AI-pilled Go - Live from San Francisco with James Heller, and interviewing with Paul Querna | Sponsored by C1.ai
This is Cup a Go for 08/26/2026. Keep up to date with the important happenings in the Go community in about twenty minutes per week. I'm Shay Nehmad.
James Heller:And I'm James Heller.
Shay Nehmad:And we're recording this episode in beautiful sunny San Francisco in C1's offices. We'll hear more from our beautiful sponsors in just a bit. First of
James Heller:all, James, how are you? I'm doing well. Thanks. I'm very excited to be here, my first time on the show.
Shay Nehmad:We really appreciate you coming. If anybody's, like, missing John, don't worry. He'll come back on soon. Today, we're gonna do a slightly shorter news segment, and I have a thing I wanted to bring up first, but it's gonna be like an AI pilled episode. We are recording in San Francisco.
Shay Nehmad:It is required by law to speak about AI at least three times an hour here. My first thing I wanted to share is a blog post from the Go team, came out about a month ago called Why Go Is An Ideal Language for AI Assisted Software Engineering. And the first thing to note is that it's not on the Go blog. You know what blog I'm talking to? Like, the Go Dev blog?
Shay Nehmad:What's your favorite post from there?
James Heller:Oh, gosh. I mean, so many, but I'm I'm a big fan anytime they're talking about performance, any sort of performance post. That one really gets me Bolgian, entertained
Shay Nehmad:and interested. So this post is from the Go team. It's like Cameron he's the group manager, and Richard, and like chief evangelist, and all those people. The crux, I recommend you read it. It's actually not that long.
Shay Nehmad:I think now the vogue is to have your agent read it for you and summarize, but we still know how to read, I think, some of us. It's all about why, A, choosing what language you choose to develop in is important in AI development, which is already controversial. Just before we heard in the talk in this meetup, someone like, Oh yeah, I write Rust, I don't really like it, but my agents do it for me, so it doesn't matter. So, A, they are saying the language you choose is actually important, and b, maybe not surprising, they're saying Go is a really good language for agents. So I wanted to ask you before we dive into why they're saying Go is good for agents, does the language even matter?
Shay Nehmad:Aren't we all just programming in English, just writing prompts, code doesn't matter, you can regenerate everything?
James Heller:I mean, that's that's a really interesting question. I it's hard, I think, to say that there's a definitive answer, but I think we all probably have somewhere that we lean depending on what it is. I think for me personally, I am still a
James Heller:you know, a year ago, and it's probably gonna keep going that way in the future. But for me, having that fallback where I can can look at it and still try to figure out what's going on and, you know, have a chance of being able to stumble through it, I think that's still huge, Go really does benefit from that with the readability and keeping things simple in the language.
Shay Nehmad:Yeah, so the point, I think, of post is agreeing with you that we used to care a lot about how writing code feels and how hard it is, and now we review a lot more. You have this like AI teammate, they call it. I think this post, you know, it's coming from Google. It's in the developer Google blog, so it's way more corporate. Of course, they would wanna push you to use more tokens and Gemini and their whatever, they just bought Windsurf, so it makes a lot of sense for them to say, of course, everybody has an AI teammate and a claw and a whatever, and that your work has become more like, let me review the code and do the architecture and things like that, and that Go is a good fit for that for a few reasons that they list that I think you'd probably agree with.
Shay Nehmad:The first one is that Go is not just for like coding, it's for software engineering. It has a lot of opinions, it's very simple,
Shay Nehmad:has few keywords, it has a strong community, an ecosystem and security built in and stuff like that. I guess you'd agree with that, right? Oh yeah, absolutely. Then you'd have Go as a platform where everything is like batteries included, right? You have fuzzing, you have concurrency, you have all these things that in many other languages are just not built in or bolted on after many, many years or stuff like that.
Shay Nehmad:I think that's pretty easy to agree with, like Go test is part of Go, so that's easy, and turns out what they're saying is this was good for humans and turns out that agents and humans have very similar needs when it comes to developer experience. These ones I think are pretty easy to agree with, and then they're coming in with like the opinions of like Go is more readable than other languages. I'm just putting this on the table. Is Go more readable than other languages?
