Which Skills Will Matter Most for Software Engineers? | Kunal Kushwaha — Transcript
Full transcript
- 0:00What apart from AI do you bring to the
- 0:03table is the question you need to ask
- 0:04yourself. I don't want to talk about
- 0:06code.
- 0:06>> Yeah.
- 0:06>> Don't talk about code with me. Talk to
- 0:08me about this website. What would you
- 0:10break down? How would that change the
- 0:12market? How would that change the
- 0:14industry? And what changes would you
- 0:16make if you had full creative control?
- 0:18>> So, it's really about focusing on this
- 0:20product mindset, I think, is what I'm
- 0:21hearing.
- 0:21>> Product mindset. Ship fast. Use AI to
- 0:24achieve all the tasks and goals,
- 0:26>> but make sure you're not completely
- 0:28relying on it. If I'm in my 40s and I
- 0:31built my career on being very
- 0:32knowledgeable in Kubernetes, for
- 0:34example, what did I do to remain
- 0:36relevant?
- 0:41Hello and welcome everyone to another
- 0:43episode of uh the merch, our podcast
- 0:45brought to you here live from the code
- 0:47rabbit headquarters in San Francisco.
- 0:49Today I'm super excited because I'm here
- 0:50with Kunal.
- 0:52>> Hi. So so excited.
- 0:54>> Yeah, good to have you here. Kunal,
- 0:55you've had an amazing career. you've
- 0:58been um making big waves in terms of the
- 1:00developer community. You've been a CNCF
- 1:03ambassador, GitHub star, uh amongst a
- 1:06bunch of other things. So, I'm very
- 1:07grateful that you took time today to
- 1:09continue our conversation actually from
- 1:11scale all the way now here to San
- 1:12Francisco.
- 1:13>> Always a pleasure and uh this is such an
- 1:16incredible office by the way. I've not
- 1:17been to the code rabbit office before.
- 1:19>> Yeah.
- 1:20>> Uh yeah, absolutely incredible. Thanks
- 1:22for having me. Well, I I am honored to
- 1:24be here with you and I think uh there's
- 1:27a lot of interesting things we can talk
- 1:28about. So,
- 1:29>> yeah. So excited. The weather's been a
- 1:31little interesting this week.
- 1:34>> It is it is San Francisco August. We we
- 1:37call it August actually. [laughter]
- 1:39>> Yeah.
- 1:40>> So, what you see outside here is is
- 1:42actually KL. Um K is the fog or the name
- 1:45of the fog I should say. Uh and um every
- 1:48time in August, you know, the all the
- 1:49fog comes in because there's so much
- 1:51heat being built up on the central
- 1:53continental mass of California creates
- 1:55an uplift and then it pulls in the
- 1:57evaporated air from the ocean and it
- 1:59just collects over San Francisco.
- 2:00>> It kind of reminds me of home back in
- 2:03London.
- 2:04>> Yeah. [laughter]
- 2:05But no, uh I like California.
- 2:08>> Yeah, it's a good place to be, right?
- 2:10>> Yeah.
- 2:10>> Yeah. For us coders, it's not too bad
- 2:12because we we have one more excuse not
- 2:14to leave the house.
- 2:14>> Yeah. Yeah. I mean, I'm in San Francisco
- 2:16every month, so it's like second home,
- 2:19>> I guess. Third home in this case, right?
- 2:21>> Third home now. Uh, I go I'm in San
- 2:23Francisco more than I'm in India.
- 2:25>> Oh, really?
- 2:25>> Um, yeah.
- 2:27>> So, this would be second home.
- 2:28>> Yeah, [laughter] that's actually an
- 2:30interesting point. I think we can um
- 2:32dive right in there actually because the
- 2:34last time we talked, which I mentioned
- 2:36earlier, was at ScaleCon, Pasadena. Um,
- 2:39and we had a very, very quick chat. We
- 2:41had our setup there. So you kind of
- 2:42sweeped in and we touched on a couple
- 2:44interesting points, but one thing we
- 2:45didn't get to talk about it was a bit of
- 2:48your upbringing, I guess you could call
- 2:50it. So, uh, I digged a little bit deeper
- 2:52into after you told me that you were
- 2:54going to go and tell Yeah. [laughter]
- 2:56>> So, I have some spicy takes, but I just
- 2:58I just thought I'd give you the
- 2:59opportunity to to tell the story
- 3:01yourself before I do. Um, can you tell
- 3:03us a little bit about how you ended up
- 3:05here?
- 3:05>> Well, where do I start?
- 3:06>> Yeah.
- 3:07>> So, I wanted to be a doctor when I was a
- 3:08kid. When I was five years old, I was
- 3:10like in an accident and um everything
- 3:14worked out fine. Since then I was like
- 3:16okay doctors are cool you know I want to
- 3:19be a doctor.
- 3:21Um by the time I was like 15 you have to
- 3:24pick a career path in high school. You
- 3:26have to either pick like science uh be
- 3:29that tech or u in India be that tech or
- 3:33medical or if you want to do accounts so
- 3:35commerce um or if you want to learn like
- 3:38history or whatever there's arts for
- 3:39that and um I realized I am terrible at
- 3:43memorizing things can't memorize
- 3:45anything I had my chemistry exam because
- 3:47you know in chemistry there's a lot of
- 3:48rules and for every rule there's like so
- 3:51many exceptions like okay this rule
- 3:52works fine but it's not going to apply
- 3:54for this compound because it's an
- 3:56exception. I could never memorize things
- 3:58like that, you know, like bookish jargon
- 4:00knowledge cuz I can just look it up. Why
- 4:02do I need to memorize, you know?
- 4:04>> Yeah.
