Arjay McCandless: Sam, welcome to the show. Congrats on joining Anthropic man. That's so exciting.
sam inteo: This is a conversation with Sam Gardner, a new member of the GoTarket team at Anthropic. Now, Sam is not a software engineer, but he's one of the more technical non engineers that I've talked to, who represents a new class of operator that sits somewhere between sales, product, engineering and go to market.
Sam Gardner: Thank you. Super excited to be here.
Arjay McCandless: So you're now on the GTM team, right? Can you kind of tell us a little bit about what that's like and how it's changed from your work at Stainless?
Sam Gardner: Yeah. ⁓ so I guess I'll start with my work at Stainless. I was sitting like the business operations team or product operations team. so anything that didn't directly fall to the go to market or the engineering team kind of fell into my hands. ⁓ and at a startup that means just like wearing a lot of hats and doing anything that people don't have time to do. ⁓
sam inteo: And all these roles are kind of stitched together by code and AI.
Arjay McCandless: And wait, for sorry, for people who don't know, can you give us like a one sense about what stainless is and what they do? 'Cause it's a super cool company.
Sam Gardner: Yeah, so stainless generates SDKs. so for people aren't familiar, a lot of engineers like working with SDKs instead of working with a raw API. And for companies that are offering API services to other engineers, it's very difficult to write SDKs in multiple different languages. and so Anthropic created a generator that automatically
sam inteo: In this episode, we talk about what it's like to join Anthropic right now. how people across different job functions internally are using AI.
Sam Gardner: creates SDKs ⁓ for whatever language that companies are looking for.
Arjay McCandless: Sainless created that generator, right? Sorry, you said it though. Yeah, no worries. Yeah. Yeah. And you guys were servicing a ton of major companies everyone's heard of, Anthropic being one of your major customers.
sam inteo: How Sam himself uses AI in his day-to-day work. How he got into multiple top-tier companies like Ramp, Stainless, Salesforce, and Anthropic.
Sam Gardner: Sorry. Sorry, ⁓ stainless. Yeah. Yes, and that's why like the acquisition happened. Anthropic was using us very heavily and they wanted to move these that services in house.
sam inteo: and where you should actually be using AI versus where you should be staying away from it.
Arjay McCandless: Cool. So you're on the BizOps team kind of working between engineering and GTM. What what did that kind of look like?
sam inteo: And I'm just getting this podcast off the ground, so if enjoy this episode, please subscribe and share it with your friends.
Sam Gardner: think was the epitome of my day was changing like all the time, ⁓ or my role was changing all the time. sometimes that was like up to go-to-market tooling, like Salesforce and Clay and any other tool that they needed. Sometimes that was like running a product-led growth effort, ⁓ fixing pricing, ⁓ or like building product ⁓ for like that the engineering team didn't have time to do. So really like
sam inteo: Let's get into the conversation.
Sam Gardner: It was a sixty person company, and you just had to do everything.
Arjay McCandless: What level of structure did you have in your work? Would someone come and say, Hey Sam, today we kinda wanna get this done or would you kind of just look at what was going on and decide for yourself what was the best thing to do that day?
Sam Gardner: It was a little bit more planned out, like, okay, this is what we're trying to achieve in the probably like the next month. And so trying to focus or stack rank the top three things that we're looking to achieve ⁓ and then focus on those. Otherwise it gets like a little chaotic.
Arjay McCandless: Yeah, that makes sense. And you were working, you know, you're building these SDKs, generating stuff from open API spec and all that. And you don't have a traditional engineering background. Like how did you kind of navigate that and get into the whole scene of writing code and building out these tools?
Sam Gardner: yeah, that's a really good question. So I was like a history major in college. I have no technical background. ⁓ I started my career at Ramp, ⁓ briefly on the sales team, and then most of my time ⁓ on the risk team, which was like slightly tangential to like engineers that they would help build a lot of our tools for us. ⁓ and I got introduced to like SQL pretty early in my career and how to build. retools in like using SQL. so that was kind of the start of it. And ⁓ really l like loved just writing SQL to like help me with my work. and that expanded into Python, ⁓ which I like can dabble in a little bit. but mostly just, you know, I'm not a very technical person, but I like working with technical teams a lot. it's like stainless was like a really good Because like the the engineers were were super strong there.
Arjay McCandless: Do you like AI has kind of drastically changed your level of technical prowess or like now you can do a lot more than you used to be able to do, or how does that kind of look?
Sam Gardner: ⁓ what do you mean by that?
Arjay McCandless: Hmm, okay, I'll let me let me speak to myself. when I was first learning how to code, you know, I could build like a simple website in ⁓ a week, let's just ⁓ And then, you know, moved to Amazon. Now I can build like a microservice in a quarter. And now it's like I can build entire products in a week or a of days and not really fully understand the tech stack. So I'm curious, if you feel like your initial understanding of like SQL and Python before AI. kind of transformed into you knowing kind of what's possible and what you're able to build by knowing that code side, but now you're able to execute more
Sam Gardner: I guess like I I I try like much more ambitious stuff with AI. Like I really push the boundaries of like what's possible and I'm like, hmm, I wonder if I'm capable or like able to do this at all. ⁓ and then just like throw stuff at the wall and see like if I'm if I'm able to do it. I find the more that I work with AI, the the actually more technical or like the more lower level that I get. Like I wanna start opening BS code and like looking at the code.
Arjay McCandless: Yeah.
Sam Gardner: which I don't do now. Like I'm strictly like vibe coding ⁓ in the way that I work, but it's it's me that like like engineers are super valuable. ⁓ and like I don't think that they're going away. I think it's really important to like the technical expertise that engineers have is like extremely valuable. ⁓ and yeah. I don't know if that answers your question.
Arjay McCandless: Yeah. Yeah, it makes sense. what is a specific big goal that you try take on with AI? Because I do feel like a lot of people are still thinking too small in like what types of tasks you can give it.
Sam Gardner: Yeah, so I think that like the way I'm starting to think about using AI is like it's getting really good at these like long running tasks. ⁓ if you ha if AI has access to like tools ⁓ and some like background information with like a system prompt or a user prompt, and then you give it access to like tools and looping, ⁓ that's where it can like really go off on its own and just like Iterate on itself and like think through like a very hard and complex problem. ⁓ that's where like I've to like really do the most with it. I'm trying to think of like specific problems or a task where like that ⁓ has happened. It's very much true in like in the go-to-market space, like we use like Salesforce, for example. ⁓ and if I have like an email thread where someone wants to ⁓ with ⁓ anthropic. ⁓ I can very easily like update all of Salesforce, log calls, do all of like the busy work of a salesperson in like one or two clicks. Like all of that, like repeating tasks happens automatically in Salesforce.
Arjay McCandless: Okay, that's pretty cool. When you talk about, you know, loops and like longer running goals, are you I guess like what does your tool set look like? Are you using dynamic workflows? Are you using fable, mythos, haiku? Are you always using the most powerful models now that you have unlimited tokens? And how does that kind of look?
Sam Gardner: Yeah, so at the moment it's just Opus because Fable is unfortunately yeah, no Fable. ⁓ But yeah, otherwise I am using like the most powerful model. and fortunately, like we get all unlimited access to those models, which is awesome. ⁓ and what I'm finding is just like the the models are really good at these like long running complex tasks.
Arjay McCandless: No, no fable, yeah. When you go about okay, like I guess specifically, what harness are you using? What exact like how are you formatting the goal you give it or the prompt you give it? And how do you close the feedback loop to make sure it can actually get that done in like one session if that's what you're doing?
Sam Gardner: Yeah, so I use Claud Code. and frankly, like not too much of my work is using like loops or goals. ⁓ I I think Cloud Code has like a built in looping system if it like if it ⁓ it needs to like think on it. ⁓ I know like coworkers that have talked like really highly of goals and loops and so like I wanna try using that more, but I haven't been able to like in my work necessarily.
Arjay McCandless: Yeah, that makes sense. I feel like I've I haven't needed it beyond just saying, accomplish this, you know, and I don't have to use like a specific framework to get it to do that and it'll run for forty minutes and like work on it. Especially if I give it away to check its own work, like a, you know, playwright MCP server or a USB connection to a phone.
