What AI Looks Like Inside Rent the Runway and Stand+

Added: 2026-05-21

[00:00:00]operators.

[00:00:06]>> Welcome to episode nine of the AI operators. I am Craig. He is Ben. We are here to help you cut through the noise and apply what matters. We are here because of our friends at Rich Panel and Afterell. Ben, it has as always been a week, but this week in particular, man, felt like a lot. How how you living? How you doing?

[00:00:29]I feel like we say that every week, but that's just how it's how it is for us.

[00:00:33]Um or or for everyone following AI. Um doing well, living the dream. Can't complain. Excited to talk about everything that happened this past week, specifically the Google IO and and other interesting uh news. How about you?

[00:00:48]>> It is a I'm I'm good. It is a full-time job to stay ahead of this stuff and I am so happy that I get to work with you on keeping our listeners up to speed and you have scaled a house of brands uh to over a billion dollars in lifetime sales and you've done it as a technical operator who has rolled out AI and aggressive AI adoption including agents, swarms, training across your entire team of 150 people. And one of the things that I'm most excited about is you're really 6 months ahead of where the bulk of our kind of listeners and candidly the world is. And so you've done and learned so much that you're now able to share with our listeners on your own technical journey driving the roll out for the company. So thank you for investing the time in helping our listeners uh based on all you're doing.

[00:01:35]>> Thanks Craig. We we do our best over at Scale Media and um as always happy to host this with you. Uh uh Craig, I know you you honed your skills with AI uh in uh Crocs uh when you managed AI over there, but since then you have learned so much and now you teach so much uh working with hundreds uh of brands over at Chat Walrus. So uh this is fun, man.

[00:02:02]I'm excited for this.

[00:02:04]>> It's really fun and it is so nice to give back to this e-commerce community that we are so lucky to be a part of. I love working with you on this. And if you are an ecom operator at a company doing, let's call it 1 million to 250 million a year, this show is for you.

[00:02:18]Our job is to do two things. Most importantly, it is to get leading operators who are applying AI in their day-to-day in front of you, sharing the ways they're leaning into the tools. And the second thing that we are here to do is to help you cut through the noise and just know what matters and how it applies to your brand. Our goal is that after listening to the show and pressing pause, you will have three, four, five new ways to lean into the technology at your company immediately. And it is a privilege to get to work with you on all of this. All right, let's get into the show today. We are here to cover this week's news which includes the Google IO conference that Ben mentioned. It includes a surprising new hire at Anthropic in Andre Karpathy. It includes new ad units and rollouts at chat GPT.

[00:03:02]It includes new services and tools within Anthropics. Claude for small to medium-siz businesses. And we are going to hear from two amazing operators, close friends of mine, Zoe, who led marketing at Cumulus Coffee and is now joining Rent the Runway as a leader there. And Rob, who has scaled his business, Stan Shoes to over $10 million in sales with a team of two humans and a ton of AI. And we will close it out with Ben walking us through agent harnesses.

[00:03:28]Here we go.

[00:03:32]If you've ever rolled out an AI support tool and watch it hallucinate or go off-brand in front of a customer, you already know most of them just don't work. Ridge Panel is different. You paste your website URL. In about 20 minutes, the AI has crawled your site, learned your brand voice, connected to Shopify, and is resolving order status, returns, and subscription cancellations end to end. About 40 cents per conversation versus $2 to $10 for a human agent. And here's the kicker. 50% of your tickets resolved autonomously in 30 days or full refund. Go to richpanel.com/demo source AI ops. They'll set it up live on your actual store on the call.

[00:04:14]>> All right, Ben, some breaking news.

[00:04:16]Andre Karpathi has just joined the team at Anthropic. And my question is, does the fact that one of the co-founders of OpenAI has paused his own work to go join Anthropic tell you which Frontier Lab brands should be betting their entire tech on. Ben, what's what what happened here?

[00:04:37]>> Look, I think Anthropic definitely earned uh some street cred with the the AI crowd by bagging one of the the biggest legend in the field. So Andre Karpathi uh you know was a founding member of of Open AI and also led AI and uh Tesla and their self-driving division and now part of anthropic. Um I think yeah it's definitely a sign.

[00:05:03]>> It's really exciting. I I have learned so much from Andre Karpathy. His videos on how LLM's work on YouTube are like required viewing. And I think what's interesting is the bet that these guys are making, right? I think what Karpathy will be leading at anthropic is this idea of recursive learning and can they build some sort of future where the models continue to get smarter on their own. This guy is incredible. We talked about his recursive self-improving loop.

[00:05:31]We talked about the wiki uh that he created through his Obsidian files, the second brain, and there is no doubt that under the guidance of such an expert that Anthropic will be set up for success. So, more to come. Uh, and I'm excited to see what he builds over the next year, year and a half. All right, on to the next. All right, Ben, some more breaking news. Google's IO conference was yesterday, and they announced a lot of stuff, right? Like I'm most particularly interested in some of the brand things that they launched across Flow, Pomelli, Stitch, and it being a direct attack on a lot of brands design systems. But what did you take away from from yesterday's conference?

[00:06:08]>> I'm actually really excited about Spark, Google Spark. It's not out yet. I think they're releasing a a betaup version for uh the ultra um AI users next week, but it's essentially a 247 AI assistant, right? So, it's uh going to be similar to Cowwork. Uh it's going to be similar to OpenClaw, but it's connected to about 30 uh different SAS products. Um and it's using the entire Google ecosystem.

[00:06:38]So that's something that's super excited exciting to me because I'm using the Google ecosystem. I think a lot of people are. Uh so I'm curious to to use Spark and see how how good it is.

[00:06:49]>> Yeah. I think what I'm focused on and first of all kudos to both Ben Thompson and Cla Vo who had great analyses same day. I think to me it's the investment that they're making in video and candidly what it means for brand teams today and world models and robots that they'll be building going forward. So you can now take a product and through Google services just turn it into a video and then you can use human text to edit that video. And these aren't just like short two second clips. These are like 10 second kind of like ads effectively and some of the other investments that they're making in terms of design and branding and the ease of launching websites is really incredible.

[00:07:26]Particularly if you are a cash constrained early company, let's call it doing1 to $3 million. Like the days of raising a bunch of capital for crazy gotom market initiatives with expensive branding agencies like you can be scrappy with Google's tools now and get candidly like A+ production quality work through some of what they're building.

