Are we heading toward a bizarre future where your engineering salary is paid in AI compute tokens instead of cash? Andrew and Ben tackle the latest tech industry shakeups, starting with Meta's acquisition of Moltbook and the controversial idea of making inference limits a core employee benefit. They also break down Charlie Guo's harness engineering playbook, the growing pains behind recent AWS AI-driven outages, and the toxic pressure to constantly run dozens of autonomous agents. Finally, they wrap up by sharing their own agentic weekend projects and debating the catastrophic risks of vibe-coding your laptop's file permissions.
Show Notes
- Silicon Valley is buzzing about this new idea: AI compute as compensation
- The Emerging "Harness Engineering" Playbook
- Meta acquired Moltbook, the AI agent social network that went viral because of fake posts
- “A spate of outages, including incidents tied to the use of AI coding tools”, right on schedule
- Every minute you aren't running 69 agents, you are falling behind
Transcript
(Disclaimer: may contain unintentionally confusing, inaccurate and/or amusing transcription errors)
[00:00:00] Andrew Zigler: So Ben, did you hear that Meta acquired Moltbook?
[00:00:04] Ben Lloyd Pearson: Oh yeah. I mean, I, I, I heard it too many times at this point. It seems like everyone wants to talk about it. I mean, yeah. I don't know what, what's, what's Mark Zuckerberg's like obsession with, like, monetizing personal relationships with AI. Like, what's up, what's up with that? Like, it just feels like this is another step in that journey.
[00:00:22] Ben Lloyd Pearson: I don't know. What do you think?
[00:00:23] Andrew Zigler: do you mean? It's, it's the perfect thing to add to his repertoire, another website
[00:00:27] Ben Lloyd Pearson: Yeah, I mean, I get that.
[00:00:28] Andrew Zigler: Another website full of bot users pretending pretending to be people. I think that it's, uh, another option as belt for sure, but actually I think it's really interesting to think that, uh, it would be something that is, uh, hungry for an acquisition and not, nothing in my mind really struck me as
[00:00:41] Ben Lloyd Pearson: Yeah. No,
[00:00:42] Andrew Zigler: book.
[00:00:43] Ben Lloyd Pearson: yeah, I'm, I'm like, what, what insane experiment can I create that some company will come along and buy,
[00:00:50] Andrew Zigler: Well, I do think there's an interesting telling there about maybe this being like a wink to where things are going and like the new consumer economy, because obviously
[00:00:58] Ben Lloyd Pearson: Yeah.
[00:00:59] Andrew Zigler: [00:01:00] in their ad marketplace making the bet that in the future they're gonna be selling to your agents. More than they're gonna be selling to you.
[00:01:07] Andrew Zigler: So you might as well start owning the places where those agents go.
[00:01:10] Ben Lloyd Pearson: Yeah.
[00:01:11] Andrew Zigler: really interesting kind of transformation of the internet, just depending on if it's artificial readers end up outnumbering. Its real ones
[00:01:19] Ben Lloyd Pearson: Um, kinda makes me wonder, like, what happens? Can you just make agents that create enough value on their own that they become like an acquirable target? but yeah, I'm just like reflecting on like how chaotic, like this, like OpenClaw, like has made the, this, the AI space, you know, like. Its inventor went to OpenAI.
[00:01:37] Ben Lloyd Pearson: Now we have like the social network that was spawned out of it, that's gone on to meta. But then like you also just have like companies like Anthropic over here that are just like quietly building cowork and it's like using all the same conventions, you know, but that, that are letting like smart people like you and me just like solve our own problems with them, you know, rather than having to like get a product out of it or something.
[00:01:59] Ben Lloyd Pearson: But
[00:01:59] Andrew Zigler: It [00:02:00] is also
[00:02:00] Ben Lloyd Pearson: yeah.
[00:02:00] Andrew Zigler: in the organizations. Anthropic somewhere. I think they built cowork in what, like 10, 12 days or something.