Sam Gardner: Yeah. Yeah. I like I definitely need to it it depends on like the type of work you're doing. If you can like have the system check its or have Claude check its work, that's like super valuable. ⁓ also another thing to add on, I like try to add as much context as I can ⁓ to the model or like to the instruction. ⁓ so if that's like an email thread, I like give it the full email thread and say, Go check Salesforce as well.
Arjay McCandless: Yeah. Do have a specific prompting framework or anything you do to get the best results? Cause I will literally just get on like voice mode and yap for three to five minutes and like there's no structure. I'm just like dumping my raw thoughts into it and I get pretty good results. But I'm curious if you found a more like better way to do it.
Sam Gardner: No, I actually think that's like a really good ⁓ way to do it. ⁓ I also find that like my prompting gets worse in the evening as I'm like more groggy. Like it's actually important like to think through what you want to say to the model. ⁓ and like, okay, what is my goal here? What am I trying to achieve? ⁓ especially if you're like creating products ⁓ and you're thinking about like
Arjay McCandless: Yeah.
Sam Gardner: what you're trying to code. It's it's pretty important to like thinking, be thinking about the prompt and be thinking about what you're trying to create.
Arjay McCandless: So when you're crafting the prompt, you're really focused on like what the outcome is. And do you go into specifics about how you want it to be created? Or do you just kind of say, like, I want to accomplish this?
Sam Gardner: That's a good question. Sometimes if I'm like familiar enough with like an architecture decision, that can be very helpful to include in the prompt. but otherwise it's more goal goal oriented. A lot of the times I'm like using plan ⁓ mode to like have a conversation with the AI. I think it's like super, super helpful to have the AI ask you questions and kind of pick at how you're thinking about a problem before actually coding. So I use that a lot.
Arjay McCandless: What kind of threshold of a task do you feel like is worth planning for? Cause if it's ⁓ be, you know, five, ten minutes of work feels like overkill, but is there a point where you're like, Okay, I should plan this out?
Sam Gardner: I think it's pr it's a little difficult, but like anything that's like a big project, I try to plan out. ⁓ anything that's like above just like a small level ask, I'll I'll do use plan for. or just to like have AI pick at what I'm thinking. Like sometimes I'm just like not thinking clearly enough and I need someone to like question my thoughts.
Arjay McCandless: Yeah, that makes sense. Refine the scope down. In terms of kind of using AI, both I'm curious to hear about Anthropic Stainless. I think ramp you were there too early for this to be relevant, but now it's clear they have awesome systems for this. But how does kind of tool sharing and like practice sharing work across the company? Cause for me, you know, now I work alone, but even before this, when I was at Maven, every single person had their own.
Sam Gardner: Yeah.
Arjay McCandless: Pocket of skills, workflows, prompts. And this was like basically completely single player, if you will. We all had our own things, which were totally separate. And eventually we started committing skills to GitHub. But if you know, I would clone down someone's skill and not really get how it works. I didn't build it and maybe have to make some changes. I'm curious if you guys now have like a standard set skills or tools that everyone uses and kind of can talk about internally.
Sam Gardner: Yeah, that's a really good question. I and I imagine like a lot of people will have this issue. ⁓ that's still an issue that we have at Anthropic. ⁓ you'll they we do have like a standard set of skills. ⁓ but like sometimes they'll be ⁓ customed to like the person that created the skill and they just pushed it to GitHub and you're like, ⁓ I need to like fully adjust that. so I think it's just like you ⁓
Arjay McCandless: Yeah.
Sam Gardner: The role of an internal tooling person is really important now. where they'll actually be creating products for the go to market team or the engineering team. The role of an internal tooling person is really important now. where they'll actually be creating products for the go to market team or the engineering team. I guess engineering's pretty good at creating their own tools, but ⁓ yeah, like an internal tooling person is pretty important to create this like standardization across like go to market.
Arjay McCandless: Yeah, and you've worked in internal tooling a lot. I'm curious how, as the internal tooling person, you go about making sure these tools are actually exactly what people need if you don't do the ops yourself already,
Sam Gardner: What do you mean by that?
Arjay McCandless: Like, okay, so for me, I'll give you an example. I was making a video editing skill ⁓ as the leader of like our social media team. And I know exactly how I edit videos. I know exactly what my output looks like. But I'm not a hundred percent convinced that an internal tooling person would know that. So, like as an internal tooling person, you know, building for GTM sales partnerships, ⁓ whatever, do you go about really understanding what your team wants and building the correct tool?
Sam Gardner: I think that's like a good question for like any product person. Like you really have to talk to your customers. ⁓ for my instance, like I'm doing the work of go to market right now, and so I know exactly what to build. So that's super helpful. I also think a lot about like this has been on my mind a lot recently where ⁓
Arjay McCandless: Yeah. Yeah, so you.
Sam Gardner: in the past, engineers had VS Code or whatever IDE they were working in. in the past, engineers had VS Code or whatever IDE they were working in. And those were like super customizable and engineers love to trick those out. And that's gonna be the the same, in my opinion, for future go to market And those were like super customizable and engineers love to trick those out. And that's gonna be like the the same, in my opinion, for like future go to market tooling where like salespeople will have that same ability to trick out their their tooling. And so that's when I build tools for other teams, I really think about that process of how can I help this individual person trick out the tools so that they can make it perfect for their workflow.
Arjay McCandless: That's actually a really cool insight. I like that. Like just thinking about all your GTM stuff basically in the same way as VS Code. I mean, it would be awesome too if you had like the marketplace of unlimited plugins and I can put my VS code pets in the corner. I don't think it's quite there yet, but but you know what I'm talking about. So when you're using Claud when you see, you you're walking through the anthropic office, do you see people more in the terminal and the desktop app and does it matter?
Sam Gardner: Yeah. like ⁓ older go to market people, it might just be the the clo like cloud chat or like cloud cowork. I think we we're pushing like everyone should use c cowork ⁓ over chat just because it's so much more powerful. ⁓ and I think it's a lot of like desktop and then you'll see a a good amount working in the terminal as well, ⁓ just like cloud code.
Arjay McCandless: Yeah. Yeah. Between ⁓ for people that are less technical, do you see much of a difference between Cloud Code and Cowork? And my last company, I I I sat down with some of our partnerships team and I was teaching them Cloud Code. And at the time, I think cowork was still kind of in its earlier days. And I I personally just prefer Cloud Code. I I think it's better. And, you know, I'm a developer, so whatever. Like it's I'm very biased, but yeah, I'm curious, like kind of
Sam Gardner: Yeah.
Arjay McCandless: Do you have any feelings w in specific cases where you would pick up cowork over code, or is it kinda like if you can use code, you just prefer it?
Sam Gardner: I don't know the exact differences between the two. I know they're pretty similar. I heard once where like if you're afraid of a terminal, use ⁓ lose cowork. but they're very similar and coworkers like very similarly powered as cloud code. So if you're not using cloud code or you're scared of a terminal, I would I would use cowork and and try it out.
Arjay McCandless: Yeah, that's my impression. Yeah. Yeah. Okay, let's take like one step back. So, you know, your career's kind of been all over the place. and now you're in GTM, and I feel like that's actually kind of consistent with a lot of GTM folks that meet. especially on the GTM engineering side, where now you're really kind of building a lot more than you are. Just like a traditional salesperson. can you kinda walk me through your past and how these skill sets have led you to where you are now?
Sam Gardner: Yeah. This is like a little contrarian, but I've kind of taken my like twenties to explore like what I've wanted to do. So I've tried out a lot of different roles. was briefly sales at Ramp before spending most of my time on the risk team. where I like was able to work with technical teams. I moved to Salesforce to be on their sales team. I started my own company, which like you have to do everything for. and then at Stainless I like Doing that mix of biz ops and product ops. I think that journey is picking up little and pieces. It was a very like breath first search kind of way of finding like what you want to do. ⁓ and it's like helped me speak. ⁓
Arjay McCandless: Yeah.
Sam Gardner: It's helped me do my go to market job at a much higher level. Like I'm able to build my own tooling ⁓ as a go to market person. I'm able to speak with technical teams, ⁓ like for prospects looking to buy Anthropic. so I'm able to just like do my job at like a much higher level because I've seen so much more of like different roles ⁓ worked with so many different people.