James Heller:I would say yes, at least for me. I think the the language itself is simple. It has fewer keywords than a lot of other languages. There there aren't as many ways to do the same thing as you might find in other languages. And so I think that, you know, in general, that that does make it more readable just because there is less of a chance that you'll be surprised by some syntax you haven't seen before.
Shay Nehmad:Yeah. I'm not surprised at all when I see, like, if error is not nil, like it's
James Heller:And you see
Shay Nehmad:that a lot. Yeah. One funny thing, I don't know if it's still true or relevant, but three ish years ago, maybe even more now, there was a talk by Natalie Bistunovich, I hope I'm saying her name right, from Go Time, like the previous Go, Big Go podcast in Go for Con, Go for Con Israel, where she said Go is great for AI. This was like really early, and one thing there was all the formatting looks the same, and unlike Python that uses spaces, it uses tabs, so it ends up being fewer tokens, which is just like dumb luck. I don't know if it's actually true or if it matters with the context windows we have today, but back then, people were like, Oh my God, I can hold a whole file in my conversation.
Shay Nehmad:And with Go, you can hold many more files because it uses tabs. It's like four times less tokens, which is just a funny tidbit. I, of course, agree that Go is very readable in my opinion just because you don't have magic, you don't have exceptions, things can't be like rubber duck replaced under your thing. There's a talk that I remember off the top of my head from Peter Boggan, where he says like, one of the things I like about Go is there's no magic. You read the lines and the lines that you read are what's gonna run on the machine.
Shay Nehmad:That's cool. Another one is Go is reliable. What do you feel about that one? I think that
James Heller:it is. I mean, there's always the naysayer because there's the runtime, and so that introduces variability into the running program. But I think for most of our applications, that's not really a major impact on what we're trying to do, especially if you're, you know, building a web service.
Shay Nehmad:I don't know about this one. I think it really depends on how you use it. Like, I've definitely seen unreliable Go applications, but, you know, they're like, well, you have GoVonCheck and you have fuzzing and you have all these things that can help you make Go code pretty reliable pretty easily. They also say Go is more maintainable on average than other languages. Like, after day one, it's easier to maintain it, which to me, the only thing that holds true is it's gonna like the Go one compatibility thing, like it's not gonna break if you upgrade.
Shay Nehmad:Other than that, it's totally up to how you use it because you can make like unmaintainable software. I do it all the time, and I think they're trying to push people to use Go with AI, but I don't fully understand what's the angle. Like, who's not picking Go because it's good for web services? Like, what's their angle here in your opinion? Is it just get people to use more AI because some of them will use tokens from Google?
Shay Nehmad:Is it get more people on Go because they wanna increase the language adoption? Is it just look at the cool characteristics of the language without any, like, hidden motive?
James Heller:Yeah. I'm I'm sure they're looking at a, you know, a circular marketing flywheel here of, you know, Go came out of Google and Google has AI with Gemini that you can use. And, you know, it all kinda circles back into staying inside of their ecosystem. So I'm I'm sure that was a factor in driving it, but I I think that this isn't the first time that we've heard that Go is good with AI. And even at Gopher Con, think the last two or three Gopher Cons, we've heard somebody up on stage say similar things along those lines.
James Heller:And so my guess is that, you know, they are pushing a narrative, but I think it's a narrative that already has some grassroots growth here. And so, you know, it it makes sense for them to keep pushing it.
Shay Nehmad:With that in mind, like, Go is good with AI, the next blog post we wanted to discuss is from Filippo Valvassori, friend of the show.
Filippo Valsorda:Hi, everyone. So I'm Filippo Valsorda. I've been maintaining the Go cryptography standard library since 2018. I've done that first at Google as the lead of the Go security team, and these days, I'm doing it as a independent open source maintainer.
Shay Nehmad:So you had a blog post you wanted to talk about, which is also AI, but, like, AI, the bad parts.
James Heller:It it hones in on the idea that, you know, with with AI and with these LLMs, it's getting much, much easier to produce these these reports. And so, you know, much like we've seen huge volumes of PRs, I think we all probably saw the recent GitHub incident report where, you know, basically, they're on huge growth in terms of usage, and it's largely driven by by AI opening lots of pull requests. And similarly, you know, there are a lot of of, you know, incident reports that are going out and, you know, the same idea in that people are inundated. And so security teams are are suffering just like anybody else might be. And, you know, he he's very much interested in security and has spent much a huge amount of time with the the Go team in security, and so he definitely has a great perspective here.