- 4:05>> And uh I I was like, "Okay, I can't be a
- 4:07doctor cuz, you know, it's just so much
- 4:09memorizing and reading and whatnot." Um
- 4:13I was always into tech. Wasn't like a
- 4:15hardcore programmer.
- 4:16>> Started when I was in high school with
- 4:18C++, but nothing major. In India, when
- 4:20you're in high school, you're preparing
- 4:22for your university entrance exams.
- 4:24Here's a huge country, 1.4 4 billion
- 4:26people. So we have millions and millions
- 4:29of students who graduate every year. So
- 4:31a lot different than a much smaller
- 4:33country let's say somewhere in Europe or
- 4:36even the US compared to India because
- 4:39these millions of engineers graduate
- 4:41every year and now obviously they're all
- 4:44looking for jobs. So it's very
- 4:45competitive. So from the very young age
- 4:47we're put into that competitive mindset.
- 4:51Um there are various ways to get a role.
- 4:54I think in tech at least um throughout
- 4:56the you know last few decades um there
- 5:00are some really prestigious colleges in
- 5:02India because there's just so many
- 5:03students fighting for seeds
- 5:05>> um getting into those universities is
- 5:07not easy and it's not always about your
- 5:09interest it's about what the university
- 5:11wants maths maths physics and chemistry
- 5:14>> it's like I think the second toughest
- 5:15exam in the world I think China has the
- 5:17toughest one I can be wrong I don't know
- 5:19but it's like definitely up there yeah
- 5:22>> so yeah that didn't work out cuz I
- 5:24didn't I wasn't good at chemistry or as
- 5:26good as physi at physics. Um and then um
- 5:31then yeah what happened then I took a
- 5:33year gap and that's sort of like looked
- 5:35down upon in India. I know ear gaps in
- 5:37Europe are great. I'm going to take an
- 5:38ear gap. I want to travel.
- 5:40>> Yeah. So it's looked down upon in India
- 5:43just like okay why are you taking an ear
- 5:44gap because you know the race moves so
- 5:46fast.
- 5:47>> Yeah.
- 5:48>> Took an ear gap. tried for the exam
- 5:49again to get into the top ID Indian
- 5:52Institute of Technology. Didn't work out
- 5:53cuz
- 5:55>> I mean a million students take that exam
- 5:57every year. I think they have around
- 5:595,000 seats.
- 6:01>> Wow.
- 6:01>> Look at the odds.
- 6:02>> No.
- 6:02>> Even if you're great, you won't get
- 6:04computer science. But there's one option
- 6:06for me. Like I could get a really nice
- 6:08college, but then I wouldn't be studying
- 6:10computer science.
- 6:11>> I would be studying something else like
- 6:13metal energy or well, god forbid
- 6:15chemistry.
- 6:16>> Yeah,
- 6:17>> I don't hate chemistry. Okay. there's
- 6:18any chemistry lovers. I love chemistry
- 6:20but it's just uh I can't memorize the
- 6:23>> compounds. But anyway, um then I was
- 6:25like okay let's just get whatever I'm
- 6:27getting.
- 6:28>> So I did that and I was like okay in
- 6:30life I'm always like whatever happened
- 6:33whatever has happened has happened.
- 6:35>> Yeah.
- 6:35>> I mean sure that is just life. If
- 6:38sometimes things don't work out
- 6:39according to your plan
- 6:42>> you can't time travel. That's didn't get
- 6:45a college didn't get a college. What
- 6:46what what would I do you know? Yeah, of
- 6:48course.
- 6:48>> Um, so I can't time travel back, but
- 6:50whatever you have in your resources
- 6:52right now, just use that to make the
- 6:55best out of it. And that's what I do. So
- 6:56I was like, okay, I'm in a decent
- 6:58college now. And there are examples that
- 7:00I've seen. The CEO of Microsoft didn't
- 7:02go to IIT.
- 7:03>> Satya Nadella, he went to MIT. It's a
- 7:05private college in India. I got into
- 7:07MIT. So I'm like, not not MIT US. It's a
- 7:10Manifal Institute of Technology in
- 7:12[laughter] India. Um, so I'm like,
- 7:14"Okay, these people are doing great, you
- 7:16know, things in life.
- 7:18>> The IIT tag is not the only thing." So I
- 7:20was like, "If they can do it, I can do
- 7:21it."
- 7:22>> Yeah.
- 7:22>> And then then I was like, "Okay, I'm
- 7:23just going to start learning and
- 7:25building and just do a bunch of cool
- 7:27stuff." That's when I found open source
- 7:29changed my life. Um,
- 7:32and yeah, I mean then that's that's how
- 7:34it started.
- 7:35>> Yeah. How do you say open source changed
- 7:37your life? I mean it's there's so much
- 7:40competition in the market you know um uh
- 7:44think from a student's point of view
- 7:46they're looking for internships first
- 7:47then they look for full-time roles
- 7:49>> it's a standard thing uh but when you go
- 7:52for an internship you have to send a
- 7:54resume in and in the resume they mention
- 7:57what have you done in the past and the
- 7:58students like I haven't done anything
- 8:00because I didn't have an internship but
- 8:02you need to put something in your resume
- 8:03to get the internship just like deadlock
- 8:06>> um open source is great because um you
- 8:09don't need to interview at a company.
- 8:11You can just pick a nice cool project.
- 8:12>> Yeah.
- 8:13>> Um contribute to it, put that on your
- 8:16resume,
- 8:18career takes off. So if you're a
- 8:19contributor, if you're a hardcore
- 8:20contributor in Kubernetes, like you
- 8:22wouldn't have issues at least getting an
- 8:24interview from some top tech companies,
- 8:27you know, like I got uh from Red Hat um
- 8:31some other companies and um it just
- 8:34opens a lot of doors for you and you get
- 8:36to meet a lot of cool people and I think
- 8:38that was the main main thing. Um yeah, I
- 8:42contributed to Kubernetes when I was in
- 8:44sophomore year.