Arjay McCandless: How do you go about landing these jobs, ramp, sales force, stainless, anthropic? I mean, this is like an eleven star resume. ⁓ do you have any secret sauce for how you got into all these great companies? Especially when you're switching roles all the time and not a textbook candidate. ⁓
Sam Gardner: Yeah. Yeah. I don't know if I should like advise switching roles that frequently. I would just like LinkedIn is like a place, you know, like meet people on LinkedIn or reach out to people on LinkedIn. ⁓ going down the traditional, like if are applying to different places, don't go down applying through their website. if are applying to different places, don't go down like the ⁓ applying through their website. I see who's recruiting for that role and message them directly on like a simple LinkedIn message on just that you notice a role. so for stainless for I think I reached out to the CEO on like that he was hiring.
Arjay McCandless: Nice. And you know, what what do you say to him? Hey, I'm running my own company, but I like what you guys are doing. Like, what does this look like?
Sam Gardner: Yeah, something like that. I think it's really important, like the why. Like why are you interested in the role? And I always try to think, even before I start like looking at a new company, I'm always thinking like why do I want to work at this place? or what do I want to do? Actually, before I like even think about the company, what exactly do I want to do? And then matching that to like which companies I'm to. And it it gives a good interview process. Like I know exactly why I'm in this interview. I know what I want to do here.
Arjay McCandless: Yeah. This is funny because, you know, this is the second episode and the last guest said the exact same thing. Like, do not apply with your resume on their website without doing anything else. And I posted a clip of it and it was like people were like, I don't know why I would spend more than three minutes, you know, trying to get into a company when they're just gonna auto-reject me with AI. And I think the point that everyone's making, which I think you're making too, and I agree with, is like You don't want to just put your resume in the pile with everybody else. You have to have some other contact into the company because yes, if you put in your resume in two minutes to apply, you're probably gonna get rejected in two minutes. So yeah.
Sam Gardner: Yes, you will. And everyone does. Like I would get rejected instantly.
Arjay McCandless: Yeah, even I get rejected inst like everyone I know who applies with their resume just gets rejected instantly. And sure you can send out a thousand applications and maybe get a couple hits. But so when you're doing this kind of like targeted cold outreach strategy, how many companies like do you have in mind? Was it, you know, stainless is my dream company or is it I'm looking for something, maybe pick five or six, shoot a message, you know?
Sam Gardner: I don't remember exactly. It was pretty like tight list, maybe like ten or so companies. Stainless was definitely one of the if the top, just because of like how strong their engineering team was. I could tell very early on that they cared a lot about their product. and that mattered like a lot for me.
Arjay McCandless: And where are you looking to get this information? LinkedIn, X, company website, engineering blog? Like where exactly do you try to ascertain that about a company?
Sam Gardner: I think just like in doing research, like you like heard the Alex the CEO, talk on a podcast ⁓ and like to how he thought about the product and what the vision for the company was was very compelling to me. and so yeah, just like people people that care, people that are, you know, not trying to build a company just to make money, but you know, to make something that's actually valuable for others.
Arjay McCandless: Yeah, that seems very aligned with Anthropic's vision. So don't know if you can talk much about this, but during the acquisition, was there interview or was it kind of just like, Hey, welcome, you're on this team now? ⁓ Really? Wow. I assume not really able to talk about that much.
Sam Gardner: I wish it was I wish it was the latter. ⁓ yeah, we went through the full interview loop. I did like ten interviews. Yeah. Yeah. Um yeah, I mean it was just like it was a normal hiring process. ⁓ I didn't get like a free ticket onto Anthropic. but I did get like to skip maybe the introduction call otherwise.
Arjay McCandless: Yeah. Wow. That's actually pretty crazy 'cause, you know, from the outside looking in, it's like you get acquired. Congrats. You now work here. Did most of your company kind of join then?
Sam Gardner: Yeah. think it was like around sixty percent. Like Anthropic is very like picky on who they want. we didn't bring along like some very, very talented people. so yeah, if anyone's looking to hire a AEs or or engineers, like I know some fantastic people to work with.
Arjay McCandless: Awesome. I'll make a note of that. Do you think there's anything in particular you to stand out?
Sam Gardner: I think it was help like it it helped that I used the most like cloud code in the company. Like I was the most AI pilled. ⁓ and I had so many projects, personally at the company that I was building with AI. So I could just talk a lot about that. that that was probably what stood out. So I was like already ready for that interview before
Arjay McCandless: Yeah. Yeah. Person. Yeah. Yeah. Personally, what are you kind of building outside of work that you're most proud of right now?
Sam Gardner: The acquisition happened. my current project is strictly work focused and so I'm spending most of my time on that. in the past I've done some really cool projects. I like playing a lot of chess, so I made a camera that watches the chessboard and uses ML to track your moves. I made a Calci trading bot, I've stopped by the way.
Arjay McCandless: Ha
Sam Gardner: That was an interesting one where I was like, I'm just gonna put a hundred dollars into this. And I strictly wanted to see if I could use AI to to build something that I knew nothing about. Like I know nothing about weather trading. And so I was curious how far the intelligence would take me. And that actually didn't work that well. And it it approved an interesting point, or I don't know, approved, but it was one data point that you can't just like go into a field you have no idea about and and try to win it. you need to actually have expertise and use AI as a tool, not like a crutch. you need to actually have expertise and use AI as a tool, not like a crutch.
Arjay McCandless: I completely agree with you. I think the best use case for AI right now, and is maybe possibly the only one in my opinion, is being very good at something that you already are good at. Because I've tried to go zero to one on things that I have no idea how to do and it is not helpful. mean, and maybe maybe I'm doing it wrong. maybe what I should be saying is teach me how to do XYZ, not do XYZ. But when I say do XYZ and I get something back, I'm like, okay. You know, like I'm yeah.
Sam Gardner: Yeah, you have to trust it.
Arjay McCandless: Like and it's this is not the perfect example because everyone talks about taste, but I have zero design sense. And when I go to design a website or something, it'll make a website. And I'm like, Looks like a website to me. like, this is a website. And but I have no idea, you know. So I I'm often finding myself texting a designer, a friend and being like, Yo, is this good? And they're like, No. it's just okay, my bad. yeah, that's that's interesting. So for your Chess ML project, how ⁓ How much like are you writing, you neural nets or is it more like import, C V, whatever?
Sam Gardner: No, yeah, it's just like taking stuff off the shelf, open projects, maybe finding a data set online and like stitching everything together. that one was pushing the boundaries of what I knew. so I was trying to learn and and push the boundaries as well. Yeah.
Arjay McCandless: I think it's a good, like interesting point there where people feel like to be building AI projects, you have to be, you know, fine tuning models or building things from zero. And I just I think there's a ton of value to be built and captured on like the application layer of AI. Like you can import these APIs and do like ridiculously cool stuff. and those are very valid AI product projects that companies want to see on your resume man. They want to see the open sourcing and building. And I think People get like they don't think that's as valuable as it is, I guess. So it's nice to kind of hear you say that's the kind of projects you're working on without being super deep in the weeds.
Sam Gardner: Yeah, and I just like building. Like it's just something that if I have a free weekend, I'll maybe tinker around for a little while. so yeah, I mean I would just like do the smallest project you can think of and just and work on it. Yeah.
Arjay McCandless: do you do anything to these projects into the wild? Like are you open sourcing them, posting them on LinkedIn, X, telling your friends about them, kind of just putting them on your resume?
Sam Gardner: don't even know if I put on my resume. I need to do a much better job of like I strictly do these projects because I think they're fun or useful. I posted the chess one on LinkedIn because I just thought it was cool and like people also thought it was cool. But no one was reaching out to me and like, hey, this is really cool. You should ⁓ yeah, I'm gonna hire ⁓ so it's strictly like was doing it for fun and and I open sourced like the chess ML project.
Arjay McCandless: Yeah. Yeah, I'm gonna hire you. Mm-hmm. Nice. I always tell people when you're picking projects, just pick something that you want to do, because that's how you're gonna finish it and you're gonna learn all the important skills without forcing yourself to build some thing that you're not interested or gonna give up on after a couple hours. That's cool. What do you think it's gonna take for you to really crush at anthropic?