[00:07:44]and Domas uh you know who I believe Dom Sebis uh I'm probably messing up his name a little bit but the guy who leads Google DeepMind he has been very clear that what they're building towards is world models that in this case can help with robots and other things. So keep an eye out for what Google is doing on the video and branding side. A lot more to come.

[00:08:03]>> Yeah. So they they just released this new world model um called Omni and I think this is this is where everything is going. We just heard uh news recently that runway which you guys all remember runway from the early uh video genai uh products they just switched to basically compete with Google and all the frontier labs by building uh a world model. So this is where everything is going towards world models. Uh so it's going to be really interesting. I haven't used it yet. I'm very curious to use that.

[00:08:35]And what's the difference between a world model to to an LLM? I'm not quite sure yet, but we're going to find out very soon.

[00:08:41]>> A little breaking news. Nvidia is now worth more than every single country on Earth, except for two. They've got nearly a $6 trillion market cap, which makes them bigger than every economy in the world, minus the two powerhouses, the US and China. So more or less before we get into the e-commerce news today, I just wanted to mark this right because it's like the kind of number that resets your sense of scale for everything that we're about to talk about. So very quickly, Nvidia hit $6 trillion in market cap in the middle of May, just a couple days ago, after a 4% intraday surge, which put them past Germany's projected GP, excuse me, which put them past Germany's projected GDP of 5.5 trillion this year. So, we've touched on the fact that only the US and China are bigger. Nvidia alone is now 7% of the entire S&P 500 by index weight. And unfortunately for your boy here, he doesn't really have a lot of exposure to that particular stock. And that's a little bit of a bummer. It makes Nvidia's uh you know segment larger than the entire energy and utilities sector.

[00:09:51]So every AI workload, every hyperscaler data center, every foundation model still runs through Jensen and the team at NVIDIA's silicon. If you are curious to dive more into what Nvidia does, there is a woman named Natalie on Tik Tok who does these amazing doodle explainers. I watched the Nvidia one like three times and would suggest that you do the same. Okay, on to the next chat. GPT has gone full Google. Open AAI will automatically create ads from your product feed. So here's what happened.

[00:10:22]The team at OpenAI has effectively mirrored Google's ad infrastructure where you can simply upload your product feed up to a million SKUs and CHPT will automatically create the ads for you.

[00:10:34]Some of the early data uh sourced by Critio is pretty eye opening. three times the click-through rate, two times the conversion rate on traffic coming from ChatGpt through this new ad unit.

[00:10:45]So, Ben, how are you thinking about this as an operator uh of a multi-billion dollar brand? What's your take?

[00:10:51]>> Yeah, the this is classic bottomfunnel um ads in Google. It's part of PMAX.

[00:10:59]Used to be shopping campaigns. Uh used to be PLA or product ads. Uh these are like highly converting ads. I think it's definitely time to start testing this.

[00:11:09]Uh we're actually running a test right now at scale and and I'll uh definitely report back with uh the results.

[00:11:16]>> Yeah, it's interesting. A couple of weeks ago, we shared that OpenAI was pulling back from instant checkout, right, where they were facilitating the transactions themselves and taking a cut of the total AOV. It seems like they've just gone a little upfunnel and now they want to live in the advertising less in the checkout world. So excited to hear back from you in a few weeks. Okay, on to the next. Anthropic has gone down market with a $20 back office for small to medium-sized businesses. And our friend Mike Beckham from Simple Modern was part of the pilot group. So Claude for small business can now run your entire QuickBooks, HubSpot, Slack, and Canva. What they've done is really incredible. through co-work just packaged up a series of connectors and skills that allow small to medium-sized business teams query and action data almost immediately. Things that would take hours are now done in seconds. Ben, what did they announce and how are you thinking about it? Yeah, I think the really cool thing here is that on top of the eight native connectors uh that uh to the software most businesses uh use to run back office, they they also ship uh this product with 15 package ready to go workflows that you can just run right off the bat. So things like payroll, monthend close for financials, uh lead triage, campaign attribution, cash forecast. Before this release, if you wanted to use AI uh uh for for these type of workflows, you need to create a a custom uh uh agent or aic workflow and that's not an easy thing to do. So now you can just use this uh claude for small business product and the 15 package workflows uh to get this down.

[00:13:01]Now, now for bigger, more most more complex businesses, this might not be 100% what you need. Uh, but what you can do is you can use this and you can see where where the gap is between what you need for your business and and then you just create a custom workflow to to bridge that gap rather than start from scratch to create uh custom agents. But if you're running a sub10 $10 million brand, uh I think this uh can probably work out of the box and can save uh on some overhead immediately.

[00:13:34]>> Yeah, I would encourage everybody, you know, Min and Finn have flagged the the landing page, go watch the video. It makes it incredibly clear what this is.

[00:13:41]And candidly, it's just smart user onboarding. What Claude has done, to your point, Ben, they've taken the tools that most SMBs use. They've made it incredibly accessible as rich context to start to action. And they've given you the skills to when you query the data set to have a better output. And what they've done is they've just made it a lot easier for a lot of small companies and entrepreneurs to get ingrained into the ecosystem to have their breakthrough moment where they see how powerful it is and then start to apply it in a bunch of different ways. Nothing new here from a technological standpoint, but a lot new in terms of ease of onboarding and quality of responses for SMBs. Your ads are already running in Amazon's Rufus platform and you as an operator did not sign up for it. I candidly am not as ingrained in the Amazon ecosystem, but lucky for us, we've got an expert who is. So, I'm going to step out of the way and Ben, tell us what has happened.

[00:14:31]Look, this is basically a new ad unit that is layered into existing sponsored products and and sponsored brand campaigns within Amazon, right? Amazon pitches it uh as uh a 60% conversion increase uh statistic. Uh but that's a Roffus overall number, not specifically for this ad unit uh uh that is prompt specific within Rufus. So, the the long story short here is that if you're an operator spending money on uh sponsored products, sponsored brands within Amazon, uh you probably had no idea that you're currently spending money on this new ad unit. So, uh pull your prompt report within the the Amazon uh business center and see what your your uh CPC and ROI looks like since March 25th when they added this. and uh then you can decide how to optimize and move forward.

[00:15:32]>> All right, I'm passionate about this one. So, Thrive Capital, who backed OpenAI, just bought the Shopify dip.

[00:15:38]They crossed into public markets and bet $und00 million that Shopify is the AI native commerce operating system. All right. So, what I'm increasingly interested is the synergy and the relationships between these two companies, Shopify and OpenAI, and how operators need to be thinking about it.