- 8:45>> Wow. And then yes then I my whole career
- 8:48was in cloud native still am now
- 8:50obviously AI.
- 8:51>> Yeah.
- 8:51>> Um and yeah which is a bit interesting
- 8:55because the complaint I see from people
- 8:57is like open source is so complex like
- 8:59these projects are so big to contribute
- 9:00to.
- 9:02>> But it doesn't mean that it's not open
- 9:04to beginners. I mean cubernet is I think
- 9:07second largest project in the world
- 9:08after Linux.
- 9:09>> It's also one of the most welcoming.
- 9:13That's the word I'm looking for. Mhm.
- 9:14>> So they are always looking for
- 9:16contributors. There are a number of
- 9:17things you can do. Code contributions,
- 9:18non-code contributions. The key is just
- 9:21getting started. Um the one last thing
- 9:24is like if you don't put in efforts
- 9:25yourself,
- 9:26>> I mean the community people are there to
- 9:29help you but if someone sees this person
- 9:31is not putting efforts themselves, no
- 9:32one's going to help you.
- 9:34>> Well, how do you feel like that now
- 9:36changed? Because I mean it seems to me I
- 9:38have my open claw running. I can just,
- 9:40you know, pick up the repo and tell it
- 9:41to make a contribution. Like how If this
- 9:44happened now, yeah,
- 9:45>> this my story happened now. I was in a
- 9:47uni that wasn't as good and now I'm
- 9:49like, okay, I want to look for a role.
- 9:51[snorts]
- 9:51>> Okay, it's a good question. What would I
- 9:53do?
- 9:54>> Um,
- 9:55>> I feel like the way of contributing
- 9:56changed, right? Yeah, way way of
- 9:57contributing has changed but I think the
- 9:59whole narrative is the same in terms of
- 10:02building because
- 10:04programming languages and how we
- 10:06interact with the systems is just it's a
- 10:08mechanism you know 20 30 years down the
- 10:12line like sorry not down the line 20 30
- 10:14years before people were like coding in
- 10:15very low-level languages let's say even
- 10:18before that there were punch cards and
- 10:20whatnot
- 10:21>> so the problem solving mindset was still
- 10:23there so to answer your question I think
- 10:25gone are the days and companies are
- 10:27going to be like, I want to hire this
- 10:28person because they write better Java
- 10:30than this other person. I don't think
- 10:32that's going to be the case anymore.
- 10:34It's more about how much of a generalist
- 10:36are you, how can you
- 10:38>> I mean,
- 10:40how how well you communicate with
- 10:42others, for example, you know, other
- 10:43skills like empathy, problem solving.
- 10:46Um, I love the AI generation because I
- 10:49don't have to worry about syntax and
- 10:51programming languages and things like
- 10:52that. I have can focus on engineering. I
- 10:54can focus on building and you know
- 10:56problem solving and stuff like that. So
- 10:58AI takes a lot of that off my plate. Um
- 11:01so yeah I can write code as you know I
- 11:03can review code
- 11:04>> um do a bunch of stuff write docs
- 11:06testing and whatnot.
- 11:08>> So it's it's just giving people time
- 11:11>> you know to build stuff
- 11:14>> in the very short period of time. I mean
- 11:16my mother can build a website now.
- 11:18>> Yes.
- 11:18>> You know it's very simple easy for her.
- 11:20There's so many so many tools out there.
- 11:22So it's just giving people more control
- 11:24to you know um bring their ideas to
- 11:28reality. Um and this is a big thing for
- 11:30small businesses. You know I read this
- 11:32somewhere like when something becomes
- 11:33easier to do more people do it. And um
- 11:38yeah it's just how do you give access to
- 11:41you know the right tools to people at
- 11:43scale
- 11:43>> right and that is a beautiful arc I
- 11:45think to the next chapter of uh of your
- 11:48life which then focused on developer
- 11:50education. Right.
- 11:51>> Yeah. So can you tell me a bit more
- 11:53about how that how you got started? How
- 11:55what got you into that and how do you
- 11:57scale it today?
- 11:59>> Yeah. Um
- 12:01sophomore year I was in Kubernetes. I
- 12:03was contributing to Kubernetes and then
- 12:05um I was uh I got into this program
- 12:09called Google Summer of Code with Red
- 12:12Hat. It's like an open source program.
- 12:13You contribute to projects
- 12:15>> and then Google pays you.
- 12:17>> And then I was a mentor in that same
- 12:18program. And then I realized uh when I
- 12:21was learning how to code that I love
- 12:24going into the details of how things are
- 12:25working and whatnot and I love teaching.
- 12:28So I started volunteering at this local
- 12:30boot camp in Delhi and um I realized
- 12:34like I love teaching.
- 12:35>> Yeah.
- 12:35>> People told me you're so good at
- 12:37teaching.
- 12:38>> I know. Oh, I watched one [laughter] of
- 12:39your videos like you know what I really
- 12:41um it was uh crazy you I think one of
- 12:45your very early videos four years ago
- 12:47you have this scratch pad digital
- 12:48scratch pad on like a Windows or
- 12:50something.