Sam Gardner: That's a good question. It's difficult 'cause like the bar here is really, really high. Like the talent is incredible.
Arjay McCandless: Yeah.
Sam Gardner: Frankly, like I don't know right answer to that. And I'm pr I'm new enough where I'm still trying to get my bearings. I think what I'm really trying to do is just be like helpful and kind. you know, like when you when you don't know what to do or how to perform, it's just like be helpful wherever you can. that's kind of the mindset I'm going into this with. Like work hard and and be kind.
Arjay McCandless: Yeah. Yeah, yeah. Yeah. Have you been able to ascertain much about culture? is it nine nine seven or and are people helpful and you know?
Sam Gardner: People are really helpful. The culture is like fantastic here. And that's like what they're like really hiring for. is like a lot of work. Like, you know, I was up till like eleven or midnight last night, like just doing email. and so people work really hard here, because like the cause is so important, and people really do care about it. that being said, is like everyone's willing to help you. and everyone's very kind, everyone's
Arjay McCandless: Awesome.
Sam Gardner: very cognizant that everyone else is busy. yeah. ⁓
Arjay McCandless: Yeah. How do you feel about AI as a job creator versus taker and kind of where does like time saving fall into that? Because In some sense, I think to myself, shouldn't we be able to solve like responding to incoming emails or make that ten times faster and free you up to do something else? And kind of like how do you reconcile that with what work you're doing and, you know, where you're at?
Sam Gardner: I kind of see like two different camps. see instances where like I work so much harder. Like I love like what I'm able to do with AI and I'm able to like push the boundaries of what I'm capable of and what AI is capable of. and so I'm definitely working more and I think other people here are working more and they use AI as a tool. what scares me a little bit is like when you use AI as a crutch. and I feel like there are a lot of jobs that can get automated away.
Arjay McCandless: Yeah.
Sam Gardner: Anthropic like shy away from that. Like we're fully aware that our technology will automate a lot of jobs away ⁓ in America and the world. I think that will ⁓ I think over the long run, like that like people will find new jobs, but there definitely is gonna be like a tumultuous period in in between. What scares me more or like I'm more concerned of is like cognitive offload where people don't use the technology as a tool, but ⁓ like students, for example, they just like do all their homework with AI. And I think that's like actually very hard to solve ⁓ and it's gonna be very destructive.
Arjay McCandless: Do you take intentional steps to make yourself think through problems or intentionally do something without AI ever just to make sure you're keeping those muscles sharp? Or how do you handle that?
Sam Gardner: So like I'm a less technical person, so I'm not writing code, but like writing is a good example. If it's like a utility writing, I'm usually writing it with code. ⁓ excuse me, with AI. but ⁓ like Paul Graham like describes it pretty well as like writing is thinking. ⁓ if you're writing something if you're thinking through something important, it's like really important to like write it down. And AI won't be able to replace that process. It's like Taking a very nebulous thought and distilling it into like the simplest form it can be. so in those instances I try to write it myself. and it's difficult in that like that muscle can atrophy and I it definitely has been atrophying. so it is important to like in my instance continue to write. In engineers' instances, like continue to code like very important parts. I I'm kind of speaking out of my but like I ⁓ I assume engineers should do that.
Arjay McCandless: Yeah, it's interesting. Like I've definitely become significantly worse at coding by hand. I mean, if you if you had me like go one on one against myself two years ago, it would not even be close. Like night and day. But at the same time, my like output is much higher. you know, like my marginal return on time is much, much greater. Like with AI versus without it. And yeah, it's a question of I think this is something a lot of engineers grapple with too, is like
Sam Gardner: Yeah.
Arjay McCandless: Okay, the current models, let's just say write code that's 80% good as a pretty good engineer. And if that code is 80% as good and we write our entire code base like that, then you let's say Fable comes back ⁓ and can write code 90% as good as an engineer, then I can just say, fix the slop in my code base, make it a little bit better. Then we get another model that's even better and it does it again, and then eventually, like in the long horizon. It never made sense for me to be like manually checking our code now, anyways. I don't know. It's like an existential thought that I think a lot of people are having across industries.
Sam Gardner: I think and and you know better than than me on this, so correct me if I'm wrong, but a a lot of code is like utilitarian ⁓ and that it's just like serving a purpose to get a job done. but like I'm working on projects now, so much of work is like slowing down and not AI to like think through like how this like specific button should function.
Arjay McCandless: Yeah.
Sam Gardner: And maybe like the end state is like you're still using AI to to like build that that button. but again like if you're just cooking with AI and having it build everything, then like you're not thinking about getting your product like a much higher level.
Arjay McCandless: Yeah, okay. This could be a good closing thought here. I guess what is your take on when to use AI and when not to use AI? You kind of said it there, but I'm I want to hear you say it a little more like explicitly in like a kind of I don't know, repeatable way. And obviously there's not a correct answer to this. I'm just curious, like your take on it.
Sam Gardner: This is like a really hard question. ⁓ I think for for outside of coding, it's what I explained before in like if it's really important and needs to be thought through in like in terms of writing, then I'm definitely just writing it by hand and not using AI because like I need to d work through the process of thinking. With code, I think of it more as like almost like driving somewhere that I I don't know. the exact destination, but I know it's like far away and it's in this general direction. And AI can like put you on the highway and like get you in the general vicinity. And then you need to slow down. And I don't know if it's by hand. It might still be by code, but you need to be much more surgical. And being surgical with AI can get a little difficult unless you're like working in in the term in the like IDE. and then you're really just you know, the actual code or yeah, it's like a lot lot a lot less AI.
Arjay McCandless: Knowledge work, how about making a presentation, sending an email, stuff that.
Sam Gardner: Again, it depends. Like if this is for creating a presentation, if this is like the one presentation that matters, you like generate the first draft in AI, and then you're thinking about I think about like, all right, what am I trying to say? What are the exact words that are going on here? Like, I'm not using AI for that. I'm like thinking through what is my purpose with this presentation? Am I conveying that and being very surgical? In an email, if it's like an important client, it's like the same thing. It's like getting a rough draft and then like edit editing.
Arjay McCandless: Yeah. I I agree with everything you said. I think one of my biggest issues I've had with AI in human to human communication, things like an email, even a LinkedIn post or a presentation is you get too many words. Like I'm very, very about slide decks in particular. If I see more than fifteen words on a slide, it's like I'm not reading all that. You know, like And I think a lot of people too, even if I'm reading a report from somebody or something that somebody sent me and it's like three pages and it should have been three sentences, like not happy. You know, I don't want to spend my time like I would have rather have read the prompts that you gave to get this than read this. Cause this is just like downstream of that. So ⁓ yeah, don't know. That's interesting. I think there's a time and a place. I had a had an old coworker at a like a couple companies ago, and I saw him
Sam Gardner: Yes.
Arjay McCandless: Write a message to Chat GPT and then copy paste the message directly from Chat GPT to me. And it's just like a personal message, like asking for help with something. And I was like, dude, like we don't need this. You know? And I I think like just people don't fall too deep into it. We're still humans, you know? Yeah.
Sam Gardner: Yeah, yeah, yeah. Like I mean, I would recommend if it's your voice, like trying to stay away from AI. ⁓ ⁓ yeah, AI Slack messages are just like not fun to read and they're too verbose. ⁓ yeah, even emailing and go to market, like sometimes you know they're using AI and you're using AI and you're trying to convey like information, so it's helpful to use AI, but it's just like not this is
Arjay McCandless: Yeah. Yes. Mm-hmm.
Sam Gardner: This is not right.
Arjay McCandless: Yeah. Yeah. A hundred percent. All right. It was awesome having you on. Thank you for sharing everything you have.
Sam Gardner: RJ, thank you. This is this has been super fun.
Arjay McCandless: Is there anything you want to share with people, like anywhere they should follow you, anything they should check out of yours?
Sam Gardner: don't know, feel free to like check out my LinkedIn. ⁓ I have like a yeah, I check out my chess project. I have a personal website that kinda like talks about who I am as a person. ⁓ that's on my LinkedIn as well. I always like like showing that off.
Arjay McCandless: Nice. Check out your chest project. Awesome. All right. Thank you for coming on. We'll see ya.