[00:15:55]Right. You've got leaders from Shopify moving to become the president of OpenAI. These two companies are clearly working very closely together. And what I think about this is it's for the consumer, right? So everything for you as an operator that is consumerf facing is likely going to run through uh you know chatbt whether that's ads and media potentially creative even elements of changing your Shopify storefront building agent shopping etc. These two companies are clearly integrated. So you need to be at the forefront there. And then what I would say internally, if it's not customerf facing, if it's more for your internal business operations, that's where you want to be leaning into Claude potentially to turbocharge the speed and quality of your work. So, I think more than anything, I'm still going to put the $100 that I have to my name on this poly market bet that at one point OpenAI will acquire Shopify. But ultimately, what you need to be thinking about here is taking advantage of the relationship between the two brands that are increasingly becoming closer when it comes to the consumer shopping experience. Make sure that you're leaning into all of the rails and Aentic builds that Shopify has made available directly through ChatBT and other tools.

[00:17:02]All right, on to the next.

[00:17:05]I am so excited to be joined by my friend Zoe who I've been working with over the past year. Zoe, tell us a little bit about yourself.

[00:17:14]>> Hi, I'm so so happy to be here. Um, I am a marketer, really a brand and community marketer who has been thrown into marketing um at consumer businesses big and small and forced to figure that out.

[00:17:27]So, I'm heading to Red the Runway to start a new role as director of integrated marketing. Um, and will really be bringing all of my marketing skills that I've, you know, gathered over the past many years working in the marketing industry across social, influencer, paid ads, really, you name it. Um, and recently was at Camille's Coffee where I connected with Craig where I was overseeing a very scrappy marketing team um, as VP of marketing and really had to figure out how to lead a team and and build an org um, using AI rather than headcount.

[00:18:06]>> I um I know that you were both thrown into the marketing world but also thrown into the AI world, you know, about six months ago. So, why don't you walk us through a couple of the ways that you've been leaning into AI at at Cumulus?

[00:18:19]>> Yeah, absolutely. So, as VP of marketing there, we had a really scrappy team. We had no creative resources. Um, and then we had two marketers who worked with me, one who was more brand focused and one who was more digital focused. So really what I did as VP of marketing was set out the overarching strategy. Um, build our budgets, build our plans, and I did all of this using Claude as my thought partner. So I'd say first I had to figure out, okay, how do I plan an entire DTOC business um, in a spreadsheet? I've never been the person to do that before. I am a brand marketer through and through. I always had a growth professional or a commercial professional to help me with it. So, first had to build a project that was basically my CRO or my thought partner in how do we plan for this? Um, so built that project. No skills involved yet.

[00:19:14]This was like very early in me figuring this out, built a project and really just through chat and Claude was able to give seasonal context and campaigns that we were planning to help really take the top down plan that our head of finance gave us and help build a bottoms up channel plan that was like, okay, here's how many machines we're going to sell each day uh based on the seasonality, based on the calendars, um, and what we want to achieve and then map budget to that. So like it all started with my spreadsheet that became my bible for the year and how we would plan on marketing.

[00:19:49]And then from there I'd say the more I learned and got comfortable, I started building out what would be my dream marketing team. So like if I'm CMO of this omni channel business, who are you know my key resources and my key team players and I actually went out and found job descriptions from similar companies and use those to build skills.

[00:20:08]So like who is my creative director?

[00:20:10]what do they need to do? Who's my creative strategist? What are they really good at? Um, let's get a copywriter in here, let's get a growth marketer in here, etc. And then ultimately that led to my like integrated marketing workflow where I would come in and say, "Okay, I need to write the brief for my evergreen March marketing plans. Like, here are the commercial goals. Here's what's happening seasonally. Here are the new product launches. Like, you are a best-in-class CMO. Let's write the brief root that to creative let's come up with three ideas bring that back into my CMO project workshop that okay now we need to bring this to the channels and basically like would just workshop these ideas and these briefs with multiple different projects who were essentially my team until I got the brief and the idea into a sharable place that we could go out and brief you know the team or another AI to to come up with the creative and and really roll out the campaigns.

[00:21:08]So, I love it. You get a tops down goal from your CFO, from your head of finance about what you need to go execute against. You're scrappy. You create a number of different projects tied to key initiatives, go to market, and then you've got effectively a team of specialists that you've created through skills. So, you're kind of like leading an AI team. Ben, are you doing anything like this at scale? Like, have you heard of anything like that? That's kind of cool. Honestly, I'm I'm listening to you, Zoe, and I'm I'm quite jealous because uh it's much harder having a bigger operation uh with more employees when you already have an existing marketing team uh to kind of uh bake in AI into these processes. But building a new marketing team in this day and age, uh used a couple of words there, uh uh a scrappy team, right? and then building it uh from scratch using AI as a thought partner, which I think is the right term, right? A lot of people call AI an assistant. Uh I I I prefer the term thought partner, >> but um building a team from the ground up using AI is something that uh sounds really appealing to me right now. And there's really many ways to skin this cat, right? So, um, your approach, um, really innovative and fun and, uh, yeah, this is really cool. Kudos.

[00:22:35]>> So, Zoe, you know, tell us a little bit about these skills. Take us in in a bit more detail about the way you're using each specialist, how you're leaning into them, etc. >> Yeah. So, I always start with my CMO chat project, um, which is really a version of a brand, which is something I learned from Craig. So, this project and skill fully trained on my brand. Um, it has our mission, our vision, our values.

[00:23:02]It has every review that we've ever received. It has Reddit commentary. It has Tik Tok comments. So, it really is seeing like how is the customer interacting with this product and this brand. And then it also has a ton of examples of like what is a best-in-class creative brief. Um, I'd say as a marketer, so much of what I do and what I've learned has always been how do you write the best brief to inspire your creative director.

[00:23:28]Um, and I've also I also feel like briefing AI is very similar to how I've been taught to brief a creative. You need to be super specific and give context. So, this first everything starts with my CMO project, um, which is really like almost my boss, I'd say. um they rate me and give feedback and make sure that you know I'm speaking to the right audience with the right language.

[00:23:51]So I always start there. Um so I'll give that the context of okay here are our acquisition targets for the month. Here are our retention targets. Um maybe you know we're launching this new product.

[00:24:04]We want to go viral with three videos on social. What are our overarching goals?

[00:24:09]Um, and then based on seasonality or what's trending in culture, we'll come up with like what is a really good insight that we can use to ground all the messaging this month. Um, and then I basically take that output, which I'd say is a creative brief, and I route that to my creative director. Then I ask the creative director usually to come up with like three big ideas. So these could be, you know, headlines or um a direction for an influencer campaign or a big event that we should do. But the creative director then comes to me with three big ideas um and like how could these live across different channels.