- 12:51>> Yeah. Yeah. You just draw like little
- 12:52figures on it but it was so well
- 12:54explained
- 12:55>> and here we we all sitting there. I
- 12:57could use a lot of videos as well with
- 12:59crazy animations and so on and I thought
- 13:01like no it can be so simple. Yeah, it's
- 13:03uh people don't watch what you do, they
- 13:06watch how you do it. Even though there's
- 13:08a lot of tutorials on Kubernetes, if
- 13:11someone can explain Kubernetes in a in
- 13:13their own way, I think they'll get a lot
- 13:15of views and they'll get their own
- 13:16audience. So, it's [clears throat]
- 13:18>> this is what the question people have
- 13:19like, hey, why would I make a video on
- 13:21React or well something like codeex or
- 13:24whatever when there's already creators
- 13:26making videos on it? The topic doesn't
- 13:28matter. What matters is how you do it,
- 13:30>> you know. Uh but I realized I'm very
- 13:31good at like uh people say I'm very good
- 13:33at teaching and then um COVID came.
- 13:37>> Uh COVID came I was in San Diego 2019
- 13:40for CubeCon and then 2020ish early COVID
- 13:42came. So all these boot camps I was
- 13:45teaching at they had their own online
- 13:46portal. So all the kids were now
- 13:49learning online. Um
- 13:52and now uh they're like kunal you're not
- 13:55on the online portal. Yeah I didn't
- 13:57record videos for them. I just teach
- 13:58online in the in the building in person.
- 14:01They're like, "No, no, I'm going to
- 14:02learn from you."
- 14:03>> And I'm like, "Oh, okay." And I started
- 14:05my YouTube channel and I'm like, "Okay,
- 14:07if if you want to learn from me, it's co
- 14:09I'll just teach on YouTube." I remember
- 14:11I uploaded five videos on Python.
- 14:13>> Yeah.
- 14:14>> And each video was like 4 hour long.
- 14:17>> It was like full Python course in the
- 14:19depth.
- 14:20>> Yeah.
- 14:20>> With some like 10 projects or whatnot.
- 14:22Like if you don't know anything about
- 14:23programming, you could be like interview
- 14:26ready. that kind of yeah
- 14:28>> that I think because of that I got zero
- 14:31to 100k subscribers organically in like
- 14:342 and a half months
- 14:36>> and I realized okay this is this has
- 14:38been good people have been very kind
- 14:40>> you found your niche this is
- 14:41>> yeah people have been very kind to
- 14:42support and then I got into devops and
- 14:44then just started pushing more content
- 14:46teaching more people making education
- 14:48more accessible
- 14:50>> um and then I was like okay now I have a
- 14:53few it just kept growing in 100 in like
- 14:55a year I had like half a million or more
- 14:57than that.
- 14:58>> And then I was like, "Okay, all these
- 15:00people are now asking me questions."
- 15:01>> Yeah.
- 15:02>> So, what do I do? Cuz if someone is
- 15:05asking me a debugging uh like wants me
- 15:07to debug, help them debug something or
- 15:09whatever. The YouTube comments are not
- 15:10the right place for that.
- 15:12>> Like I'm like, "Okay, I'll start a
- 15:13Discord server."
- 15:14>> Started a Discord server.
- 15:16>> I don't remember what it's called. It
- 15:17was called something. And uh that grew
- 15:20to like 70 80,000 people. Um and then
- 15:25that converted to V make devs. So like
- 15:28we have this whole community um I
- 15:31started giving them challenges and we I
- 15:33was like okay um let's do some hackathon
- 15:35or something you know because have this
- 15:37big community and learning online is
- 15:40great but what I would recommend is to
- 15:43everyone even in today's age because you
- 15:44asked me earlier like what I would do in
- 15:46the AI age if I was a student now I I
- 15:49don't think much would change than what
- 15:51I did earlier honestly. Henrik um it's
- 15:54just the medium would change.
- 15:55>> I would be using a lot more AI tools.
- 15:57>> Yeah.
- 15:58>> But apart from that problem solving,
- 16:00contributing to open source, going to
- 16:02hackathons, creating content, writing
- 16:03blog posts, um speaking at conferences
- 16:07as a student.
- 16:08>> Um
- 16:10>> one word that I just get started.
- 16:11>> Yeah. That resonated with me was like I
- 16:13think learn in public. Is that
- 16:14something?
- 16:15>> Yeah. Build in public, learn in public
- 16:17cuz you get um learning in public
- 16:19doesn't mean like there's a mistake
- 16:21people make. They're watching my videos
- 16:23like today I learned u support vector
- 16:26machines from Kunal Goshwa.
- 16:27>> No one who is watching that post cares
- 16:30>> like [laughter]
- 16:32that's that's what the post would be
- 16:33today I learned about two-pointer from
- 16:35Kunal Kushwa.
- 16:36>> Mhm.
- 16:37>> That's not learning in public. That's
- 16:38just you sharing about your progress
- 16:39which is great.
- 16:40>> Yeah.
- 16:41>> But think about it like someone who is
- 16:43watching that post someone who's a
- 16:45really good engineer, senior person,
- 16:47it's not giving them enough value.
- 16:49>> Mhm.
- 16:49>> Right. It's just helping my promotions
- 16:51which I appreciate [laughter]
- 16:53you know I I appreciate that please keep
- 16:56it coming but no jokes aside
- 16:58>> when you talk about learning in public
- 16:59it's more about
- 17:01>> sharing your knowledge out there
- 17:02>> so I learned this computer course from
- 17:05Kunal or whatever
- 17:06>> if you find something interesting share
- 17:08about that write that in a blog post
- 17:09speak about it at a conference
- 17:11>> credit the creator if you saw the video
- 17:14from them if you want to no pressure um
- 17:18but that's what learning in public is
- 17:19all about and one other thing is because
- 17:20you get feedback from people in the
- 17:23industry and uh quite often I mean
- 17:26Twitter can X or LinkedIn can be a
- 17:29little uh
- 17:32what's the word I'm looking for
- 17:34>> cringe
- 17:35>> not cringe but like interesting place
- 17:37>> cuz sometimes if you ask for feedback
- 17:39people will have no filter they lack
- 17:42empathy is what I'm saying
- 17:43>> oh yeah I think especially the YouTube
- 17:45comments I don't know what's going wrong
- 17:46with people on YouTube but you need to
- 17:47stop like [laughter] guys this This is
- 17:49too much.
