What can I boil this down to?" It's really appealing actually. Um, the in- the commoditization of intelligence is very real. Even goes back to the second brain stuff we've been talking about a lot on the show. \[00:25:27\] **Andrew Zigler:** Distilling that information down is really just that process too. So, um, you know, maybe another time we can talk about how I've take- I've taken a crack at some of those things for sure to, to learn and, and just be more curious with the world. \[00:25:41\] **Ben Lloyd Pearson:** All right, well, this is Dev Interrupted, brought to you by LinearB. Thank you to our listeners for giving us your time today and for sticking around all the way till the end, actually. Uh, we really appreciate the attention and the engagement that we get out on social media. And speaking of engagement, you know, we love to get to interact with our audience, so whether that's just giving us a thumbs \[00:26:00\] up or a, a like or a rating wherever you're consuming our media. Um, and then reach out to us, comment on our Substack, reach out to us on LinkedIn. We love to interact with, with everyone that's out there. So thanks again for joining us this week, and thank you, Andrew, and we'll see you next week. \[00:26:15\] **Andrew Zigler:** See you next time ## Real conversations with top engineering leaders Find us on [](https://www.linkedin.com/showcase/dev-interrupted/) [](https://devinterrupted.substack.com/) ## Your next listen [![Cover image for Why the traditional pull request has a target on its back | CircleCI’s Rob Zuber](https://assets.linearb.io/image/upload/c_limit,w_2560/f_auto/q_auto/v1/Blog_Comprehensive_DORA_Guide_2400x1256_70_8130b5ade1?_a=BAVMn6ID0)](https://linearb.io/dev-interrupted/podcast/circleci-rob-zuber-pull-request-ai-agent-cicd) Dev Interrupted [Why the traditional pull request has a target on its back | CircleCI’s Rob Zuber](https://linearb.io/dev-interrupted/podcast/circleci-rob-zuber-pull-request-ai-agent-cicd) CircleCI CTO Rob Zuber joins the show to discuss why the traditional pull request is struggling under the sheer volume of autonomous AI agents. As AI adoption... [![Cover image for Models escaping containment, intelligence becoming a commodity, and AI code review to the rescue](https://assets.linearb.io/image/upload/c_limit,w_2560/f_auto/q_auto/v1/ai_models_containment_code_review_commoditization_7ea901a97a?_a=BAVMn6ID0)](https://linearb.io/dev-interrupted/podcast/ai-agent-sandbox-escape-open-source-models-code-review-bottlenecks) Dev Interrupted [Models escaping containment, intelligence becoming a commodity, and AI code review to the rescue](https://linearb.io/dev-interrupted/podcast/ai-agent-sandbox-escape-open-source-models-code-review-bottlenecks) On this episode of the Friday Deploy, Ben and Andrew break down the Hugging Face security breach caused by an OpenAI model, the rapid rise of open-source... [![Cover image for The most underrated dataset in enterprise AI is your org chart | Rippling's Albert Strasheim](https://assets.linearb.io/image/upload/c_limit,w_2560/f_auto/q_auto/v1/Blog_Comprehensive_DORA_Guide_2400x1256_68_43be6688ce?_a=BAVMn6ID0)](https://linearb.io/dev-interrupted/podcast/rippling-albert-strasheim-employee-graph-enterprise-ai-agents) Dev Interrupted [The most underrated dataset in enterprise AI is your org chart | Rippling's Albert Strasheim](https://linearb.io/dev-interrupted/podcast/rippling-albert-strasheim-employee-graph-enterprise-ai-agents) AI agents can't transform an org they can't see. Albert Strasheim, CTO at Rippling, joins Andrew Zigler to explain why agentic transformation starts with the... ## Structured data _Machine-readable metadata (JSON-LD) embedded in the page for search/AI context — not content rendered on the page itself._ ```json { "@context": "https://schema.org", "@type": "Organization", "name": "LinearB", "url": "https://linearb.io/", "logo": "https://assets.linearb.io/image/upload/v1715628027/logo-mark-lg.svg", "description": "LinearB is the engineering productivity platform that helps engineering leaders prove AI is improving throughput without sacrificing delivery confidence, flow efficiency, or developer experience.", "sameAs": [ "https://www.linkedin.com/company/linearb" ], "award": [ { "@type": "Award", "name": "LinearB is a Leader in the 2026 Gartner® Magic Quadrant™ for Developer Productivity Insight Platforms", "dateAwarded": "2026", "awardedBy": { "@type": "Organization", "name": "Gartner®" } }, { "@type": "Award", "name": "Great Place to Work Certification", "dateAwarded": "2025-2027", "awardedBy": { "@type": "Organization", "name": "Great Place to Work" } }, { "@type": "Award", "name": "America's Best Startup Employers 2025", "dateAwarded": "2025", "awardedBy": { "@type": "Organization", "name": "Forbes Magazine" } } ], "hasCertification": [ { "@type": "Certification", "name": "SOC 1 Type 2" }, { "@type": "Certification", "name": "SOC 2 Type 2" }, { "@type": "Certification", "name": "GDPR Compliance