[00:24:45]I'll usually go back and forth on one of those until I land on one idea that I really like. Um then I bring that back to my CMO and I'm like, okay, how do you feel about this? We might workshop it again there. Um and then we land on the idea. And by the way, this is like exactly how life used to work when I was director of integrated marketing, reporting into a CMO, managing a creative director. This was exactly what my workflow was. So then basically I take that idea that we've come up with and I'll brief it into my like channel thought partner. So I say now this is like a real marketing expert who you know in your marketing or maybe this is your VP of marketing, but they oversee your director of performance, your director of growth marketing, your director of retention marketing, um, and your director of brand marketing, let's say. So I'll bring it to them and we come up with what are channel thought starters here. So we know we want to have a blog post that says this and an email and an SMS that links to the blog post. Another email and SMS that you know is more revenue focused. Um we think this could be a really good ad to try. Maybe this is the month to go into podcast etc. And then I have these like channel plans. Um, and then I usually take that and I'll bring it to my copywriting agent who can come up with like what is the launch email, what is the launch social post, etc. So we have like you know clear examples and then also a messaging hierarchy. So like the key message we need to hit this month in the brand tone is you know with cumulus it would be like cold coffee season is here super clear example and then what are the three reasons to believe that ladder up to that that every single channel needs to hit on this month. So then I would have actual copy I could use. Um and then I would take this whole package and I would send this off to our email marketing agency who we partner with, our PR agency um and our in-house marketing team so that they could go off and with their very limited resources of you know Canva as our creative tool or um our email agency who was building everything for us. these desperate real people could use this work that I had created with my AI marketing team and come back and make it look like we had this really cohesive marketing team and honestly make it look like we were punching above our weight class in terms of how consistent the messaging could be.

[00:27:03]>> I I just want to flag Zoe as you bring in and talk about the creative director and CMO and integrated marketing partner the first real person that you introduced was the email agency. you've created these kind of AI skills that that take that role, which is incredible. I'm curious. You touched on the importance of context. You mentioned Reddit forums. You mentioned customer reviews, job descriptions. As you were pumping in the skills and projects with context, were there anything that you found that gave you the best output over time? How would you guide others to think about context and what to feed the the models?

[00:27:36]>> I think giving best-in-class marketing examples is really helpful. So even when I was building this first CMO for my brand, I went through all of our emails like which emails had the highest revenue tied to it and then gave gave the project those emails so they could see the actual language and like the emotional need the emotional feeling that is resonating with our customers to drive that revenue. So um top performing content is great, top performing social posts and ads also really great. Um and then I'll give some competitive examples. So, here is another brand that's not even in our niche. So, I was at a coffee company. I often um would give different kitchen wear um items, like high-end kitchen wear, and I'd say, "Here's an asset that I love, but here's what I love about it." And this is to my point about how I have learned to brief AI based on how I brief creative. You can't just give something and say, "This is great." You really have to explain what is great about it. Is it the visual composition? Is it the way the image is sitting in your ad? is it the hook? Um, but really being clear when you're building that context of of what matters the most. But I'd say like as much context as possible in that first brand brain. Um, and then I even take that I mean that project probably has the most context, but then I'll always ask for like a onepage output that I can give to all the other projects because it feels a little overwhelming to do that like 10 times. Um, yeah, but can I turn that into like a one-page cheat sheet that everyone on the team needs to know? And I think that probably is like your real team, too, because your CMO knows the most, your creative director knows the most, but like your social media manager still needs to have access to personas and understand like what the brand is really about to do their job.

[00:29:25]>> Zoe, I I one more question before I I ask Ben a question building on this. Are you doing all of this in chat or have you gone into co-work? Are you using cloud code? Is this all in chat and project right now with connectors?

[00:29:37]>> That's a great question. It is mostly chat and project. I do a fair amount in co-work. I do not touch code. Um I need to that's like my next project I need to figure out. Um, I have a number of scheduled tasks in co-work that is really that helps me just like scrape competitive resources and competitive references and also like a lot of the reviews and that's how I actually like built my financial model was in co-work but mostly I'm doing this in chat.

[00:30:07]I am both so proud as always and also so excited for you because you are really on the cusp of some like very easy small breakthroughs in co-work and code that's going to I wouldn't say automate but just make a lot of this process so much more seamless. It's it's you're really kind of like about to have a massive um breakthrough. I What else? I'm I'm really blown away in in so many ways.

[00:30:29]One question for you, Zoe. When you're creating these briefing skills of how to brief a new, you know, brief into your creative partner, are you working with them to create and define what those best-in-class outputs look like? Are you guys collaborating on this or are you doing it independently?

[00:30:47]>> I think in the beginning, actually, I know in the beginning, I collaborated on um with my CMO on like what is a best-in-class insight, for example. Um, but ultimately this is what I've been doing for 10 years. So I feel like I really know what I want the skill to look like, but I will actually work with Claude in a fresh chat window to say, "Help me build a skill to accomplish X and then have Claude ask me questions until we build that together." So I'd say that's where I'm getting the feedback from it in terms of how to actually build these.

[00:31:20]>> I want to close with with two questions.

[00:31:22]one is I know we spent a lot of time together on chat and then moved over to to Claude chat GPT and then to Claude any guidance on your experience between the two and the process of going from one platform to the other if if we've got listeners who are really active on chat GPT but they hear everything you're doing in Claude and want to check it out what was your experience there like >> I think I was tinkering so much in chat GPT initially that I probably didn't set up as great of systems but some of the custom GPTs I'd set up. You could just export very easily to Claude. Claude has something that will give you a prompt that you can give to Chad GPT to, you know, extract data. But I would just talk to chat GPT like and tell it what you actually need to do. I want to recreate this brand. Can you please, you know, create a skill that is optimized for claude? And it will actually create that for you. Um, and then I'd also say these tools are just so smart. So, anytime I don't know what to do, I just ask it what the best next step is. Um, and it will help me learn it. So, I do that with like any marketing skill I haven't figured out, I just ask it to teach me. And I do the same and how to how to use it and how to like set up my automated tasks in co-work. Claude literally walked me through how to do that.

[00:32:36]>> What are some of the automated tasks you have running in co-work?