- 17:50>> No, one good other good thing is I get
- 17:52so many comments I just don't see a lot.
- 17:55>> Yeah.
- 17:55>> Uh criticism is fine or always welcomed.
- 17:59Let's make sure is I'm just saying like
- 18:00you will find people
- 18:01>> proportions like constructive.
- 18:02>> Yeah. You will find
- 18:04>> folks saying like hey this is not good
- 18:05but if if you're a student if you're
- 18:07just getting started
- 18:08>> I mean learning club is not just for
- 18:09students like anyone even if you're
- 18:11launching your own startup
- 18:12>> getting feedback online. I mean this is
- 18:13actually learning in public is what
- 18:15people do actually with these go to
- 18:16market sort of like activities because
- 18:18they do launch on product hunt and they
- 18:20get um sorry building in public yes
- 18:22>> which is sort of like you're learning
- 18:23from your customers or whatever.
- 18:26>> Uh but you're building in public you're
- 18:27doing soft launches on product hunt get
- 18:29a lot of feedback you're doing
- 18:30hackathons and stuff you're doing go to
- 18:32market events you're doing uh some
- 18:34webinars you're doing uh conference
- 18:36sponsorships so you're learning from the
- 18:38community in the public.
- 18:40>> Yeah.
- 18:40>> Right. It's just a different scale at
- 18:42what beginners do it at. But yeah, make
- 18:44sure the posts that you're making are
- 18:45valuable. Um, don't spam people. Don't
- 18:48use too many hashtags. Um, please stop
- 18:51the AI slop. It's just not going to
- 18:53work. It's it's not working. Use AI to
- 18:56write content. I'm totally for that.
- 18:58Okay?
- 18:59>> Use that uh to refine your content. I
- 19:02know English is not the first language
- 19:03for a lot of people.
- 19:04>> Yeah.
- 19:04>> Um, so you can use AI to sort of like
- 19:06help you with that, but some people just
- 19:08completely copy paste. We can tell.
- 19:11>> Oh, I mean I think then also the bot
- 19:12traffic and the bot commenters,
- 19:14especially LinkedIn, their banning
- 19:15algorithm is not so good yet.
- 19:17>> Yeah, it's not that good, but it's
- 19:18actually doing you more harm than good.
- 19:20>> Mhm.
- 19:21>> So, use AI. I'm very pro AI, but don't
- 19:24just copy paste uh comments. But yeah,
- 19:26for learning in public, getting feedback
- 19:28about your projects. Uh one best thing
- 19:31about learning in public is you make so
- 19:33many friends. I think that's the best
- 19:34thing.
- 19:34>> Okay.
- 19:35>> Um you make a lot of friends and you
- 19:37build your profile. Um,
- 19:40in the world is so small. Like every
- 19:42month in San Francisco, I run into
- 19:44someone that I knew from someone else.
- 19:47>> I mean, San Francisco is a surprisingly
- 19:50small city as well.
- 19:51>> The world is super small. [laughter]
- 19:54>> That's right. Well, yeah, I think from
- 19:57that arc, moving on, uh, you you talked
- 20:00about, you know, how you built this
- 20:02passion for learning and then how that
- 20:04translated into your community that
- 20:05you're building now. Can you tell me a
- 20:07bit more about that community? How big
- 20:09is it? What are what are you guys doing
- 20:10exactly?
- 20:12>> Yeah. So, when I was at conferences and
- 20:14stuff, uh when I was doing booth duty,
- 20:17it was great. There's just a lot of
- 20:19sales people in the room.
- 20:20>> Mhm.
- 20:20>> Which is fine. Sales people make it
- 20:22happen. They are the ones who bring
- 20:24money in.
- 20:24>> Yeah.
- 20:25>> But like if I want to if you want to if
- 20:27if I want to get developer feedback, if
- 20:29I want to get like what's wrong with the
- 20:31tool, what can I change? How easy was it
- 20:33to contribute? how easy was it to
- 20:36onboard and things like that. I think
- 20:39that's where you need to obviously get
- 20:40your hand get your tool in the hands of
- 20:42builders. So I think these sort of like
- 20:46go to market or engagement events that
- 20:48companies do. Let's say do like a
- 20:50workshop thingy or a webinar always talk
- 20:54always think about how engaged do you
- 20:56think people are.
- 20:58Some people they just do a webinar live
- 21:00and that doesn't really convert to
- 21:03anything. You know the argument they
- 21:04give is oh we got 100 email addresses
- 21:06yeah how many of them will reply to you
- 21:09know try getting a reply from everyone
- 21:11how many of them will get on a meeting
- 21:12with you so the coefficient factor for
- 21:16like how much engaged a person is I
- 21:19think that's something that you need to
- 21:21think about for your marketing efforts
- 21:23for developer not like you in general
- 21:25like uh people in general
- 21:27>> um hackathons are great because we we
- 21:31did one I think it was 2021 we got like
- 21:35um 2,000 something people and I hadn't
- 21:38even done
- 21:39>> any was it remote or
- 21:41>> it was remote from from my YouTube I
- 21:43think we got two two to 3,000ish or
- 21:45something
- 21:46>> and it was with elastic
- 21:47>> I remember
- 21:49>> um actually it was by Arvind
- 21:53>> Oh really Arvind for context is the
- 21:56previous VP of growth at great person
- 22:00>> great person and he was my first he was
- 22:01a very good friend so he was supportive
- 22:03and it was very small like event we did
- 22:06the budget was a thousand bucks. Um we
- 22:09got so many people share feedback and
- 22:11stuff and then uh we just started doing
- 22:14more and more events. Um I think last
- 22:16year because we make devs community we
- 22:20we renamed it like last year but it was
- 22:22so big that we were trying to figure out
- 22:23okay what do I want to do?