Sam Gardner: Thanks, Arja.
sam inteo: This is a conversation with Sam Gardner, a new member of the GoTarket team at Anthropic. Now, Sam is not a software engineer, but he's one of the more technical non engineers that I've talked to, who represents a new class of operator that sits somewhere between sales, product, engineering and go to market.
Sam Gardner: Thank you. Super excited to be here.
Arjay McCandless: So you're now on the GTM team, right? Can you kind of tell us a little bit about what that's like and how it's changed from your work at Stainless?
Sam Gardner: Yeah. ⁓ so I guess I'll start with my work at Stainless. I was sitting like the business operations team or product operations team. so anything that didn't directly fall to the go to market or the engineering team kind of fell into my hands. ⁓ and at a startup that means just like wearing a lot of hats and doing anything that people don't have time to do. ⁓
sam inteo: And all these roles are kind of stitched together by code and AI.
Arjay McCandless: And wait, for sorry, for people who don't know, can you give us like a one sense about what stainless is and what they do? 'Cause it's a super cool company.
Sam Gardner: Yeah, so stainless generates SDKs. so for people aren't familiar, a lot of engineers like working with SDKs instead of working with a raw API. And for companies that are offering API services to other engineers, it's very difficult to write SDKs in multiple different languages. and so Anthropic created a generator that automatically
sam inteo: In this episode, we talk about what it's like to join Anthropic right now. how people across different job functions internally are using AI.
Sam Gardner: creates SDKs ⁓ for whatever language that companies are looking for.
Arjay McCandless: Sainless created that generator, right? Sorry, you said it though. Yeah, no worries. Yeah. Yeah. And you guys were servicing a ton of major companies everyone's heard of, Anthropic being one of your major customers.
sam inteo: How Sam himself uses AI in his day-to-day work. How he got into multiple top-tier companies like Ramp, Stainless, Salesforce, and Anthropic.
Sam Gardner: Sorry. Sorry, ⁓ stainless. Yeah. Yes, and that's why like the acquisition happened. Anthropic was using us very heavily and they wanted to move these that services in house.
sam inteo: and where you should actually be using AI versus where you should be staying away from it.
Arjay McCandless: Cool. So you're on the BizOps team kind of working between engineering and GTM. What what did that kind of look like?
sam inteo: And I'm just getting this podcast off the ground, so if enjoy this episode, please subscribe and share it with your friends.
Sam Gardner: think was the epitome of my day was changing like all the time, ⁓ or my role was changing all the time. sometimes that was like up to go-to-market tooling, like Salesforce and Clay and any other tool that they needed. Sometimes that was like running a product-led growth effort, ⁓ fixing pricing, ⁓ or like building product ⁓ for like that the engineering team didn't have time to do. So really like
sam inteo: Let's get into the conversation.
Sam Gardner: It was a sixty person company, and you just had to do everything.
Arjay McCandless: What level of structure did you have in your work? Would someone come and say, Hey Sam, today we kinda wanna get this done or would you kind of just look at what was going on and decide for yourself what was the best thing to do that day?
Sam Gardner: It was a little bit more planned out, like, okay, this is what we're trying to achieve in the probably like the next month. And so trying to focus or stack rank the top three things that we're looking to achieve ⁓ and then focus on those. Otherwise it gets like a little chaotic.
Arjay McCandless: Yeah, that makes sense. And you were working, you know, you're building these SDKs, generating stuff from open API spec and all that. And you don't have a traditional engineering background. Like how did you kind of navigate that and get into the whole scene of writing code and building out these tools?
Sam Gardner: yeah, that's a really good question. So I was like a history major in college. I have no technical background. ⁓ I started my career at Ramp, ⁓ briefly on the sales team, and then most of my time ⁓ on the risk team, which was like slightly tangential to like engineers that they would help build a lot of our tools for us. ⁓ and I got introduced to like SQL pretty early in my career and how to build. retools in like using SQL. so that was kind of the start of it. And ⁓ really l like loved just writing SQL to like help me with my work. and that expanded into Python, ⁓ which I like can dabble in a little bit. but mostly just, you know, I'm not a very technical person, but I like working with technical teams a lot. it's like stainless was like a really good Because like the the engineers were were super strong there.
Arjay McCandless: Do you like AI has kind of drastically changed your level of technical prowess or like now you can do a lot more than you used to be able to do, or how does that kind of look?
Sam Gardner: ⁓ what do you mean by that?
Arjay McCandless: Hmm, okay, I'll let me let me speak to myself. when I was first learning how to code, you know, I could build like a simple website in ⁓ a week, let's just ⁓ And then, you know, moved to Amazon. Now I can build like a microservice in a quarter. And now it's like I can build entire products in a week or a of days and not really fully understand the tech stack. So I'm curious, if you feel like your initial understanding of like SQL and Python before AI. kind of transformed into you knowing kind of what's possible and what you're able to build by knowing that code side, but now you're able to execute more
Sam Gardner: I guess like I I I try like much more ambitious stuff with AI. Like I really push the boundaries of like what's possible and I'm like, hmm, I wonder if I'm capable or like able to do this at all. ⁓ and then just like throw stuff at the wall and see like if I'm if I'm able to do it. I find the more that I work with AI, the the actually more technical or like the more lower level that I get. Like I wanna start opening BS code and like looking at the code.
Arjay McCandless: Yeah.
Sam Gardner: which I don't do now. Like I'm strictly like vibe coding ⁓ in the way that I work, but it's it's me that like like engineers are super valuable. ⁓ and like I don't think that they're going away. I think it's really important to like the technical expertise that engineers have is like extremely valuable. ⁓ and yeah. I don't know if that answers your question.
Arjay McCandless: Yeah. Yeah, it makes sense. what is a specific big goal that you try take on with AI? Because I do feel like a lot of people are still thinking too small in like what types of tasks you can give it.
Sam Gardner: Yeah, so I think that like the way I'm starting to think about using AI is like it's getting really good at these like long running tasks. ⁓ if you ha if AI has access to like tools ⁓ and some like background information with like a system prompt or a user prompt, and then you give it access to like tools and looping, ⁓ that's where it can like really go off on its own and just like Iterate on itself and like think through like a very hard and complex problem. ⁓ that's where like I've to like really do the most with it. I'm trying to think of like specific problems or a task where like that ⁓ has happened. It's very much true in like in the go-to-market space, like we use like Salesforce, for example. ⁓ and if I have like an email thread where someone wants to ⁓ with ⁓ anthropic. ⁓ I can very easily like update all of Salesforce, log calls, do all of like the busy work of a salesperson in like one or two clicks. Like all of that, like repeating tasks happens automatically in Salesforce.
Arjay McCandless: Okay, that's pretty cool. When you talk about, you know, loops and like longer running goals, are you I guess like what does your tool set look like? Are you using dynamic workflows? Are you using fable, mythos, haiku? Are you always using the most powerful models now that you have unlimited tokens? And how does that kind of look?
Sam Gardner: Yeah, so at the moment it's just Opus because Fable is unfortunately yeah, no Fable. ⁓ But yeah, otherwise I am using like the most powerful model. and fortunately, like we get all unlimited access to those models, which is awesome. ⁓ and what I'm finding is just like the the models are really good at these like long running complex tasks.
Arjay McCandless: No, no fable, yeah. When you go about okay, like I guess specifically, what harness are you using? What exact like how are you formatting the goal you give it or the prompt you give it? And how do you close the feedback loop to make sure it can actually get that done in like one session if that's what you're doing?
Sam Gardner: Yeah, so I use Claud Code. and frankly, like not too much of my work is using like loops or goals. ⁓ I I think Cloud Code has like a built in looping system if it like if it ⁓ it needs to like think on it. ⁓ I know like coworkers that have talked like really highly of goals and loops and so like I wanna try using that more, but I haven't been able to like in my work necessarily.
Arjay McCandless: Yeah, that makes sense. I feel like I've I haven't needed it beyond just saying, accomplish this, you know, and I don't have to use like a specific framework to get it to do that and it'll run for forty minutes and like work on it. Especially if I give it away to check its own work, like a, you know, playwright MCP server or a USB connection to a phone.