[00:32:40]>> Um, I have a daily briefing that I receive. So Claude's just going through my inbox and my Slack messages and anything I'm tagged in to give me like here's your schedule for the day. Here's what's outstanding in your inbox. Um I'm a huge inbox zero person. So I have like a very detailed way that I organize my inbox, but it also means I sometimes miss emails. So now I get like priority zero, priority one, priority two, and I've kind of told it who belongs in each category. And then I also have a weekly competitive review. So, I have like a separate document that has Instagram like links to different Instagram and Tik Tok accounts and I say, you know, review all of these and then put together a weekly email that tells me like what's trending in coffee culture or what's trending in millennial kitchen culture. Um, and it will actually put together the email and it drafts it for me. It sits in my Gmail and on Friday mornings I review it, make a couple of edits if I want and then I fire it off to the whole team so that everyone is like a breast on these trends and they think that I'm, you know, collecting all this data each week when really it's just an automated task that I've set up.

[00:33:46]>> Um, Ben, I mean, we're hearing from Zoe, one of the things that I love, I have a similar automated process for competitive intel and then I've got a skill that analyzes that launch in my point of view. Um, you know, we've got a a a an amazing marketer who's clearly leaning in. You run a house of brands.

[00:34:03]You lead teams of 150 plus. How are you reacting to what Zoe's sharing here, both from a structural standpoint, but also from like a cultural standpoint in terms of how you want your employees to behave?

[00:34:16]>> Yeah, I I think a lot of what we're hearing from you, Zoe, is is how modern marketing teams should be run, right? Um really what you're describing is you're orchestrating agents, right? Um and you're adding context. You're adding examples of how you'd like outputs or structured outputs to look like. And you're adding taste, right? Which is no, I like it more like this. This is the this is the voice of my brand. This is what a 10 out of 10 uh output or or example uh would be. Um and and and that's the right way of of working with AI. I think what Craig was referring to when he was talking about, you know, you're on a cusp of a breakthrough is automating that orchestration part, right? So, we're later in the show, we're going to talk about agent harness, uh, and agent harnesses and what that means, right? Co-work is an example of an agent harness. Cloud code is an example of an agent harness uh, versus, you know, the the project you're using within cloud or custom GPTs, those are just agents. So an agent harness would basically automate uh that orchestration part uh with a set of instructions and and persistent memory and and other features. So, you know, um you see Craig shaking his head because he had that realization not too long ago and he's like hooked on on creating these uh uh agentic workflows, I like to call it, which is basically you're just tying a bunch of agents together and then even that orchestration of like moving one output from one agent to an input to another agent uh is is going to be automated. So, um, I'm excited for you, uh, Zoe, and and this is really cool the the journey you're on.

[00:36:06]>> I'll I'll close Zoe with one thing and and Ben's right. I mean, you're like 30 minutes away from just setting up a few instructions in a co-work folder >> and then having it run all day. It it's it's really it's something. We'll spend some time on it together. But um my last question for you is is Zoe when we spoke recently you said you know you came to an event and and you said you were both sort of like scared but empowered by what these things could do for you as a marketer. Can you just reflect on that um on that comment and then we will ask where folks can hear from you. Tell me tell me what you meant by scared and empowered as a marketer.

[00:36:41]>> Yeah well I think we're talking about it right now. Right. I just built this entire marketing team. Um, and I only have two direct reports and but I have 10 agents working for me. And so like that's really scary to me as someone who is, you know, 12 years into my marketing career and wants to keep growing in my leadership and my seniority. I think about if this was 5 years ago, where would I have fit in? Like would AI how quickly can AI replace me? And as I continue to grow and think about the longevity of my career, um, I feel empowered in that, I recognize I need to build these skills. I recognize that marketing looks different and might be, you know, one of the first industries that's really going to be impacted by this changing AI landscape. Um, so it's a little scary to sit there and just think about, you know, the future. What will the next 5 10 years of my career look like? But I also am someone who loves learning and I love a new challenge. And so I feel like I'm an earlyish adopter of these um of these tools and I want to use them to get ahead um as much as I as I possibly can.

[00:37:50]>> You are curious, you are eager, you are pattern matching. You are the perfect person who is going to thrive uh with this technology. I'm so excited to continue to learn from you and work with you. For our listeners, where and how can they find more about you? Cumulus runs runway, etc. >> Um, LinkedIn is probably the the best place. Um, you can look me up on LinkedIn and Cumulus is an amazing cold brew coffee company. If anyone is a cold coffee drinker who isn't, I highly recommend it. Um, and I'm so excited to head back to Rent the Runway. I worked there six years ago and I'm really looking forward to helping them in this new phase of growth. So, check them out, too.

[00:38:29]>> Thank you so much, Zoe. We'll see you soon.

[00:38:31]>> Thank you.

[00:38:34]Thank you to our friends at Afterell.

[00:38:36]Brands are drowning in AI subscriptions right now. AI copy tools, AI agents, AI CRO, AI support. Most just add another expense. But Rocked Afterell is different. It has already paid out over $1.5 billion with a B to brands like Sephora, Macy's, and Jones Road. All by monetizing a page that they already own, the post purchase confirmation page.

[00:38:58]Using real-time data, Rock Brain matches shoppers with relevant offers after checkout, generating additional revenue without disrupting the customer experience. Afterell also helps brands increase revenue across the full post-purchase journey with checkout optimization, upsells, downells, and thank you page monetization. And for many brands, the revenue generated through Rock Thanks matches their afterell subscription dollar for dollar, helping them earn and save at the same time. See how brands are turning their confirmation page into a new revenue stream at afterell.com/operators.

[00:39:34]Welcome to the show, Rob Greg. Rob, founder and CEO of Stand Plus. Rob, welcome to the show. Tell us a little bit about yourself and your company.

[00:39:44]>> Well, first and foremost, I'm uh forever a fan of Craig over here and what you guys are building and so hopeful to be here. But we make the slowest shoe in the world. the first shoe that's built for health. What we realize is that shoes are either designed to look great and not feel great or they're designed for performance, but performance is usually on a track or field. And we're doing the opposite. We're doing it for all the population stands for a living.

[00:40:09]Nurses, bartenders, chefs, recovering athletes to act more like an energy dissipation mat for your feet.

[00:40:16]>> What a great mission. Uh thanks for uh coming on the show, Rob. So, uh, first question, let's let's dive into your AI stack. What does that look like? Which tools you're leaning into? What are some of your favorites right now?