- 22:24>> Okay I want to do workshops I want to do
- 22:26inerson meetups and stuff. We we did one
- 22:28or two more hackathons and I realized
- 22:30okay this is like the thing that people
- 22:33want.
- 22:34>> Yeah.
- 22:34>> The the community wants this because
- 22:36when we do hackathons that's the most
- 22:38engagement we get.
- 22:39>> So around 10,000 people
- 22:41>> Yeah. They try out the product you can
- 22:42be sure that everybody who attends will
- 22:44actually have physical touch with the
- 22:46tools of the sponsors.
- 22:48>> Yeah. So for for the attendees uh and
- 22:50then I was like okay let's just do
- 22:52hackathons. That's what's working. Let's
- 22:53just do that.
- 22:54>> There are two strains to it. One is why
- 22:56would the audience care? Mhm.
- 22:57>> Why do the companies care?
- 22:59>> So for the audiences, I think they get
- 23:01to because there's the trend is shifting
- 23:03so fast. There's a new tool come out, a
- 23:05new inferencing tool will come out, a
- 23:07new monitoring tool will come out. Um
- 23:10>> the way we look at it differently at PMC
- 23:12Devs is we are we are riding with that
- 23:14trend.
- 23:15>> Something cool comes out, we give people
- 23:16the platform to build with it and share
- 23:19their feedback. That's what I care about
- 23:21the most because we have this incredible
- 23:24community behind us. The feedback that
- 23:26they give us is worth so much to well
- 23:31essentially all the companies in the
- 23:33world, the tech companies.
- 23:34>> Yeah.
- 23:34>> Because you know when a VC firm spons
- 23:37wants to invest in a company, they will
- 23:38go to like the research firms and some
- 23:40independent people. They pay them like
- 23:42two 300k for doing that background check
- 23:45about the industry analysis and whatnot.
- 23:48um big big companies like yeah like
- 23:52Gartner and stuff they publish their
- 23:54reports and they work with enterprises
- 23:56and share those findings and stuff with
- 23:58them. What we have found working with
- 24:00companies is um imagine this you're a
- 24:04company we're doing a hackathon together
- 24:0810,000 people test out your platform
- 24:11>> signups get up stars and whatnot
- 24:13>> I genuinely don't believe in vanity
- 24:15metrics I don't think it matters as much
- 24:18um may look good on the paper
- 24:21>> but the kind of numbers if you want to
- 24:23expect from a hackathon or like any
- 24:25conference you're sponsoring it doesn't
- 24:28matter if it's Not I mean in the end it
- 24:29doesn't matter if it's not directly
- 24:32relating to some sort of business
- 24:34outcomes
- 24:35>> you know because in the end that's what
- 24:37people care about that's what the come
- 24:38that's what's going to pay the salaries
- 24:40and whatn not
- 24:41>> well I mean these events often they're
- 24:42there for generating pipeline for right
- 24:45>> yeah but it's sometimes you know most of
- 24:47the time it's like indirect as well so
- 24:49let's say if you're doing a hackathon
- 24:50with students
- 24:52>> um we do like professionals senior
- 24:55engineers and students and everyone
- 24:57included but that ROI you will get when
- 24:59the students graduate and now they're
- 25:02working at a company and they would know
- 25:03like okay if I want to do take an
- 25:07example I want to do code reviews
- 25:09>> I did code reviews when I was in u my
- 25:11uni
- 25:12>> and I used code rabbit so they you know
- 25:16kids they become like a community for
- 25:19their
- 25:21career
- 25:22>> it it transitions into the career
- 25:23because when they're new in the career
- 25:25they have to answer all these questions
- 25:27I've given this task ask him what do I
- 25:28do? Oh, I used this tool like a few
- 25:30years ago. Why do you think all these
- 25:31big companies like uh you know all the
- 25:33hyperscalers they all have their student
- 25:35programs?
- 25:36>> Oh yeah, I know. I for me I
- 25:37[clears throat] studied meatronics in in
- 25:39university and I remember still working
- 25:41without a desk for my CAD project. So
- 25:43>> yeah,
- 25:44>> so it translates it just carries on and
- 25:47the time moves by so fast. These people
- 25:48will graduate they'll become seniors and
- 25:51now you have your own sort
- 25:52[clears throat] of like community
- 25:53members. So even though if like let's
- 25:55say you're at an event, you're just
- 25:57scanning people. Let's say you're at the
- 25:58booth duty. Someone comes in, you're
- 26:00like, "Hey, can I scan you? I'll give
- 26:01you a t-shirt." And the guy and the
- 26:03person's like, "Yeah, sure." Scan them,
- 26:05give them a t-shirt. Then you go to your
- 26:07VP, "Hey, we got a thousand lead scans."
- 26:10And then you reach out to the thousand
- 26:11lead scans and there's just no reply,
- 26:13nothing.
- 26:14>> It just looks good on paper.
- 26:15>> Yes.
- 26:16>> That hey, we got thousand leads from the
- 26:17event. Looks good for the CMO or
- 26:19whatever. But
- 26:20>> is that translating into anything? So
- 26:22how do you promote uh ROI from your
- 26:26events?
- 26:27>> So for us it's about the depth rather
- 26:29than just the numbers. There's this
- 26:30thing called value hours. The idea is
- 26:32that
- 26:33>> you did a hackathon or you did a webinar
- 26:35100 people came 1 hour was the average
- 26:36view time
- 26:37>> and there's a coefficient to it which is
- 26:39how engaged do you think the people
- 26:40were.