Shay Nehmad:Yeah. The I think the crux of the of the post is when you I really like the opening sentence where he says, like, when you work in a public open source thing like Google, like Go, I mean, issue or every pull request is a present, not an obligation. Like, don't actually have to address every issue. You don't actually have to merge every PR. You don't have to handle everybody who's trying to contribute to your project.
Shay Nehmad:It's more like a present. You can decide to ignore it. It used to be the case, at least according to Filippo, that vulnerability reports weren't They had like a different echelon. Like when you got a vulnerability report, you had to treat it. It was like part of your responsibility for users, and now with LLMs, it's not that case anymore.
Shay Nehmad:He's like, Ah, it's easier for me to just generate the And I'll probably do a more accurate job of like finding these myself. So vulnerability reports are not special anymore. It used to be case that the security researchers did like a very specialized work that only they could do and it was very important. Now it's not that anymore. To me, this is like another case in the big folder called Slop is ruining everybody's life, a novella or whatever, I don't know, in my company or in any like dev org that I hear, like people are used to bring like an issue, but now they'll give you a pull request.
Shay Nehmad:Five years ago, if someone gave me a pull request, was like, oh my God, they wrote all this code, they went through all this effort, I have to respect it. Now, like I can point my claw at the same originating thing and generate the same code, so it doesn't carry the same weight. I think this is the same thing, but just for vulnerability reports. What I'm wondering is what real important vulnerability reports are getting lost in the noise, or how will you Like now that it's 2026 and it's not true anymore that these are special, how do you find the special ones from, like, real security researchers that have found something special?
James Heller:Yeah. I mean, that that's just a hard problem, and I hope that we get better at finding the the signal in that noise, but it you know, it's it's certainly something that we're going to have to struggle through collectively as as we're all working towards these things. You know, it's it's one of those kind of secondary effects that I don't know if well, I'm sure lots of people saw this coming. I I didn't see them the issue until, you know, it's here knocking on our door. But these are things that are are hard to plan for, and and it's it's hard to prepare ahead of time.
James Heller:And now it's here. And so, you know, we're we're seeing people who are very well qualified at handling these things, but they they are themselves kind of, you know, scratching their head a little bit thinking, you know, what what is the future going to look like as we tackle these problems?
Shay Nehmad:It's interesting. One thing you write is like, if there's already an existing trust relationship, like if there's someone I already know who's producing high quality reports, then I'm gonna address them on a different level, which is exactly the same problem that everybody who's a junior right now trying to get into the industry is facing, that people who like the rich get richer sort of situation, people who already have cred are gonna keep having cred, people who don't have cred are like, ah, you did it with LLM, I can do the same thing. I don't know when they break, like, that glass ceiling, I guess, is the word I'm looking for. But this is another interesting blog post from the other side of, like, AI on end go. The final thing is a release I've found.
Shay Nehmad:I've been doing a ton of MCP recently. What's what's your favorite MCP server, James? I don't know if you have one.
James Heller:Don't have one.
Shay Nehmad:Turns out they're all just wrappers around APIs anyway, this was a huge mistake, an anthropic SIOP, but whatever. In San Francisco, these are called just like this is a religion, so you can't. But we, I wanted to talk about a specific MCP server that I actually use, which is the GitHub MCP server. I don't like using this MCP server because I can just tell my agent to use the GHC align, that usually works pretty well, but for a few remote agents that I've set up, I don't really want it to have the GA because it doesn't have the auth login, I want it to use an API token with a specific thing. It's just easier for me that way.
Shay Nehmad:Also, I have to do it for my job where I look at MCPs and try to break them from a security perspective, but it's interesting. Do you remember, like, four months ago, five months ago ish, what was the zeitgeist around tokens and token pricing? It was like, everybody get max subsidized, and it was all like you didn't worry a lot about how many tokens you were spending. Right?
James Heller:What about now? Well, now we're all paying a lot more attention. And they're they're also, when you're looking at people using agents, skills are, you know, the the new hotness in in new I put in quotes because it's not new anymore. But, you know, there there was that kind of contention of, hey, MCP can bloat your context, which I I think, you know, this this release here does a lot to address it.