[00:40:31]>> You know, if you asked me a month ago, it'd be different. And if you asked me two weeks from now, it's probably going to evolve. I think what I've been trying to do is make sure I don't get into scope creep of getting too many things at once, but I find my foundations. And so it involved uh starting in chat GPT I moved into Claude although I feel like lately Claude's getting a little wishy-washy probably because of the popularity. Um but I try to put controls in place. So I start instead of a chat I evolve it into a project and in that project I've got a set of instructions uh and markdown files. So I set guard rails of what is this conversation going to be about and how are we going to engage. I was doing a lot of projects and I found a limitation was that the projects weren't communicating between each other. So, uh I also found that there's a lot of repeated tasks and I wanted to get similar results. So, for repeated tasks, I use a cloud project to create a skill. So, we do custom shoes for teams and I need to get a logo, turn it into a TPU patch, send it as a mockup and out to the team. Well, rather than go back into the projects, this is what I'm looking to do. I have a skill that I just say create these and I've got an API integrated into Nano Banana and Higsfield. So skills are wonderful for always a repeatable task. Where I've leveled up over the past four or five months is with cloud code. And cloud code is I'm thinking about how do I do things that are specific to my business that save me time. So, there's a lot of buzz around AI for creation of new things, and I'm mostly using it to optimize what exists. We're a small team. We're an 8 figure break business with one employee, uh, a handful of contractors as well, but as you can imagine, there's a lot that just eats up the day, and they're really important tasks. And the way I'm using AI is to figure out what are the things that take the most of my time. They're important for the business, but I can put systems in place and in lie of hiring an individual that gets tired or there's human bias that's involved. I'll give you some specific examples as well.

[00:42:38]We'll dive into it, but I'm using it to get time back in my day and to be able to find the things that human bias misses from inventory management to how do I actually organize all the inbound messages and DMs and you tell me uh where we can be most valuable for the audience and I'll go an inch wide and a mile deep and I'll go through exactly how I built it. And I think the cool thing is I am not a technical founder by trade. This is a brand new thing. I don't think anyone's truly an expert. I think there are people who are at the forefront of fascination and knowledge gathering uh that can help disseminate that. Craig being one of those folks.

[00:43:18]Non-paid plug here. Uh just a huge fan of the community that's being built. I really can't say enough about it. Um and chat walls. But um I think it's just having the the courage to start more than anything is the biggest barrier.

[00:43:32]The technical barrier has never been lower. the ability to go in and code or vibe coding as people call it. It's never been easier. I can talk about some things I did wrong and some things that I learned along the way to do it right.

[00:43:44]But for anybody that's listening to this, today is the day. It's only going to get more crowded, more expensive, and that gap to the start line is just going to keep getting further and further away. So, if there's any one takeaway, just today is the day. It's okay to suck at it. You're going to to start. You'll get real good real quick.

[00:44:07]Rob, uh, super inspiring. I I mean, you're living the dream. This is what this is all about, right? AI is here to help us, uh, uh, rid us of the mundane work. Uh, allow us to to have more time to do more interesting things, answer harder uh, questions, solve harder problems, be more creative, be more strategic. So, you're doing it right.

[00:44:29]Um, Craig, I I I see you uh uh looking like a proud coach over there. What What What are you feeling right now? What are your thoughts?

[00:44:38]>> I'm feeling gratitude for Rob and his friendship. And I remember a year ago kind of spending some time with him.

[00:44:43]That's what kickstarted my company and to see how far he's come. Candidly, I now need to learn from from Rob. I mean, Rob, you you built your entire website with AI, right? Didn't you like redesign the entire site around tell us a little bit about that?

[00:44:57]Yeah. So, uh, I wanted to redesign the website, but I wanted to go a step further and rather than hire an agency with best practices. I'm not really interested in best practices. And I think that that's something where AI is really good at is best practices, but I want to do something different because our consumers deserve something different. And what we're creating really hasn't been built in this exact way, shape, or form. So, how to start?

[00:45:21]It's going to sound crazy, but I thought of it like managing a hedge fund. So hedge funds are huge on data. They're trying to predict markets and behaviors.

[00:45:29]And I thought, why not apply that to a website? I'm not designing this for me.

[00:45:34]I'm designing this for our customers. So where I started was a project and it was a research tool. Go far and wide. Find out what our con consumers are saying about us. Go on social media. What's in the trade news? What's in the press? Is it a bull market or bare market? Are people looking for value right now or are they looking to cost save? So I wanted to get a snapshot of our demographic and people who have not been customers so far. What's their mentality like? How are they buying? What are they looking for? So starting with that as the base, it helped uncover things about the site that I would have otherwise missed because previous I was going let's do a new banner image, let's organize where these flow, but I was missing really key pieces of information that AI was help able to help me uncover. An example on that energy dispersion. visualize that. People don't know what that is. And so it said, "We're gonna create a visual. I've got an API integration. We're gonna create this in Higsfield and Nano Banana." I blew me away and I was actually pretty mad at myself because I think I'm a creative guy. And I I said I don't really use that much for creative, but man, it it humbled me very quickly. Is a phenomenal insight to be able to put in.

[00:46:44]So I was using consumer sentiment to drive what should actually be on the content. And the rest of it was pretty easy. We've got a repository of assets.

[00:46:53]I do have some API integrations for image gen. But what it was able to do first it gave me a wireframe and I said okay this is great but now I need to convert this into J or into liquid for Shopify said convert it into liquid. So I put it into liquid. I jumped into the cloud code and they said I need you to set up a new theme. Paste this liquid in there and I need you to do QA on it because I'm not going to push this live without making sure nothing breaks. So between Claude and Claude Code, Claude was my coach. So it gave me the ideas, it converted into liquid. I went into Claude Code to do a session with my Shopify store to put it in the format instead of live. And I went between the two. The executor being claist being cla can view what's going live on my stream.

[00:47:40]And by the way, these are tools that uh there's um I use railway versel superbase things terms that were brand new to me. Uh I'm not an expert in it. I didn't go read novels around it. I didn't go to school for this. I I just asked what is this? What's the value? Do I need this in my tech stack? Is this going to eat up time or save me time?

[00:48:02]And the filter I have it's got to save me time and be able to scale. And so I pushed it live and I Q8 it since then.

[00:48:09]We've improved our conversion rate by full percentage point for ecom that's almost unheard of. And this was without writing any code. This was that without doing any new photo shoots, without having to really think behind it. It just allowed AI to really get in the mindset of my consumer base and present something that answers the questions that they're looking for. Just incredibly powerful tool.