- 26:41>> Webinar people are not as engaged as
- 26:43inerson workshop. So let's put a
- 26:44coefficient of 0.5 100 into 1 hour into
- 26:470.5 50 hours of value hours. So that's
- 26:50like measuring the kind of depth how
- 26:52engaged the person is in your event. For
- 26:55our hackathon specifically, u it helps
- 26:57because it's hands-on.
- 26:58>> Yeah.
- 26:59>> Um the most important thing when we work
- 27:02with like enterprises, open source
- 27:03projects, dev tools, we got a lot of
- 27:05professionals and senior senior
- 27:07engineers. So we need to give incentive
- 27:09for them to come.
- 27:10>> So one of the incentives is you're
- 27:12riding the eye wave. You're basically
- 27:15any new cool tool comes out, we do a
- 27:17hackathon around it. any new industry
- 27:19sort of like term comes out we do a
- 27:21hackathon around it we get experts on
- 27:22this you know in the stream and you
- 27:24share resources that's the second part
- 27:27making it easy for people to contribute
- 27:30making it easy for people to build with
- 27:32these tools
- 27:33no one's judging you
- 27:35>> I think that's why people come to
- 27:36hackathons they want to learn they want
- 27:37to see what's out there
- 27:39>> surprisingly I thought it was going to
- 27:41be just students but we get like
- 27:43>> so many engineers and professionals
- 27:45>> yeah I mean
- 27:45>> the one we're doing in SF the one we did
- 27:47at GitHub hub 35% executives came
- 27:51uh 30 or something other person senior
- 27:54engineers and uh a lot of developers and
- 27:57people
- 27:58>> yeah over here you even see a lot of VC
- 27:59funds they sent some of their scouts to
- 28:01these tubby events too
- 28:03>> exactly yeah
- 28:04>> very interesting trajectory I want to
- 28:06pull us a little bit further um because
- 28:08I think one of the most interesting
- 28:09things that you experience every day
- 28:11through running all of these these
- 28:12events you're kind of so close to the
- 28:14developers of today and the developers
- 28:16of tomorrow really So can you give me of
- 28:18an impression like maybe you can start
- 28:19sharing of one of the most impressive
- 28:21projects that you've seen in the latest
- 28:23weeks or two weeks or month. Yeah. Um
- 28:25and then I'll be interested also to
- 28:26learn a little bit more how you see like
- 28:27the development workflow shifting on
- 28:30these events as well.
- 28:31>> Yeah. So what to what I was mentioning
- 28:34earlier like measuring ROI it's around
- 28:37for a dev tool if if I take as an
- 28:39example they care about we want feedback
- 28:43we want quality people like uh from all
- 28:46backgrounds when I mean quality people
- 28:48is people who are actively engaging with
- 28:49the platform can be students can be
- 28:51engineers want feedback for them from
- 28:53them we want to see use cases of our
- 28:55tool uh from the community and if they
- 28:58end up paying in the future we become
- 29:00like a sales pipeline or whatever that's
- 29:01a different story. Um the last hackathon
- 29:05we did to answer your question,
- 29:07>> it's a interesting story. We did it on
- 29:09Saturday.
- 29:10>> Yeah.
- 29:10>> Um at the Bright Data Loft with um port
- 29:15um and uh we had um
- 29:19Bright Data obviously we had people from
- 29:21Anthropic and stuff. It was a really
- 29:23nice big hackathon.
- 29:24>> Um we had a lot of people registered I
- 29:27think over 800.
- 29:28>> Wow. venue can already hold um close to
- 29:31200.
- 29:32>> So you had to cut people off at the
- 29:33door.
- 29:34>> Yeah. No, no, no. Uh we had weight list
- 29:36beforehand.
- 29:37>> Ah okay.
- 29:37>> So we had a weight list beforehand and
- 29:41uh this one kid came.
- 29:43>> He was at I think 12 years old.
- 29:45>> Oh wow.
- 29:45>> And he was on the wait list. So he
- 29:47didn't get in.
- 29:47>> Okay. And he's at the door and he's like
- 29:49his dad dropped him off and he's like
- 29:50hey I registered for the event and I'm
- 29:52looking on Luma and this guy is like 12
- 29:54years old on the Luma. he's not
- 29:56accepted, you know. And then I couldn't
- 29:58say no to him. I'm like, dude, just
- 30:00come. I'm coming. I'm just come on in.
- 30:02Yeah.
- 30:03>> So, I came [snorts] he came in and then
- 30:05um he was sitting I just gave him a
- 30:07desk. His some of his friends as well
- 30:10like yeah, just sit there build.
- 30:12>> In in my mind, I was like, "Okay, he's a
- 30:14kid. I mean, 12 years old." I was going
- 30:16to do
- 30:16>> when I was when I was 12, what was I
- 30:18going,
- 30:19>> you know?
- 30:20>> Yeah. His project was
- 30:24top three projects
- 30:26that I have received in the last 10
- 30:29hackathons we did.
- 30:30>> Wow.
- 30:30>> His project was insane.
- 30:31>> Thousands and thousands of submissions.
- 30:33>> His project was insane. He blew me out.
- 30:35He gave me imposter syndrome and I don't
- 30:37[laughter] get imposter syndrome.
- 30:40>> All right, tell me all about it. I want
- 30:42>> incredible. It was It was not mainly
- 30:43just about project. It's how first of
- 30:45all his demo video was so cool and cute.
- 30:48It was like a role play and he was role
- 30:50playing and he was acting and it was
- 30:52cool. He built like a go-to market sort
- 30:54of like app where you put a idea in his
- 30:57app. It's going to build the whole
- 30:59website and stuff. It's going to do
- 31:00research. It's going to pinpoint people
- 31:02in like San Francisco and stuff that can
- 31:04um you know um that you can engage with.