Shay Nehmad:Yeah. So I I really like this release where they it's clearly, in my opinion, clearly a response to tokenomics. Like, people are like, I'm asking for a release or something or a commit or a branch just because I need a tag or just I need this, like, one piece of detail to get this AI workflow working, but when the response came back from the tool, it included all the fields from that specific response, So now the MCP server has a parameter called fields, well, sort of like, what's that thing called, a GraphQL, where you tell it, like, which things you need when you get it back. So you, like, filter the response dynamically. You're like saying, well, please get me the latest release, but out of that latest release, just tell me who are the contributors.
Shay Nehmad:I don't care about the notes, I don't care about the title, I just wanna know who are the new contributors in this release. So when you call the tool, you shrink the tool response size and you optimize context usage. I thought this was interesting, like a whole new release. They release this pretty often, and this is a Go project. It's actually a pretty good one to read if you're looking into like how to build MCP servers.
Shay Nehmad:I've read the code, it looks pretty good, but just like a whole release all around that's basically all around tokenomics. Like, the only thing this thing does is save money. So I thought this was another interesting AI related release, and I think that does it for the news segment for this episode. Stick around after the ad break to hear an interesting interview with C1's CTO. And thank you, James, for cohosting with me.
Shay Nehmad:Where can people find you online?
James Heller:Oh, you can find me you can find me on LinkedIn. And if you look at the Apartment 304 website, you can find me through that too.
Shay Nehmad:Thanks all for coming on the episode. Let's give it up for James. Welcome to our Ad Break. This episode is sponsored by C1 dot ai, and who's better to talk about it than the person who made this meetup actually happen, Blake. Blake, why don't you tell our beautiful listeners and also the people in this room what is all about?
Shay Nehmad:And if there are Go developers listening, which I hope they are because this is a Go podcast, what do they wanna do with y'all?
C1.Ai:Yeah. So hello, everybody. We are C1AI. We securely connect humans and agents to data and different systems. If you're a Go developer, we're hiring.
C1.Ai:So and not only so if you found Paul's talk interesting and that's something you wanna do with your, I don't know, life or years or whatever, you should reach out to us. You can look at c1.ai, or you can also email me directly, Blake@c1.ai or blake haggardy on LinkedIn or seven zero seven food on Instagram or no. Okay. Anyways, I think the other big plug is kind of exciting coming right up here. We're hosting our first conference in a month.
C1.Ai:So October 6, it's gonna be at the Clancy Hotel, which is just right across the street over there. There'll be some events going on here, events going on there. Gonna have some pretty awesome speakers like Travis McPeak, who's the head of security for Cursor. Carl McGinnis, he's the ex chief product officer for Okta. You know, senior staff engineers from Brex are gonna be speaking.
C1.Ai:It'll be rad. You should go. It's free to attend. And if you want to, you can scan that QR code right over there on that table, the little tabletop thing right there, and register or go to Partifen or Luma or any one of those things as well.
Shay Nehmad:Thank you, Blake. Other than that, we wanna thank our beautiful Patreon supporters. This show is supported by you, as we mentioned at the top of the show. To support the show, you can find all the links at cupago.dev. Due to popular demand, which means one guy, we introduced a new tier on our Patreon, which is even more expensive.
Shay Nehmad:So if you're swimming with money and you're just looking where to put it, you can join our new board member tier level on Patreon, which is like $20 a month, which is more than I'm willing to pay for any sort of content, so I don't know who wants to pay that. But if you like Cup a Go, you can do that. Also, our swag store is going out of business pretty soon just because Wix is too expensive, so unless they're gonna sponsor us in the future in case Wix is awesome. But yeah, we're gonna take that down and replace it with, like, way more complex email based, like you actually email us and we package it and send it to you, which is gonna be a lot worse. So if you wanna grab stuff, just do that before we actually get to that to do.
Shay Nehmad:It'll probably take us a month, we're very lazy. And if you wanna follow-up and join, you know, the community discussions or whatever, our Slack channel, A Cup A Go on the gopher Slack. That's it for the Ad Break. Stick around for an interesting interview with Paul from C1AI. Well, I'm recording this episode live here in beautiful San Francisco in C1's offices, but I'm not seeing anyone, C1, hiring as hard as c one.