[00:48:35]Rob, this is so awesome to hear. Um, Craig often calls me an engineer and says I'm very technical and he's not technical. I'm like, you you don't need to be technical, right? You just need exactly what you're describing, Rob. You just need the curiosity and just to learn and uh you know, Claude would tell you the next step. It's it's it's that easy. You just have to be persistent um and and learn these things. So, uh kudos to you, Rob. This sounds uh really uh inspiring. Um, let's move on to uh um a more specific use case for AI uh that you unlock for your business that that our listeners can um can maybe use in their business.

[00:49:16]>> Yeah, it started with what eats up the most of my time at the end of the day and what is the thing that I save for the end of the day, meaning I'm avoiding it the most. And that was responding to social messages and DMs. We don't have a social team. I manage that. Uh, and I was not doing that well because it comes 1:00 am, 2 am, 3:00 am, and there I am with a whole stack of messages. I'll save what it looked like before, but it was close to a 12step process because I wanted to organize it. And just the the process of managing it all took up most of the time, not actually communicating directly. So, what I did uh is I jumped into cloud code and I said, I need a cleaner way to take all these inputs, help me come up with find the inefficiencies. So, it said, "Well, you're tracking everything here. You have to log in to Instagram here. You've got to open that page here. You've got to categorize it here. Send emails here.

[00:50:10]And there's a whole web of things." I said, "I just want it in one clean place. It's got to save me time. I don't want you to answer the DMs. I want you to save me time." So, what that looks like now, uh, I've got on the left is a stack of all my unread DMs. In the middle, I've got an image as well as the top nine. And I'm doing this in a way that doesn't violate terms and conditions. It's not actually integrated into Instagram. It takes the public facing information which is free and clear to do because it would say we can do this. There's gray area. It said don't do gray area.

[00:50:43]>> Let's do this by the book.

[00:50:45]>> And so if you push it, it will find creative solutions. And then on the right it categorizes it. It says who are they? Why might they be interested? I also took a year's worth of my DMs and I uploaded it in a secure server and said, "Understand my voice.

[00:51:00]>> I don't want you to write the messages for me, but I need you to understand when do we engage, when do we do partnerships, when do we offer discount codes, when's it customer support?" And so feeding that data in, it was able to categorize who's reaching out and what might my most likely response be and then categorize it. So now what my flow looks like, I see the DMs on the left. I click one. I see the snapshot. I see the ranking. Should we engage with them and why and what category? I see a copy.

[00:51:31]Here's some thought starters. I can click the button to open the screen. I hit send. And when it's done, I hit mark sent. It moves into the next bucket. And then when it's all said and done, anytime they post content, I click on a new tab. All the content lives there.

[00:51:45]everyone that's created content, tagged us in content, I can now share this externally with any of our marketing folks and they go rather than say, "Hey, is there any new year to generate content? What's the chat history?" It all lives there. So, it saved me about two hours a night in my day, which as a new dad with a 18-month-old is huge.

[00:52:06]>> Okay. as a founder that wants to try to get a little bit extra sleep. I used to love the grind culture and I was up till 4 a.m. Now I'm going, man, if you went to sleep at 9:00, that is the dream.

[00:52:16]>> I'm I'm working towards getting there.

[00:52:18]Um, but really saving time in my night and to start too, it got really overwhelming if I'm being honest. It took me about 13 hours to build this.

[00:52:27]Not because it takes 13 hours to build, but it's because I'd never done this before. So, at the end of the session, what I did, I said, "This took a really long time." why what I do wrong where can we create efficiencies how if I'm going to do this on a repeated basis can I do this better so I asked it to debrief our session and it came up with some really interesting tips and tricks to streamline and then eventually I wrote a playbook a markdown file and they said before we start any new claude session reference this markdown file it's goes through this checklist it's a great book called the checklist manifesto 99% of things can be avoided if you just do these three simple rules.

[00:53:06]So, I was able to create rules and now my build time takes about 4 to 6 hours for a massive build that usually would have taken four to 6 months with a massive team.

[00:53:19]>> That's so awesome, Rob. Really, really exciting to hear um everything uh you're doing with AI. I think you know people talk about the time savings, right? you get two hours more to spend with your family, but really it's more than that because now you're going to be more fresh when you wake up the next day to go to work uh and you'll perform better uh as a founder. Um that's uh that's great. Craig, any thoughts about that?

[00:53:45]You're working with with hundreds of brands doing similar things. Have you heard of a similar use case to what Rob is doing?

[00:53:53]I think what I'm hearing from Rob and I'm starting to hear from others and candidly we heard from you know Matt Aila and bar at Boous is this idea of kind of building bespoke SAS on your own unique to your company and your needs rather than relying on outside vendors necessarily Rob for you as a you're not a soloreneur you've got a you know a partner in Graham and you know you you've scaled the business to eight figures how are you thinking about SAS spend and how have things maybe evolved uh you given the skill set that you've uh yeah that you've got over the last year.

[00:54:26]>> Yeah, that's a great question. I think SAS is no longer something I'm interested. Data is and I'm always interested in where I can get the richest data from. If there's a solution that has really rich data that I can tap into to leverage insights, that's going to be interesting to me. But what I'm finding is where the richest data comes from is internal. And I'll give something that's going to apply to anybody that's got a physical product.

[00:54:52]And this is such a huge unlock for us because we talk about optimizing ads and our landing page and conversion rates, but there's a very real piece that for us, we felt that pain last year. We grew 100% year-over-year, but we were stipened or stifled because we ran out of inventory more often than we were in stock. And that's a real pain point is opportunity cost. So, I'll walk through very quickly what the opportunity cost was, how we identified it, and what we built. So, I said, "Go tell me when things are out of stock." He goes, "I need access to your back end." And he go, "No, you don't. If it's grayed out, it means it's out of stock. So, build a tracker and tell me in real time when things go out of stock.

[00:55:30]>> Let's go a step further from my own data, not a database. Let's plug in Google Analytics. I need you to tell me when something goes out of stock, what is our missed revenue?" And how it's calculating that when something's in stock, it knows the conversion rate 3.3%. It's on an instock skew. In this particular style, we're getting 800 900,000 people a day to this exact skew that's out of stock, which means if the conversion rate held, you're leaving, I think at one point we had $53,000 a week we were leaving on the table. So, what do we do with that? The next step is, well, let's take a look at our inventory, what our current production cycle looks like, and our POS. Are we ordering the right things?