[00:02:06] Andrew Zigler: Meta's Super Intelligence lab That's a place where a bunch of smart people and interesting ideas go, never to be heard from again. So far, so far it's really just the Willy Wonka factory of AI and so, uh, I, I just don't know what this means even for Moltbook, but definitely something to watch.
[00:02:22] Ben Lloyd Pearson: yeah, Well, anyways, welcome to the Friday Deploy. I'm your host, Ben Lloyd Pearson.
[00:02:27] Andrew Zigler: And I'm your host, Andrew Zigler.
[00:02:29] Ben Lloyd Pearson: Here's this week's news, getting paid in GPU time, a playbook for harness engineering, a spate of outages, including incidents tied to the use of AI tools, and are you falling behind if you're not running agents every minute of the hour. Andrew, let's just start right at the top of that and get, talk about these people who are, are, are getting paid in AI compute.
[00:02:50] Ben Lloyd Pearson: What's the story?
[00:02:52] Andrew Zigler: Yes, so OpenAI is, Greg Brockman said Inference compute is increasingly driving software productivity, but also becoming part of compensation [00:03:00] packages and one submission on the levels that FYI even lists. Did a copilot subscription as a benefit. Uh, so it's really interesting to think that AI usage could represent, uh, upwards of even 20% of total engineering compensation based on some, uh, VCs kind of estimates on this news.
[00:03:17] Ben Lloyd Pearson: Yeah. You know, I think both of us can kind of relate to this. Uh, you know, we've, we've had discussions about how like, quickly just raw token costs can become like one of our biggest budget needs. Uh, and it can just like suddenly ramp up, like overnight once we, we find a new workflow that we're going to use.
[00:03:36] Ben Lloyd Pearson: I mean, the moment that I started like doing agent orchestration, you know, I, I immediately felt this like anxiety around my token usage limits. Like, you know, I didn't really like being in this place where I had to like temper my goals and expectations around what I could think, what I thought I could fit into, like a five hour token session.
[00:03:54] Ben Lloyd Pearson: there was a number in here that I, I saw that the, uh. There was an estimate that [00:04:00] like token costs could reach a, a, a level of like 20% of, of an employee's salary. you know, I think just looking at the horizon that exists today, like if the trajectory continues, I, I could see that being a reality that like you actually would be consuming so many tokens that.
[00:04:16] Ben Lloyd Pearson: it would be a significant portion of the cost of hiring someone. So, yeah, I don't, I don't know. What do you think, Andrew? Like, especially around this, like it being a part of compensation packages?
[00:04:27] Andrew Zigler: As an engineer and someone who consumes a lot of tokens. Actually, I think that this whole idea is totally unacceptable and it
[00:04:36] Ben Lloyd Pearson: Oh.
[00:04:36] Andrew Zigler: face of what I think, uh, the idea that the inference is unlocking for companies is supposed to be doing. You know, we've talked extensively on the show about the value transfer of being a 10x a 100x engineer and using all of that time and inference costs to. To create benefits that ultimately get captured by your employer. That's the, that's the structure in which we work [00:05:00] inside. And so this is like telling somebody working in the.com boom, that they're gonna get paid in bandwidth. You know, if you want me to buy more tokens or want me to be more on the cutting edge of whatever technologies we're using, then uh, make that infinitely available to me as part of my job.
[00:05:16] Andrew Zigler: Isn't part of the narrative that these teams are supposed to be getting smaller because one person's doing five 20 persons people's work. So why
[00:05:24] Ben Lloyd Pearson: Yeah.
[00:05:25] Andrew Zigler: five or 20 persons people's budget? Um, it, it doesn't make sense to me to tie it into a compensation package. Uh, I think it's extremely accessible to get access to a near infinite amount of tokens without. Breaking the bank. And uh, frankly, a lot of engineers are already tapping into both sides of this with their work consumption and their personal consumption anyway. So the idea of trying to make it part of the compensation doesn't make sense as much as it is. Uh, you should definitely roll it into the cost of the employee. Um, but that's kind of the agentic workforce we're stepping into. Maybe that employee [00:06:00] is two or three times more expensive or something, uh, and a large part of that's going to their tokens, but you're getting five x the output, 10 x the output. And so, uh, for me, I think it's a fascinating, it made me actually think of the song, uh, um, where he is singing like, oh, you load 16 tons and what do you get in another day,
[00:06:18] Andrew Zigler: older and deeper in debt? But it's like, in this case, you just get what? A co-pilot subscription. Like, no, it's
[00:06:25] Ben Lloyd Pearson: Yeah.