Sam Gardner: Yeah. Yeah. I like I definitely need to it it depends on like the type of work you're doing. If you can like have the system check its or have Claude check its work, that's like super valuable. ⁓ also another thing to add on, I like try to add as much context as I can ⁓ to the model or like to the instruction. ⁓ so if that's like an email thread, I like give it the full email thread and say, Go check Salesforce as well.
Arjay McCandless: Yeah. Do have a specific prompting framework or anything you do to get the best results? Cause I will literally just get on like voice mode and yap for three to five minutes and like there's no structure. I'm just like dumping my raw thoughts into it and I get pretty good results. But I'm curious if you found a more like better way to do it.
Sam Gardner: No, I actually think that's like a really good ⁓ way to do it. ⁓ I also find that like my prompting gets worse in the evening as I'm like more groggy. Like it's actually important like to think through what you want to say to the model. ⁓ and like, okay, what is my goal here? What am I trying to achieve? ⁓ especially if you're like creating products ⁓ and you're thinking about like
Arjay McCandless: Yeah.
Sam Gardner: what you're trying to code. It's it's pretty important to like thinking, be thinking about the prompt and be thinking about what you're trying to create.
Arjay McCandless: So when you're crafting the prompt, you're really focused on like what the outcome is. And do you go into specifics about how you want it to be created? Or do you just kind of say, like, I want to accomplish this?
Sam Gardner: That's a good question. Sometimes if I'm like familiar enough with like an architecture decision, that can be very helpful to include in the prompt. but otherwise it's more goal goal oriented. A lot of the times I'm like using plan ⁓ mode to like have a conversation with the AI. I think it's like super, super helpful to have the AI ask you questions and kind of pick at how you're thinking about a problem before actually coding. So I use that a lot.
Arjay McCandless: What kind of threshold of a task do you feel like is worth planning for? Cause if it's ⁓ be, you know, five, ten minutes of work feels like overkill, but is there a point where you're like, Okay, I should plan this out?
Sam Gardner: I think it's pr it's a little difficult, but like anything that's like a big project, I try to plan out. ⁓ anything that's like above just like a small level ask, I'll I'll do use plan for. or just to like have AI pick at what I'm thinking. Like sometimes I'm just like not thinking clearly enough and I need someone to like question my thoughts.
Arjay McCandless: Yeah, that makes sense. Refine the scope down. In terms of kind of using AI, both I'm curious to hear about Anthropic Stainless. I think ramp you were there too early for this to be relevant, but now it's clear they have awesome systems for this. But how does kind of tool sharing and like practice sharing work across the company? Cause for me, you know, now I work alone, but even before this, when I was at Maven, every single person had their own.
Sam Gardner: Yeah.
Arjay McCandless: Pocket of skills, workflows, prompts. And this was like basically completely single player, if you will. We all had our own things, which were totally separate. And eventually we started committing skills to GitHub. But if you know, I would clone down someone's skill and not really get how it works. I didn't build it and maybe have to make some changes. I'm curious if you guys now have like a standard set skills or tools that everyone uses and kind of can talk about internally.
Sam Gardner: Yeah, that's a really good question. I and I imagine like a lot of people will have this issue. ⁓ that's still an issue that we have at Anthropic. ⁓ you'll they we do have like a standard set of skills. ⁓ but like sometimes they'll be ⁓ customed to like the person that created the skill and they just pushed it to GitHub and you're like, ⁓ I need to like fully adjust that. so I think it's just like you ⁓
Arjay McCandless: Yeah.
Sam Gardner: The role of an internal tooling person is really important now. where they'll actually be creating products for the go to market team or the engineering team. The role of an internal tooling person is really important now. where they'll actually be creating products for the go to market team or the engineering team. I guess engineering's pretty good at creating their own tools, but ⁓ yeah, like an internal tooling person is pretty important to create this like standardization across like go to market.
Arjay McCandless: Yeah, and you've worked in internal tooling a lot. I'm curious how, as the internal tooling person, you go about making sure these tools are actually exactly what people need if you don't do the ops yourself already,
Sam Gardner: What do you mean by that?
Arjay McCandless: Like, okay, so for me, I'll give you an example. I was making a video editing skill ⁓ as the leader of like our social media team. And I know exactly how I edit videos. I know exactly what my output looks like. But I'm not a hundred percent convinced that an internal tooling person would know that. So, like as an internal tooling person, you know, building for GTM sales partnerships, ⁓ whatever, do you go about really understanding what your team wants and building the correct tool?
Sam Gardner: I think that's like a good question for like any product person. Like you really have to talk to your customers. ⁓ for my instance, like I'm doing the work of go to market right now, and so I know exactly what to build. So that's super helpful. I also think a lot about like this has been on my mind a lot recently where ⁓
Arjay McCandless: Yeah. Yeah, so you.
Sam Gardner: in the past, engineers had VS Code or whatever IDE they were working in. in the past, engineers had VS Code or whatever IDE they were working in. And those were like super customizable and engineers love to trick those out. And that's gonna be the the same, in my opinion, for future go to market And those were like super customizable and engineers love to trick those out. And that's gonna be like the the same, in my opinion, for like future go to market tooling where like salespeople will have that same ability to trick out their their tooling. And so that's when I build tools for other teams, I really think about that process of how can I help this individual person trick out the tools so that they can make it perfect for their workflow.
Arjay McCandless: That's actually a really cool insight. I like that. Like just thinking about all your GTM stuff basically in the same way as VS Code. I mean, it would be awesome too if you had like the marketplace of unlimited plugins and I can put my VS code pets in the corner. I don't think it's quite there yet, but but you know what I'm talking about. So when you're using Claud when you see, you you're walking through the anthropic office, do you see people more in the terminal and the desktop app and does it matter?
Sam Gardner: Yeah. like ⁓ older go to market people, it might just be the the clo like cloud chat or like cloud cowork. I think we we're pushing like everyone should use c cowork ⁓ over chat just because it's so much more powerful. ⁓ and I think it's a lot of like desktop and then you'll see a a good amount working in the terminal as well, ⁓ just like cloud code.
Arjay McCandless: Yeah. Yeah. Between ⁓ for people that are less technical, do you see much of a difference between Cloud Code and Cowork? And my last company, I I I sat down with some of our partnerships team and I was teaching them Cloud Code. And at the time, I think cowork was still kind of in its earlier days. And I I personally just prefer Cloud Code. I I think it's better. And, you know, I'm a developer, so whatever. Like it's I'm very biased, but yeah, I'm curious, like kind of
Sam Gardner: Yeah.
Arjay McCandless: Do you have any feelings w in specific cases where you would pick up cowork over code, or is it kinda like if you can use code, you just prefer it?
Sam Gardner: I don't know the exact differences between the two. I know they're pretty similar. I heard once where like if you're afraid of a terminal, use ⁓ lose cowork. but they're very similar and coworkers like very similarly powered as cloud code. So if you're not using cloud code or you're scared of a terminal, I would I would use cowork and and try it out.
Arjay McCandless: Yeah, that's my impression. Yeah. Yeah. Okay, let's take like one step back. So, you know, your career's kind of been all over the place. and now you're in GTM, and I feel like that's actually kind of consistent with a lot of GTM folks that meet. especially on the GTM engineering side, where now you're really kind of building a lot more than you are. Just like a traditional salesperson. can you kinda walk me through your past and how these skill sets have led you to where you are now?
Sam Gardner: Yeah. This is like a little contrarian, but I've kind of taken my like twenties to explore like what I've wanted to do. So I've tried out a lot of different roles. was briefly sales at Ramp before spending most of my time on the risk team. where I like was able to work with technical teams. I moved to Salesforce to be on their sales team. I started my own company, which like you have to do everything for. and then at Stainless I like Doing that mix of biz ops and product ops. I think that journey is picking up little and pieces. It was a very like breath first search kind of way of finding like what you want to do. ⁓ and it's like helped me speak. ⁓
Arjay McCandless: Yeah.
Sam Gardner: It's helped me do my go to market job at a much higher level. Like I'm able to build my own tooling ⁓ as a go to market person. I'm able to speak with technical teams, ⁓ like for prospects looking to buy Anthropic. so I'm able to just like do my job at like a much higher level because I've seen so much more of like different roles ⁓ worked with so many different people.