Shay Nehmad:So I have Paul from c oh, sorry. That was really bad. Usually, Jonathan is on these on these intros to, like, curb these bad jokes, but I'm I'm on my own this time. So everybody give it up to Paul, who's gonna tell us a bit more about himself and C1.
Paul Querna:Yeah, thank you. Yeah, I'm Paul Querna. I'm the CTO, one of the co founders at C1. I've been a Golang developer from, I don't know, twenty twelve, twenty thirteen era. I love Go.
Paul Querna:I don't know. We built our entire company around Go, and we use a lot of it. And, you know, we're doing stuff like everyone, but really focused on kind of agentic identity, agent security, how agents talk to different systems. And Go is great at that because we're always moving data around, proxying data, doing cryptography, things that Go is great at.
Shay Nehmad:I'm not gonna bury the lead. We're gonna talk about how is hiring, not because C1 is sponsoring this episode, which they do, and that's very nice. Thank you for hosting us. But because we had an interesting discussion before the meetup started on how have you changed your interviewing practices from, let's say, total pre AI era five years ago to now, and I think it's a topic that a lot of our listeners are actually interested in, and this is a topic that comes up on the Slack every now and then anyway of like, oh, I was interviewed and I was not allowed to use AI, or I'm trying to build an interview and I want people to use AI, or whatever. So if you could walk us through what's the difference, assuming, you know, five years ago you had a coding exercise as some part of the hiring process.
Shay Nehmad:How has that changed? Because people who are here on the meetup know that this guy is totally AI pilled, so it would be interesting to hear from you, especially that the company is actively hiring, again, in their beautiful, beautiful offices in San Francisco.
Paul Querna:Yeah, so I think when you think about hiring, you have to go back to the first principles around hiring. So I think of it into kind of two extremes. So one is you want to filter out bad people. Not that people are bad inherently, many people are great, but you don't wanna have a filter function that you're trying to identify reasons someone shouldn't be here. But that is like the secondary concern, honestly.
Paul Querna:The primary concern is you wanna figure out, is there a reason someone should be here? That you'd be ecstatic to have them on your team. And so when I think about that, yeah, I used to run coding interviews, whiteboard interviews, all kinds of ways to try to figure out, am I excited to work with someone? And what we've changed to in the last year is we do effectively a Zoom based or in person pair coding using generally Claude code or another harness, whatever person's most comfortable with. And we'll actually give the candidate a Claude Max API key for the interview.
Paul Querna:Because, you know, if you're personally running these agents, you don't necessarily wanna spend $50 to come to an interview at C1. So we try to make that easy and not a big deal. And then it's really about problem solving. It's back to, like, what I would call, like, one of kind of whiteboard style stuff where you're trying to break down a problem, figure out how to do it. The agent's there to assist you.
Paul Querna:We're not trying to stump you on how to build a B tree. We're trying to figure out, like, if you need a B tree, can you help the agent figure that out? And that's really how we think about the problem is, like, collaboration with the agent, and we're trying to test for, it's like a test, is this person who is someone who is understanding agents, how do you get leverage out of them, and are they someone that has just already figured out the basics and is already kind of living in this future state of leveraging agents to the maximum?
Shay Nehmad:Like, I get that, but just to play the and not to saying that I disagree, I'm just like playing devil's advocate because I'm the interviewee, I'm the interviewer. Wouldn't you like to know if the person you're interviewing knows the shortcut of how to open their terminal in their IDE or knows how to set a breakpoint and reach it, or like knows how to add a log themselves, like even if they can analyze the high level problem, just being able to like read a JSON log themselves and like pipe it to JQ and find a thing, not because agents can't do it anymore, but just because if you know how to do some of these, like, I'm picking on troubleshooting specifically just because I had a very troubleshooting day, but some interviewees that will come to you legitimately don't know how to set a breakpoint and reach it because they've never done it. And I don't know, five years ago, to me, that was a negative signal. Like you've never debugged the thing before, someone else on the team helped you, or you just added a million logs and everything took you a long time. Does that like not matter anymore if Cloud Code can do it for you?
Paul Querna:So my perspective, and this is someone who started out as a C programmer in my career. So I started out doing C programming and then everyone started writing PHP. And I was like, oh, PHP is terrible. What if they don't know how to open the assembly and read it? And then everyone got into Python, and then eventually Node.