[00:06:26] Andrew Zigler: to me.
[00:06:26] Ben Lloyd Pearson: Yeah, yeah. Uh, yeah, you know, I think that's a good, that's a good way to frame it. Like it really is something that, um, you know, I, I, I often compared like, uh, in the early days of copilot subscription to like a grammar checking software subscription. Like, we use Grammarly for example, and it's like we, we don't really like.
[00:06:48] Ben Lloyd Pearson: Uh, talk about whether or not it's something I should have. Like, we don't really debate it, you know, it's just, I, I, I walk in the door at any job and it's something that I'm given because we recognize that it's something that makes me more productive, uh, [00:07:00] to have access to that. So, the story really just highlights just how in demand all of these services really are becoming at this point. Uh, and not only that, but how much value we we're extracting from them. Now when we, when you figure out how to do agentic orchestration. So, uh, yeah, it's an interesting conversation nonetheless.
[00:07:19] Andrew Zigler: For sure.
[00:07:21] Ben Lloyd Pearson: All right, so moving into our next story. So this story covers the emerging harness engineering playbook. And I really like this article 'cause I think it illustrates a lot of great examples of what we're describing of these teams that need high token costs be or high token consumption because of the thing, the types of things that they're doing.
[00:07:40] Ben Lloyd Pearson: Uh, so this, this article comes from friend of show, Charlie Guo he walks through a whole bunch of situations of, of engineering teams using agentic AI to accomplish some pretty amazing things. Like at OpenAI, there's a team that builds a 1 million line products internal product with over a few months with just three engineers.
[00:07:58] Ben Lloyd Pearson: Um, you know, Stripe is [00:08:00] doing tons of, agentic PRs that are, uh, being produced every single week. So, Andrew, I'm curious what you think about, uh, this article after reading it.
[00:08:08] Andrew Zigler: Uh, this article is a really great, uh, nuanced breakdown of the challenges that are facing engineers right now, because we've talked a lot on the show and everyone's heard this saying, a lot of times of engineers have to become managers, right? And engineers
[00:08:20] Ben Lloyd Pearson: Yeah.
[00:08:21] Andrew Zigler: think in the level of how their manager would typically assign and, and, and figure out what's gonna have the highest impact of work, right?
[00:08:28] Andrew Zigler: And that kind of, uh, transformation, how you think about your work is, is one journey. But then there's another journey happening here as well is once you've decided what needs to happen and you're. Acting upon it. It's more than just telling your agents to do it, like you know exactly what to do. You also have to curate the right environment and give them the right tools and set the right expectations so that they can not only get to where they need to go, eventually, but do it quickly, safely, and on task.
[00:08:55] Andrew Zigler: Because as we've talked about, like over 70% of agentic coding time is [00:09:00] typically spent refactoring past agent's work. And
[00:09:02] Ben Lloyd Pearson: Mm-hmm.
[00:09:03] Andrew Zigler: the idea of having proper alignment. And guardrails from the very beginning is so important, and that's what this article dives into. the idea of harness engineering, being that you have to manage the work environment just as much as you're managing the work that the agents are doing
[00:09:18] Ben Lloyd Pearson: Yeah.
[00:09:19] Andrew Zigler: acknowledging that it's not a step one, step two process. They feed into each other, you switch between them and they enrich each other. And there's actually as much as there is to learn, uh, and reflect on from engineering managers and product leaders and how they think about and delegate technical work. There's also a lot to think about just in general. practices, leadership practices. If you've ever been in front of a, group and and doing a presentation, leading a team in any capacity, you know that you have to put the right tools in your people's hands in order for them to get the job done. And this is what's now happening on a micro level, not just on the engineer level, but inside all of their individual projects.