Arjay McCandless: How do you go about landing these jobs, ramp, sales force, stainless, anthropic? I mean, this is like an eleven star resume. ⁓ do you have any secret sauce for how you got into all these great companies? Especially when you're switching roles all the time and not a textbook candidate. ⁓
Sam Gardner: Yeah. Yeah. I don't know if I should like advise switching roles that frequently. I would just like LinkedIn is like a place, you know, like meet people on LinkedIn or reach out to people on LinkedIn. ⁓ going down the traditional, like if are applying to different places, don't go down applying through their website. if are applying to different places, don't go down like the ⁓ applying through their website. I see who's recruiting for that role and message them directly on like a simple LinkedIn message on just that you notice a role. so for stainless for I think I reached out to the CEO on like that he was hiring.
Arjay McCandless: Nice. And you know, what what do you say to him? Hey, I'm running my own company, but I like what you guys are doing. Like, what does this look like?
Sam Gardner: Yeah, something like that. I think it's really important, like the why. Like why are you interested in the role? And I always try to think, even before I start like looking at a new company, I'm always thinking like why do I want to work at this place? or what do I want to do? Actually, before I like even think about the company, what exactly do I want to do? And then matching that to like which companies I'm to. And it it gives a good interview process. Like I know exactly why I'm in this interview. I know what I want to do here.
Arjay McCandless: Yeah. This is funny because, you know, this is the second episode and the last guest said the exact same thing. Like, do not apply with your resume on their website without doing anything else. And I posted a clip of it and it was like people were like, I don't know why I would spend more than three minutes, you know, trying to get into a company when they're just gonna auto-reject me with AI. And I think the point that everyone's making, which I think you're making too, and I agree with, is like You don't want to just put your resume in the pile with everybody else. You have to have some other contact into the company because yes, if you put in your resume in two minutes to apply, you're probably gonna get rejected in two minutes. So yeah.
Sam Gardner: Yes, you will. And everyone does. Like I would get rejected instantly.
Arjay McCandless: Yeah, even I get rejected inst like everyone I know who applies with their resume just gets rejected instantly. And sure you can send out a thousand applications and maybe get a couple hits. But so when you're doing this kind of like targeted cold outreach strategy, how many companies like do you have in mind? Was it, you know, stainless is my dream company or is it I'm looking for something, maybe pick five or six, shoot a message, you know?
Sam Gardner: I don't remember exactly. It was pretty like tight list, maybe like ten or so companies. Stainless was definitely one of the if the top, just because of like how strong their engineering team was. I could tell very early on that they cared a lot about their product. and that mattered like a lot for me.
Arjay McCandless: And where are you looking to get this information? LinkedIn, X, company website, engineering blog? Like where exactly do you try to ascertain that about a company?
Sam Gardner: I think just like in doing research, like you like heard the Alex the CEO, talk on a podcast ⁓ and like to how he thought about the product and what the vision for the company was was very compelling to me. and so yeah, just like people people that care, people that are, you know, not trying to build a company just to make money, but you know, to make something that's actually valuable for others.
Arjay McCandless: Yeah, that seems very aligned with Anthropic's vision. So don't know if you can talk much about this, but during the acquisition, was there interview or was it kind of just like, Hey, welcome, you're on this team now? ⁓ Really? Wow. I assume not really able to talk about that much.
Sam Gardner: I wish it was I wish it was the latter. ⁓ yeah, we went through the full interview loop. I did like ten interviews. Yeah. Yeah. Um yeah, I mean it was just like it was a normal hiring process. ⁓ I didn't get like a free ticket onto Anthropic. but I did get like to skip maybe the introduction call otherwise.
Arjay McCandless: Yeah. Wow. That's actually pretty crazy 'cause, you know, from the outside looking in, it's like you get acquired. Congrats. You now work here. Did most of your company kind of join then?
Sam Gardner: Yeah. think it was like around sixty percent. Like Anthropic is very like picky on who they want. we didn't bring along like some very, very talented people. so yeah, if anyone's looking to hire a AEs or or engineers, like I know some fantastic people to work with.
Arjay McCandless: Awesome. I'll make a note of that. Do you think there's anything in particular you to stand out?
Sam Gardner: I think it was help like it it helped that I used the most like cloud code in the company. Like I was the most AI pilled. ⁓ and I had so many projects, personally at the company that I was building with AI. So I could just talk a lot about that. that that was probably what stood out. So I was like already ready for that interview before
Arjay McCandless: Yeah. Yeah. Person. Yeah. Yeah. Personally, what are you kind of building outside of work that you're most proud of right now?
Sam Gardner: The acquisition happened. my current project is strictly work focused and so I'm spending most of my time on that. in the past I've done some really cool projects. I like playing a lot of chess, so I made a camera that watches the chessboard and uses ML to track your moves. I made a Calci trading bot, I've stopped by the way.
Arjay McCandless: Ha
Sam Gardner: That was an interesting one where I was like, I'm just gonna put a hundred dollars into this. And I strictly wanted to see if I could use AI to to build something that I knew nothing about. Like I know nothing about weather trading. And so I was curious how far the intelligence would take me. And that actually didn't work that well. And it it approved an interesting point, or I don't know, approved, but it was one data point that you can't just like go into a field you have no idea about and and try to win it. you need to actually have expertise and use AI as a tool, not like a crutch. you need to actually have expertise and use AI as a tool, not like a crutch.
Arjay McCandless: I completely agree with you. I think the best use case for AI right now, and is maybe possibly the only one in my opinion, is being very good at something that you already are good at. Because I've tried to go zero to one on things that I have no idea how to do and it is not helpful. mean, and maybe maybe I'm doing it wrong. maybe what I should be saying is teach me how to do XYZ, not do XYZ. But when I say do XYZ and I get something back, I'm like, okay. You know, like I'm yeah.
Sam Gardner: Yeah, you have to trust it.
Arjay McCandless: Like and it's this is not the perfect example because everyone talks about taste, but I have zero design sense. And when I go to design a website or something, it'll make a website. And I'm like, Looks like a website to me. like, this is a website. And but I have no idea, you know. So I I'm often finding myself texting a designer, a friend and being like, Yo, is this good? And they're like, No. it's just okay, my bad. yeah, that's that's interesting. So for your Chess ML project, how ⁓ How much like are you writing, you neural nets or is it more like import, C V, whatever?
Sam Gardner: No, yeah, it's just like taking stuff off the shelf, open projects, maybe finding a data set online and like stitching everything together. that one was pushing the boundaries of what I knew. so I was trying to learn and and push the boundaries as well. Yeah.
Arjay McCandless: I think it's a good, like interesting point there where people feel like to be building AI projects, you have to be, you know, fine tuning models or building things from zero. And I just I think there's a ton of value to be built and captured on like the application layer of AI. Like you can import these APIs and do like ridiculously cool stuff. and those are very valid AI product projects that companies want to see on your resume man. They want to see the open sourcing and building. And I think People get like they don't think that's as valuable as it is, I guess. So it's nice to kind of hear you say that's the kind of projects you're working on without being super deep in the weeds.
Sam Gardner: Yeah, and I just like building. Like it's just something that if I have a free weekend, I'll maybe tinker around for a little while. so yeah, I mean I would just like do the smallest project you can think of and just and work on it. Yeah.
Arjay McCandless: do you do anything to these projects into the wild? Like are you open sourcing them, posting them on LinkedIn, X, telling your friends about them, kind of just putting them on your resume?
Sam Gardner: don't even know if I put on my resume. I need to do a much better job of like I strictly do these projects because I think they're fun or useful. I posted the chess one on LinkedIn because I just thought it was cool and like people also thought it was cool. But no one was reaching out to me and like, hey, this is really cool. You should ⁓ yeah, I'm gonna hire ⁓ so it's strictly like was doing it for fun and and I open sourced like the chess ML project.
Arjay McCandless: Yeah. Yeah, I'm gonna hire you. Mm-hmm. Nice. I always tell people when you're picking projects, just pick something that you want to do, because that's how you're gonna finish it and you're gonna learn all the important skills without forcing yourself to build some thing that you're not interested or gonna give up on after a couple hours. That's cool. What do you think it's gonna take for you to really crush at anthropic?