Paul Querna:Js came around. I was a contributor to Node. Js, I really liked it. And eventually Go, obviously. The difference is, yes, I can read Assembly.
Paul Querna:I don't anymore. I think, you know, when I think about the new generation developers, if you're in college today learning about computers, you're not gonna learn assembly. You're not gonna learn how BIOS works. You're not usually you're not gonna compile your own Linux kernel. Like, at some point, these are low level concepts.
Paul Querna:And the only difference with AI is that it's just happening really, really fast. You know? Like, the transition speed is extreme, but I would have said the same thing, you know, ten or fifteen years ago if I was interviewing someone for a C programmer and they didn't know how to do these things. I'd be like, you cannot be a C programmer. But you still could be a darn good PHP programmer or a pretty good TypeScript programmer without knowing C.
Paul Querna:And so I think we will evolve into a new level of abstraction that is higher, that has a different skill set, and someone starting their career today will never learn how to use a debugger. And I think the world will be okay. It might take time, and we're in this transition moment. But the difference is the transition is happening really, really fast. So at the end of the day, I'm trying to think for first principles, but am I excited to work with this person more than I'm trying to figure out, like, do they know how to use debugger?
Paul Querna:I think AI will be
Shay Nehmad:fine at debuggers. And specifically with Go, because very relevant to listeners of this show and people in this room in the meetup right now, T1 is hiring for Go developers specifically right now, right?
Paul Querna:Yes.
Shay Nehmad:Yes. Humans, not agents. Specifically for Go, there are some tricks that I feel, and maybe you shared the same thing, that because you've known Go for a while, you know. Like, you know that the race detector, right? And someone who is strictly, you know, they don't know Go, but they're pretty good with like Claude Foo or whatever, they would know to ask the agent like, oh, can you bring in fuzzing, or can you bring in like, you know, profile guided optimization just because they don't listen to the show?
Shay Nehmad:In the interview, do you bridge that gap by giving them like a hint, like saying, hey, this is gonna use Go. You should probably brush up on Go concepts or whatever, or is it completely abstract that they can solve it in whatever language you want? Not to give people too much prep if they are actually gonna interview. I'm just interested in, like, how you've built it.
Paul Querna:The actual interview, we don't really prescribe a programming language. A lot of people do it in Go. A lot of people do it in TypeScript. We we tell them our code base is mostly Go. So hopefully, you like Go if you wanna work at C1.
Paul Querna:But I think the the skill thing is very real. I you know, we have a class of interns. We try to tell them about weird computer science stuff and things we do in the business that maybe our more senior developers know. I think some of it we encode in markdown files, in agents markdown and skills files and things that more senior developers might build. I don't know how this evolves a couple years from now as the workforce turnover.
Paul Querna:I mean, okay, years, hopefully, before I turn over. But as people retire and change and they're gonna have a different workforce, yeah, they're just never gonna have those skills. Maybe they'll get lucky and encounter it. Yeah, PGO is a great example. That's a thing that I learned when I was writing C, and I knew that it happens in Go.
Paul Querna:And, like, I know we use it. It's great. But, like, when will you encounter that if you never have? And so topic discovery, think, is one of those unsolved problems. I think in large existing companies, you
Filippo Valsorda:will
Paul Querna:have documentation, markdown files, agent markdown files that help bridge that and produce kind of knowledge graphs that agents can consume. But if you're starting a new company with young people that have never worked at a large business before, it's gonna be a weird code base. But maybe the business outcome is the same. I think that's the other thing you have to remember is at the end the day, we are building software generally. I mean, I have a love of the game software on my GitHub, but I generally work to make money.
Paul Querna:And so I think that's the other thing is the business outcomes are as we're gonna drive behavior at the end of the day. So if the business doesn't need PGO experts, they won't have any.
Shay Nehmad:That makes a ton of sense. I assume these interviews have been going for a while. Have you been getting feedback from interviewees that this is like, Oh, this was really out there. I enjoyed it. I hated it.