certification" }, { "@type": "Certification", "name": "ISO 27001" } ] } ``` ## More on linearb.io ### Top navigation - [Book a Demo](https://linearb.io/book-a-demo) - [AI Code Reviews — Catch security risks, bugs, and spec mismatches](https://linearb.io/platform/ai-code-reviews) - [AI & Productivity Insights — See how AI tools affect cycle time and delivery speed](https://linearb.io/platform/ai-developer-productivity-insights) - [Measure AI Impact — Track AI adoption and tie it to delivery outcomes](https://linearb.io/use-case/measure-ai-impact) - [MCP Server — Chat with your data to spot patterns and boost output](https://linearb.io/platform/mcp-server) - [Resource Allocation — Cost initiatives and shape your investment strategy](https://linearb.io/platform/resource-allocation) - [Cost Capitalization — Capitalize engineering costs with audit-ready reports](https://linearb.io/platform/cost-capitalization) - [Dev Team Management — Set targets and tie throughput to business outcomes](https://linearb.io/platform/goals-and-reporting) - [DevOps Workflow Automation — Policy-based PR routing, approvals, and tests](https://linearb.io/platform/ai-workflow-governance) - [AI Powered Support — Unify AI and human code delivery in one clear view](https://linearb.io/use-case/ai-powered-support) - [Optimization — Surface friction with feedback and MCP insights](https://linearb.io/platform/developer-experience) - [Reporting — Spot what's working and what needs attention](https://linearb.io/use-case/measuring-developer-experience) - [Surveys — Turn developer feedback into actionable signals](https://linearb.io/platform/developer-surveys) - [Platform overview](https://linearb.io/platform/overview) - [Watch now](https://linearb.io/event/engineering-productivity-gap) - [Customers](https://linearb.io/customers) - [Pricing](https://linearb.io/pricing) - [Why choose LinearB — Explore your data. Measure performance. Act to improve it.](https://linearb.io/why-linearb) - [APEX framework — The operating model for AI-era engineering teams](https://linearb.io/resources/apex-framework) - [Anti-FAQ — The questions other vendors won't answer](https://linearb.io/why-linearb/anti-faq) - [Security — Enterprise-grade compliance and zero code access](https://linearb.io/security) - [Build vs. buy — The hidden cost of building it yourself](https://linearb.io/resources/build-vs-buy) - [Dev Interrupted Podcast — Conversations with engineering leaders](https://linearb.io/dev-interrupted/podcasts) - [Reports & Guides — Deep dives on productivity and delivery](https://linearb.io/resources) - [Webinars — Expert sessions on productivity and AI](https://linearb.io/resources?category=workshops) - [Metrics Benchmarks — See how your engineering org stacks up](https://linearb.io/resources/software-engineering-benchmarks-report) - [Blog — Product updates and practical insights](https://linearb.io/blog) - [Help Center — Documentation, setup, and support](https://linearb.helpdocs.io) - [API Docs](https://docs.linearb.io/api-overview) - [Status](https://www.linearbstatus.com/) - [Integrations](https://linearb.io/integrations) - [LinearB Library](https://linearb.io/library) - [Engineering metrics](https://linearb.io/library/engineering-metrics) - [Platform engineering](https://linearb.io/library/platform-engineering) - [Engineering glossary](https://linearb.io/library/engineering-glossary) - [Developer productivity](https://linearb.io/library/developer-productivity) - [AI in software development](https://linearb.io/library/ai-in-software-development) - [Engineering management](https://linearb.io/library/engineering-management) - [Developer experience](https://linearb.io/library/developer-experience) - [DevOps](https://linearb.io/library/devops) - [Engineering operations and the context layer](https://linearb.io/library/engineering-operations) - [Engineering efficiency](https://linearb.io/library/engineering-efficiency) - [Software delivery](https://linearb.io/library/software-delivery) - [Research and data](https://linearb.io/library/engineering-benchmarks-and-research) - [LinearB is a Leader in the 2026 Gartner® Magic Quadrant™ for Developer Productivity Insight Platforms](https://linearb.io/resources/gartner-magic-quadrant-dpi-platforms-2026) - [Sign in](https://app.linearb.io/login) - [Enterprise](https://linearb.io/solutions/enterprise) - [Contact](https://linearb.io/contact-us) - [About us](https://linearb.io/about-us) - [Careers](https://linearb.io/careers) - [Service agreement](https://linearb.io/services-agreement) - [Privacy policy](https://linearb.io/privacy-policy) - [DPA](https://linearb.io/data-processing-agreement) - [Security FAQ](https://linearb.io/security-faq) - [Substack](https://devinterrupted.substack.com/) ### Footer _Additional links from the site footer, not repeated from the top navigation above._ - [GitHub](https://github.com/linear-b) - [LinkedIn](https://www.linkedin.com/company/linearb) - [Twitter](https://twitter.com/LinearB_Inc)