Sam Gardner: That's a good question. It's difficult 'cause like the bar here is really, really high. Like the talent is incredible.
Arjay McCandless: Yeah.
Sam Gardner: Frankly, like I don't know right answer to that. And I'm pr I'm new enough where I'm still trying to get my bearings. I think what I'm really trying to do is just be like helpful and kind. you know, like when you when you don't know what to do or how to perform, it's just like be helpful wherever you can. that's kind of the mindset I'm going into this with. Like work hard and and be kind.
Arjay McCandless: Yeah. Yeah, yeah. Yeah. Have you been able to ascertain much about culture? is it nine nine seven or and are people helpful and you know?
Sam Gardner: People are really helpful. The culture is like fantastic here. And that's like what they're like really hiring for. is like a lot of work. Like, you know, I was up till like eleven or midnight last night, like just doing email. and so people work really hard here, because like the cause is so important, and people really do care about it. that being said, is like everyone's willing to help you. and everyone's very kind, everyone's
Arjay McCandless: Awesome.
Sam Gardner: very cognizant that everyone else is busy. yeah. ⁓
Arjay McCandless: Yeah. How do you feel about AI as a job creator versus taker and kind of where does like time saving fall into that? Because In some sense, I think to myself, shouldn't we be able to solve like responding to incoming emails or make that ten times faster and free you up to do something else? And kind of like how do you reconcile that with what work you're doing and, you know, where you're at?
Sam Gardner: I kind of see like two different camps. see instances where like I work so much harder. Like I love like what I'm able to do with AI and I'm able to like push the boundaries of what I'm capable of and what AI is capable of. and so I'm definitely working more and I think other people here are working more and they use AI as a tool. what scares me a little bit is like when you use AI as a crutch. and I feel like there are a lot of jobs that can get automated away.
Arjay McCandless: Yeah.
Sam Gardner: Anthropic like shy away from that. Like we're fully aware that our technology will automate a lot of jobs away ⁓ in America and the world. I think that will ⁓ I think over the long run, like that like people will find new jobs, but there definitely is gonna be like a tumultuous period in in between. What scares me more or like I'm more concerned of is like cognitive offload where people don't use the technology as a tool, but ⁓ like students, for example, they just like do all their homework with AI. And I think that's like actually very hard to solve ⁓ and it's gonna be very destructive.
Arjay McCandless: Do you take intentional steps to make yourself think through problems or intentionally do something without AI ever just to make sure you're keeping those muscles sharp? Or how do you handle that?
Sam Gardner: So like I'm a less technical person, so I'm not writing code, but like writing is a good example. If it's like a utility writing, I'm usually writing it with code. ⁓ excuse me, with AI. but ⁓ like Paul Graham like describes it pretty well as like writing is thinking. ⁓ if you're writing something if you're thinking through something important, it's like really important to like write it down. And AI won't be able to replace that process. It's like Taking a very nebulous thought and distilling it into like the simplest form it can be. so in those instances I try to write it myself. and it's difficult in that like that muscle can atrophy and I it definitely has been atrophying. so it is important to like in my instance continue to write. In engineers' instances, like continue to code like very important parts. I I'm kind of speaking out of my but like I ⁓ I assume engineers should do that.
Arjay McCandless: Yeah, it's interesting. Like I've definitely become significantly worse at coding by hand. I mean, if you if you had me like go one on one against myself two years ago, it would not even be close. Like night and day. But at the same time, my like output is much higher. you know, like my marginal return on time is much, much greater. Like with AI versus without it. And yeah, it's a question of I think this is something a lot of engineers grapple with too, is like
Sam Gardner: Yeah.
Arjay McCandless: Okay, the current models, let's just say write code that's 80% good as a pretty good engineer. And if that code is 80% as good and we write our entire code base like that, then you let's say Fable comes back ⁓ and can write code 90% as good as an engineer, then I can just say, fix the slop in my code base, make it a little bit better. Then we get another model that's even better and it does it again, and then eventually, like in the long horizon. It never made sense for me to be like manually checking our code now, anyways. I don't know. It's like an existential thought that I think a lot of people are having across industries.
Sam Gardner: I think and and you know better than than me on this, so correct me if I'm wrong, but a a lot of code is like utilitarian ⁓ and that it's just like serving a purpose to get a job done. but like I'm working on projects now, so much of work is like slowing down and not AI to like think through like how this like specific button should function.
Arjay McCandless: Yeah.
Sam Gardner: And maybe like the end state is like you're still using AI to to like build that that button. but again like if you're just cooking with AI and having it build everything, then like you're not thinking about getting your product like a much higher level.
Arjay McCandless: Yeah, okay. This could be a good closing thought here. I guess what is your take on when to use AI and when not to use AI? You kind of said it there, but I'm I want to hear you say it a little more like explicitly in like a kind of I don't know, repeatable way. And obviously there's not a correct answer to this. I'm just curious, like your take on it.
Sam Gardner: This is like a really hard question. ⁓ I think for for outside of coding, it's what I explained before in like if it's really important and needs to be thought through in like in terms of writing, then I'm definitely just writing it by hand and not using AI because like I need to d work through the process of thinking. With code, I think of it more as like almost like driving somewhere that I I don't know. the exact destination, but I know it's like far away and it's in this general direction. And AI can like put you on the highway and like get you in the general vicinity. And then you need to slow down. And I don't know if it's by hand. It might still be by code, but you need to be much more surgical. And being surgical with AI can get a little difficult unless you're like working in in the term in the like IDE. and then you're really just you know, the actual code or yeah, it's like a lot lot a lot less AI.
Arjay McCandless: Knowledge work, how about making a presentation, sending an email, stuff that.
Sam Gardner: Again, it depends. Like if this is for creating a presentation, if this is like the one presentation that matters, you like generate the first draft in AI, and then you're thinking about I think about like, all right, what am I trying to say? What are the exact words that are going on here? Like, I'm not using AI for that. I'm like thinking through what is my purpose with this presentation? Am I conveying that and being very surgical? In an email, if it's like an important client, it's like the same thing. It's like getting a rough draft and then like edit editing.
Arjay McCandless: Yeah. I I agree with everything you said. I think one of my biggest issues I've had with AI in human to human communication, things like an email, even a LinkedIn post or a presentation is you get too many words. Like I'm very, very about slide decks in particular. If I see more than fifteen words on a slide, it's like I'm not reading all that. You know, like And I think a lot of people too, even if I'm reading a report from somebody or something that somebody sent me and it's like three pages and it should have been three sentences, like not happy. You know, I don't want to spend my time like I would have rather have read the prompts that you gave to get this than read this. Cause this is just like downstream of that. So ⁓ yeah, don't know. That's interesting. I think there's a time and a place. I had a had an old coworker at a like a couple companies ago, and I saw him
Sam Gardner: Yes.
Arjay McCandless: Write a message to Chat GPT and then copy paste the message directly from Chat GPT to me. And it's just like a personal message, like asking for help with something. And I was like, dude, like we don't need this. You know? And I I think like just people don't fall too deep into it. We're still humans, you know? Yeah.
Sam Gardner: Yeah, yeah, yeah. Like I mean, I would recommend if it's your voice, like trying to stay away from AI. ⁓ ⁓ yeah, AI Slack messages are just like not fun to read and they're too verbose. ⁓ yeah, even emailing and go to market, like sometimes you know they're using AI and you're using AI and you're trying to convey like information, so it's helpful to use AI, but it's just like not this is
Arjay McCandless: Yeah. Yes. Mm-hmm.
Sam Gardner: This is not right.
Arjay McCandless: Yeah. Yeah. A hundred percent. All right. It was awesome having you on. Thank you for sharing everything you have.
Sam Gardner: RJ, thank you. This is this has been super fun.
Arjay McCandless: Is there anything you want to share with people, like anywhere they should follow you, anything they should check out of yours?
Sam Gardner: don't know, feel free to like check out my LinkedIn. ⁓ I have like a yeah, I check out my chess project. I have a personal website that kinda like talks about who I am as a person. ⁓ that's on my LinkedIn as well. I always like like showing that off.
Arjay McCandless: Nice. Check out your chest project. Awesome. All right. Thank you for coming on. We'll see ya.
Sam Gardner: Thanks, Arja.