[00:56:11]>> And before it was human bias. We're guessing. We launch a new style, it sells out quickly. I don't know if there's just a thousand people waiting for that one style and there's not a thousand in one. I don't know if there's 100,000 because there's human bias that's required. But because we've got the web traffic and data that we can plug into it, I'm actually writing the POS and they're going straight down to our factory. It knows our minimum order quantities. It knows any of our production rules. and it uncovered things. Hey, you're good in black and white, but that new style you launched, huge pent-up demand. And what that's doing is allowing us to be in stock with the products that people are looking for. And again, so much time and energy goes towards what's on fire. Where's the squeaky wheel? Well, if you don't have something in stock for people to buy, you're going to miss that revenue.

[00:56:58]Without that revenue, there's no growth.

[00:56:59]And so, that's a very practical tool that anybody can do today. just listen to the episode and follow through those steps and it will build it for you just like that. It just it's incredible. But that's such a huge unlock for me. So to answer the question as a long way of saying I'm looking at our own data as the most powerful source. I think Ben, I'll I'll I'll hand it back to you to close, but I I want to say one thing, Rob, which is you just touched on the most important thing and I think the next year or two years, it's all about internal context that you're able to give the model to deliver better outputs and candidly, right now we're limited in our ability to do that. Yes, there are connectors that you can connect to outside data sets. Yes, you can give it background as you're prompting, but the big challenge for organizations over the next one to two years will be how to standardize the process of sharing as much details with the AI about the company as you can and then making it easy for employees to at scale call on that context to action it in the right way. Um, there is so much there. Uh, and it's just it's interesting there. Yeah, I mean the the way that we're working is is changing and you are very clearly an exemplar of how that's the case. So Ben, I'll let you kind of close us out.

[00:58:12]>> Yeah, I think you both are touching on where this is all going, right? And for for small teams and new businesses, it's it's really easy to kind of I wouldn't say easy but easier to to build from the bottom up rather than uh you know make changes to existing businesses which is uh the operating system of a business.

[00:58:30]The operating system is AI. It is about connecting to data sources to enrich the context as much as possible. The more context, the more data you feed the AI, the better the outcome would be. And then, you know, we talked about orchestrating agents. We're going to talk about automation of that orchestration, which is agent harnesses, which is uh coming right up um in AI 101. Uh but yeah, this Rob, you are I I I said it before, I'll say it again.

[00:58:58]You're living the dream. Uh you're doing a great job. This is this is what uh it's all about using AI for for small businesses. Thank you for coming on the show. Uh if our viewers and listeners want to learn more about you and Stan Plus, uh how can they uh get a hold of you?

[00:59:15]>> Yeah, I'm on LinkedIn just Rob Greg Jury GG and we're stand.com.

[00:59:21]>> Wonderful.

[00:59:22]>> Love you, man.

[00:59:24]>> Thanks for coming on.

[00:59:25]>> Thank you guys.

[00:59:27]>> All right, AI 101. Today we are going to define what an agent harness is. This is a term by Mitchell Hashimoto um that was coined in February and now it's everywhere. And I I think a lot of times these terms can get muddled. Large language model, agent, agent harness.

[00:59:42]People are using them interchangeably when they're not. So Ben, why don't you define what a harness is for our listeners?

[00:59:48]>> Yeah, no problem. So an LLM is the raw model, right? Claw, GPT, Gemini, Grock, uh which is basically the language brain. An agent is an LLM running with tools. It can search the web. It can write a file. It can edit a spreadsheet.

[01:00:04]It can send an email. That kind of stuff. An agent harness is really the orchestration layer or the operational environment that runs and coordinates agents um to perform, you know, much more complicated tasks, right? The most famous agent harness or harnesses are cloud code uh um open claw. Hermes uh is a big one. Now uh these agent harnesses they they basically orchestrate uh the agent and decide based on the task or the prompt what context to use, what to remember between sessions, uh what they're allowed to do, uh what they're not allowed to do, which agent or sub agents handles which kind of work and and really what happens when when something breaks. So an agent harness is basically a piece of software, right?

[01:00:51]Using good old deterministic code, right? if this then that which tells the LLM what to do and when to do it and and kind of how to do it as well. So uh the the the better that piece of software uh for a specific purpose let's say coding for example the more you can get out of the LLM cloud code for example which was originally created as as an agent harness to to get the best coding out of uh claw LLMs right uh and thropic realized it's really good at other things right so they spun it to create co-work which is an agent harness that can do a whole a lot more than just coding. Open Claw is another example.

[01:01:35]It's the open- source equivalent where uh you self post it on on your local hardware and it connects to communication channels like Slack, Discord, Telegram, WhatsApp. So, you can talk to your agent via these channels, right? Hermes is another one uh which is very popular right now and it has persistent memory. It has a skill library that grows over time and and also the same multi- uh channel deployment as OpenClaw. People often mix up agent with agent harness and and maybe eventually they'll they'll mean the same, but at this early stage, I think it's it's good to understand the difference and the nuance here.

[01:02:12]>> So Ben, if I take a step back and try to understand everything we've learned in AI 101 over the course of the past few months, there are a couple of key terms that I just want to define to make sure I'm understanding it properly. So you've got the lab. In this case, the lab is anthropic. They then created the large language model claude. Within the large language model claude, there are effectively subm models. There's haiku, there's sonnet, there's opus. And then as you try to use those models for more advanced use cases in cloud code or cloud co-work like we heard from Zoe and Rob earlier today, that's when you're in the harness of it all. Is that correct?

[01:02:50]>> Look, yes. I mean, let let's uh let's give a the kitchen analogy, which is is my my favorite analogy to explain this.

[01:02:57]So that the imagine the LLM is the chef, right? It knows recipes and knows technique but without tools it can't cook much, right? Uh the agent is the chef with tools, right? It has knives, pan, stove, so they can cook some dishes. The harness is basically the whole kitchen, right? The recipes which are instructions and memory, the expediter, calling orders, which is the orchestration layer, right? The prep cooks, which are the sub aents, etc. So chef is the LLM, chef with tools is the agent, and then chef with an entire professional kitchen is an agent harness. Does that does that make sense?

[01:03:37]>> I'm tracking, man. And I know that we are making some delicious meals together with our agents and the harness, etc. So people go down and down. Uh let's cook, man. All right, on to the next.

[01:03:49]>> And that is it for episode 9 of the AI operators. Please like, share, subscribe. Thank you to Maheen. Thank you to Finn. Thank you to my wonderful co-host Ben and to Rob and to Zoe. And in closing, Andre, I am so sorry. I know that we had coordinated you making the announcement on the show today about the move to Anthropic. Uh we will make sure to have you on the show next week or the following because we couldn't get to it now. Guys, this is wonderful. We will see you for episode 10 soon.

[01:04:20]>> See you next week.