[00:09:58] Andrew Zigler: And this article dives [00:10:00] into some of the nuances of that skill of building that harness.
[00:10:04] Ben Lloyd Pearson: Yeah, and specifically some of the practices that stood out to me is. Uh, you know, first of all, like, planning is the new coding. And, you know, I've been keeping this phrase in my head, be intentional. Like, I keep thinking that at all times as I'm building stuff because I found that's, that's the way that I ensure that the agents that I'm operating are moving in the direction that I want them to move.
[00:10:25] Ben Lloyd Pearson: Uh. There's also advice on documentation being your new system of record. You know, everything needs to be documented and stored and you know, I think historical context we're learning is just as important as like knowledge context. You know, you need to build that, the, the historical reference of all of the decisions that have been made, why they were made, and how it impacts.
[00:10:48] Ben Lloyd Pearson: Things moving forward, you know, you need that chain of decisions. Uh, but most of all, there's some, you know, the advice that I think we all need to just really, really keep at the center of this is just avoid the AI slop. Like you should [00:11:00] have higher standards for AI generated code than you do for human generated code.
[00:11:05] Ben Lloyd Pearson: And I think that's one of the missing nuances behind when you hear all these, these executives out there that are like we, we have 70% of our code being generated by AI. And they don't always mention the part where they also include substantially more testing for that code than they've ever required in the past.
[00:11:22] Ben Lloyd Pearson: So it's just naturally generating far larger volumes of code than, than, than, uh, what they would otherwise. So yeah, there's a ton of great advice in this. Uh, and yeah, definitely encourage everyone listening to go read the, the article.
[00:11:37] Andrew Zigler: Yeah.
[00:11:37] Ben Lloyd Pearson: All right, well, let's, let's go, let's move on to a topic that, uh, we, we seem to be coming back to over and over here on Dev Interrupted.
[00:11:44] Ben Lloyd Pearson: And I, I kind of feel like it's gonna keep, this story's gonna keep going for a little while. Uh, this comes from Gary Marcus, about these recent Amazon outages that have happened. These types of things probably are already happening like all the time. Um, but it really only makes the news when it's a company [00:12:00] like AWS that it happens to.
[00:12:02] Ben Lloyd Pearson: So, yeah. Andrew, what's, what's your take on this, this article?
[00:12:05] Andrew Zigler: Yeah, I think that's an important distinction. The call out is that this kind of process and incident, in reality, this growing pain is happening everywhere. It's just when you're Amazon, it's inescapable, uh, you know, when it affects you and your customers. All of that said, I think that this is like a really interesting, sign of. how much of the growing pains of developing those new skills around understanding what needs to be
[00:12:31] Ben Lloyd Pearson: Exactly.
[00:12:32] Andrew Zigler: also creating the right environment for it to be done as folks build those dual skills, you kind of get this awkward, one step forward, one step back, you know, lurching kind of motion. And I think that's what you're seeing is like, from this, there's obviously going to be learnings that will produce new guardrails and. Better environments, more strict, uh, security, a deeper understanding also of how the tools that are getting, uh, put on, put on your team are being [00:13:00] utilized and adopted.
[00:13:01] Andrew Zigler: I really think that's like a calling card here is like if you have incidents like this that are either visible or you think are happening under the surface, it's really important to understand how that does tie to your organization's AI velocity and what tools you're using and having a scope. Of, you know, uh, if there's a problem that's happening in an a, in an because of like an agentic process, being able to implement guardrails to prevent it is something that's so achievable now in a way that actually was, fundamentally harder when you were trying to implement these guardrails with humans and human engineers.
[00:13:35] Andrew Zigler: And so I think there's a lot of interesting ground to explore in being able to understand like. How, uh, AI adoption is tied around these kinds of incidents and then actually be able to create the best working environment for it moving forward. It's gonna be an uncomfortable process because we're effectively reinventing the SDLC.