Shay Nehmad:Like, all of our When we change hiring processes, which is something we're doing right now as well at Opsin, which is where I work, we always try to optimize for hiring the best people. Like you said, like they have to be a best fit and we have to be a best fit for them, etcetera, etcetera. But how do candidates feel about it? I assume, you know, some people learned Leetcode and grinded years away at Leetcode and you're still like, yeah, all that doesn't matter anymore versus some people who might be relieved, oh, finally I can focus on the good parts of whatever. Have you heard any feedback from the people who were trying to join C1 and went through this very new sort of process where they get an API key and get an agent to do their will?
Paul Querna:Yeah, so first of all, you have to, I think, set people's expectations in the hiring process. You have to be very clear with them. This is the interview structure. So I think we try to, we talk about our, we have a panel structure and this panel's gonna do the coding interview and here's how it's gonna So one is just education ahead of the interview so people know what they're getting into. You don't wanna show up and be like, Welcome to Zoom.
Paul Querna:Here's an API key. Good luck. So I think part of it is manage the expectations of the candidates, which I think we do a good job of. I think generally the feedback has been very positive. I think we do tell them, Hey guys, this is new.
Paul Querna:We've only been doing this for less than six months. We think it's a better way to evaluate, will you be successful at C1? But it's new to us. Help us figure it out. And that's also part about the, frankly, ambiguity of being in a startup is you're gonna be in ambiguous situations with new things.
Paul Querna:And either that you're okay with that, and that's part of being excited to be at C1, you're like, that's not really the gig I want. I don't want ambiguity in even how my interview process works. But having said that, I do think we're to set expectations. It's generally been really positive because I think it is also I think people I personally did not like the LeetCode era. It never excited me, and I never excited was excited trying to give those interviews.
Paul Querna:And so I'm thankful it's over, frankly, and I think a lot of people really are.
Shay Nehmad:Yeah, I hated those as well. Well, thanks for giving us this insight into, I don't know, I think a problem that basically every company is either facing or avoiding. I'm not surprised the feedback from your candidates is pretty positive. I've met some of the employee and whatever experienced people, and other than their hate of cars, they seem really nice. Finally, before we close out this interview, we have a stumper question we ask every interviewee.
Shay Nehmad:It changes every year. It used to be like, what's your favorite feature of Go? What's your least liked feature of Go? And then we got a complete list of all features in the language in both columns, and then we had, who was the person who inspired you, whatever. This year it's, what is your favorite third party library?
Shay Nehmad:So not standard library that you used recently or heard about or whatever. I use a lot
Paul Querna:of external libraries. I think the one that comes to mind is I think it's jackrpgx, which is kind of the, I think, the most popular kind of third party Postgres library. And what I love about it is it's just well factored. He's built out connection pooling separately. There's a whole the protocol layer's separate.
Paul Querna:It's just a well built Postgres library, and it's really fast. And we've generally loved working with it in our product. And so it just stands out as a very solid piece of software engineering. Obviously, it's a pre AI package. It was boutique.
Paul Querna:It was handwritten, but it was always well done. As someone who's like, you know, how does our connection pooling work when our API server runs out of connection pool? Like, it was just well written, well documented, well built. And the factory was always just great. Like, we actually took some of his connection pooling stuff and used it for non Postgres protocol stuff.
Paul Querna:So I just think it's a great little library, and it's one of those dependable things that you just don't really think about. And it's kind of better than the database SQL one that no one uses anymore. So it's great.
Shay Nehmad:I like well factored, because that means it doesn't need refactoring. Right? Cool. PGX. Well, thanks all for coming on the show.
Shay Nehmad:Where should people, now that they've heard all these good things about C1 and they wanna grab your API key to vibe code their own project and then goes the interview, where can people find, you know, all the links and also find you online, etcetera?
Paul Querna:Yeah. C1.ai. There's things in the footers like careers and links to our docs and all that kind of stuff. And then I am a p q u e r n a, Pcorna, on, I guess, X still. I don't really use LinkedIn much, so probably try X.
Shay Nehmad:All right. Thanks a lot for coming on the show. That wraps up our live episode for this week. Although we plan to record another episode this week, this is slowly becoming my part time job. One hundred twenty seven News, I know all the listeners have told us, one twenty seven came out, it's been two weeks, I don't know what's in it.
Shay Nehmad:Don't worry, we'll talk about it. Thanks a lot for listening. Program exited. Goodbye.
Filippo Valsorda:Program exit up. Goodbye.
Creators and Guests