[00:13:55] Ben Lloyd Pearson: Yeah. Yeah. I think just last week I mentioned the, you know, the three steps forward, two steps [00:14:00] back with, with AI or just technology progression in general. and I've seen this as I've tried to adopt all the latest tooling. there's basic capabilities that you would expect to have or want to have that just aren't there yet, because the tooling is still like very immature.
[00:14:17] Ben Lloyd Pearson: You know, even down to just like giving basic permissions, like allowing AI to, to perform certain actions on certain parts of your, your code or, or your workspace, um, versus other parts of that workspace. the tooling to do that isn't, isn't always in place without, you know, again, being very intent intentional about putting it in place in, in the first place.
[00:14:39] Ben Lloyd Pearson: So, um, yeah, there's a lot of just, I feel like fundamental capabilities that, that we're all relearning at this point. and it's even more risky when you, when you're doing this in the hands of people who don't have a lot of experience if you have, you know, more junior engineers or, or, or, uh, something like that, or just less structure around your engineering process.
[00:14:59] Ben Lloyd Pearson: Maybe it's [00:15:00] just your, your engineering team as a whole is less mature. Uh, you know, that, that open. Those are the types of things that open up these, these risks.
[00:15:08] Ben Lloyd Pearson: Is that okay? At the end of our lineup here, I wanna end it on a fun one. This is a blog post from George Hotz saying, uh, you know, every minute you aren't running 69 agents, you're falling behind. But it's really a, just kidding, it's a nod. Nod to the reality of the kind of toxic race that we all feel like we're running in sometimes. The idea that if you're not running a hundred agents right now, then you're falling behind, or you're not changing the game, or you're gonna be replaced, you know, it's, it's definitely a, a a time of like very, breakneck velocity. Right? And at
[00:15:40] Ben Lloyd Pearson: Yeah.
[00:15:41] Andrew Zigler: it can feel like there's no relief from it because you have to keep
[00:15:43] Andrew Zigler: running. And I think that this is like a nod to like acknowledging that reality that we're all running this relay, figuring out where it goes. but he also dives into a little bit about, if you know, if you are running that race, if you are, uh, waking up every day and, and, and doing your best to figure out where it's all gonna go, then you're already in in such [00:16:00] a great position.
[00:16:01] Andrew Zigler: What did you think of this one, Ben?
[00:16:02] Ben Lloyd Pearson: you know, I, I,
[00:16:03] Ben Lloyd Pearson: like I said, I think it's just something that, uh, it, it is just some the type of thing that we should always be thinking about just for our own mental health, for the health of our teams, for our companies. it is very tempting, like it's intoxicating to work with AI and see how quickly it can help you move.
[00:16:19] Ben Lloyd Pearson: and it's, it's really fun and really great, but it's also very taxing. There definitely is a lot of pressure to adopt, some of which is reasonable, but I think also a lot of which is often unreasonable. So I think it's really important that.
[00:16:33] Ben Lloyd Pearson: Uh, you know, again, I'm gonna keep saying this for, for I don't know how many weeks, but for many more weeks ahead. I think, you know, be intentional. Like if you're going to, to accelerate yourself with ai, make sure there's a clear intention behind it and you know exactly what you want to achieve with that.
[00:16:50] Ben Lloyd Pearson: And. What to, what to do with the outcomes from, from what you're doing. and I think that's the best way to just, you know, adopt AI in a, healthy way that that helps [00:17:00] you move faster without like just scattering you to the wind with all of these different priorities.
[00:17:05] Andrew Zigler: And I think a strategy for, for doing that is the, and being mindful about how you, um, are building your skills and what you're working towards.
[00:17:14] Ben Lloyd Pearson: Yeah.
[00:17:14] Andrew Zigler: what you should be thinking about is re using these newfound abilities and multipliers to remove complexity. From your job and your life, not to add it,
[00:17:25] Ben Lloyd Pearson: Mm-hmm.
[00:17:26] Andrew Zigler: only going to make that breakneck pace more untenable and more unmanageable.
[00:17:30] Andrew Zigler: But the, there's actually such a bigger opportunity to use that to pair away things that before we took for granted, that before we used as crutches, that were our own human harnesses in a world before AI and actually reinvent, um, ourselves in a simpler way so that you can be more impactful in very specific ways.
[00:17:50] Andrew Zigler: I think that's kind of the call to action. If, you know, you don't need to feel like you need to have a hundred agents running at all time. It's like there's so many people who make things and they never use them, [00:18:00] never ship them, get really close to a problem you're really passionate about, and then, uh, you'll be shocked at how much leverage you can apply to change it Now.
[00:18:09] Ben Lloyd Pearson: Yeah, I think, I think if you, if you focus on leveraging human expertise, I think that's sort of the answer to this. Um, find your experts and give, help them or have them use AI to remove the things that prevent them from spending their time being an expert. You know, I think that's really a, a healthy way to approach this because it, it frees up people for the, you know, I think most people, they want to engage in higher order work.
[00:18:37] Ben Lloyd Pearson: They don't want to be toiling away at manual effort and, and the like. So I think. Naturally. if you focus on freeing up your experts to be experts, it's a really healthy way to, to leverage ai.
[00:18:50] Andrew Zigler: Well said,
[00:18:52] Ben Lloyd Pearson: Yeah. And speaking of leveraging ai, how, what are your, what are your agents up to this week, Andrew?
[00:18:57] Ben Lloyd Pearson: I,
[00:18:58] Andrew Zigler: you caught me. They're running actually right [00:19:00] now. Oh, they were running, but during our call, um,
[00:19:03] Ben Lloyd Pearson: I hope it's not, I hope it's not 70 agents, you know? 'cause that would be the,
[00:19:07] Andrew Zigler: not 70 agents. but this case, uh, it's one of those human obstacle things where they're basically up in arms and having a revolt in my little software factory until I come and click the human button, uh, it's just one of those mornings.
[00:19:21] Andrew Zigler: What about you, Ben?
[00:19:22] Ben Lloyd Pearson: Yeah, well, I, so I'm, I'm setting up a new laptop right now and, and I've really decided to be intentional again, and, and like set this, my experience up from scratch in a way that is more agent forward. Like, I want to, I want to, as I'm working, using my laptop day to day, I want everything I'm doing to be, Supporting the agentic systems that I'm building, So yeah, I'm adopting a lot of new tools, a lot of new workflows. Like I've been spending a lot of time with Claude skills and, and holy cows. It's a lot of fun. But again, getting back to these like missing fundamental components like file permissions is like quickly becoming a nightmare.
[00:19:55] Ben Lloyd Pearson: And I'm like, man, do I need to like vibe, code of file, permission system to like, [00:20:00] like manage what the stuff that I give to my ai.
[00:20:04] Andrew Zigler: Folks, he's a dangerous territory. I listening to him talk about vibe coding, a file permission system,
[00:20:10] Ben Lloyd Pearson: Yeah.
[00:20:11] Andrew Zigler: take cowork away from him.
[00:20:13] Ben Lloyd Pearson: Yeah, I know, I know. I've seen you building all these like fancy apps to, to manage the deterministic layer of, of your agents and, and I'm not quite ready to get to that for just like my personal work, but yeah. I need something, something like that, that gives me, yeah, because right now it's like all the tools just want like full access to everything or like.
[00:20:32] Ben Lloyd Pearson: They just don't get access, you know? And I, and yeah, that's, that's how you get these horrible outages with AWS or, you know, and the like, like that's how you end up with that stuff. You,
[00:20:43] Andrew Zigler: that the truth? Except in your case, an outage on your laptop just means you can't join our Zoom call. So the stakes are a little lower, but um.
[00:20:51] Ben Lloyd Pearson: I mean, I also don't want to just like delete all of our content accidentally, you know? That would be pretty horrible.
[00:20:58] Andrew Zigler: All right, folks. Well, if next week there's [00:21:00] no podcast, it's because Ben's agents deleted all of the Dev Interrupted production work that I've built for the last year. but
[00:21:07] Ben Lloyd Pearson: Yeah.
[00:21:08] Andrew Zigler: it was really great chatting with you again this week, Ben and everyone else. I hope you have a great weekend.