- 31:08The UI was incredible. how he used the
- 31:10whole pipeline for port bright data for
- 31:12scraping the uh data um from the
- 31:17internet and the UI was so good and uh
- 31:20the documentation you know small things
- 31:22like that because you have your project
- 31:24on GitHub right so small things like
- 31:25code quality and just how well you use
- 31:27the PRs the documentation read me
- 31:30>> and this was done in eight hours
- 31:32>> wow
- 31:33>> and I'll leave you'll see for yourself
- 31:35>> I mean just the guts to go there and
- 31:37heck with like people that are way way
- 31:39older than you build entire
- 31:41>> you know AI won't take your job these
- 31:42kids will [laughter]
- 31:45>> well yeah interesting that you say that
- 31:47because I think that's my my next point
- 31:48that I want to talk to you about it
- 31:50there's a lot of anxiety in the
- 31:52developer ecosystem right now about AI
- 31:55potentially taking that job like what
- 31:57what does it mean for me as a developer
- 31:59I spend so much time learning C++ I
- 32:03spent so much time understanding
- 32:05pointers and uh you know the complex
- 32:07relationship between hardware ware and
- 32:09software in these lower level languages
- 32:11and now all that time seems like it
- 32:16doesn't account to that much anymore for
- 32:18a lot of people and that triggers an
- 32:19anxiety how do you feel about that is a
- 32:23very good point I would say the number
- 32:26one benefit of going into these details
- 32:28for the basics for example is you
- 32:31develop a superpower
- 32:33>> you know what that superpower is
- 32:35>> so the superpower is you learn how to
- 32:39escape the hype cycle
- 32:41because you know the fundamentals.
- 32:43>> Mhm.
- 32:44>> You cannot be persuaded with the hype.
- 32:47>> Okay.
- 32:48>> And that's very important.
- 32:51>> So I always recommend focus on
- 32:52fundamentals.
- 32:53>> Yeah.
- 32:54>> Um you don't have to be an expert in
- 32:56every single thing.
- 32:58>> Obviously there's pies learning. Know
- 32:59something about uh know everything about
- 33:01something something about everything. So
- 33:03pick a niche. But I think with AI now is
- 33:07I'm not going to say that it's not like
- 33:08a time to worry but it depends where
- 33:11you're coming from. There are a lot of
- 33:13people in India who are like working at
- 33:15very lower wages because
- 33:20I mean there's a big divide in India
- 33:21amongst engineers.
- 33:22>> Mhm.
- 33:23>> Some of the people are running
- 33:24companies. Some people are, you know,
- 33:26CEOs obviously at the big tech
- 33:28companies, but also there are a lot of
- 33:30outsourcing agencies that pay very
- 33:33little amount compared to their home
- 33:36country to developers in India.
- 33:38>> They get a lot of stuff done. They don't
- 33:40have a choice because they know that if
- 33:42I quit this job that's paying me $3,000
- 33:44a year, I'm not going to find another
- 33:46one.
- 33:47>> Right? The question you need to ask
- 33:49yourself is why is it that someone from
- 33:51India came from a village and they went
- 33:54to US they went to London they they
- 33:58obviously did something you know I
- 34:01didn't come from a rich background my we
- 34:02had debt and thing and um grew up in a
- 34:06small apartment no no in no way
- 34:09whatsoever I came from a rich background
- 34:11didn't even go to uni in any university
- 34:13outside India no masters nothing didn't
- 34:15have you know it's expensive studying in
- 34:18the US.
- 34:18>> Yeah. [laughter]
- 34:19>> Um so
- 34:22with AI, I think all those people are
- 34:25the ones who are going to be impacted
- 34:26quite a bit.
- 34:28So
- 34:30it is time to worry but it depends where
- 34:33you're coming from. Coming back to my
- 34:35earlier point like you're not going to
- 34:36be hired if you know one programming
- 34:38language better than the other. Um I
- 34:40think it's more about being a generalist
- 34:43in all honesty. Mhm.
- 34:45>> So if I'm giving an if I'm giving like
- 34:48um task to a student right now, student
- 34:51is like hey I have an interview and
- 34:53whatnot and how do I prepare or what
- 34:56task can you give me or like just in
- 34:57general when I'm approaching founders
- 34:59and recruiters and stuff give them a
- 35:01website and ask them break down this
- 35:03website for me. What would you change in
- 35:06this website? I don't want to talk about
- 35:07code.
- 35:07>> Yeah.
- 35:08>> Don't talk about code with me.
- 35:09>> Talk to me about this website.
- 35:11>> What would you break down? How would
- 35:13that change the market? how would that
- 35:14change the industry and what changes
- 35:17would you make if you had full creative
- 35:18control and things like that?
- 35:20>> Yeah.
- 35:20>> Um, so I think that's one and obviously
- 35:23the second the obvious answers the
- 35:25obvious answer is that how well people
- 35:28work with these AI tools. So if you're
- 35:29looking to get a job in law, what kind
- 35:32of AI tools are being used in the law
- 35:33industry? Um, if you're Kubernetes, how
- 35:36do you use AI to make Kubernetes
- 35:38smarter,
- 35:39>> right? If you're in testing, how do you
- 35:40do that? um if you're um if you're a
- 35:45sort of data science person so there's a
- 35:46lot of stuff that junior data scientists
- 35:48used to do that now AI can do for you so
- 35:51what apart from AI do you bring to the
- 35:54table is the question you need to ask
- 35:56yourself here's the other thing you know
- 35:58why this is very important the the the
- 36:00issue with people and students right now
- 36:03you see them hey new uh model came out
- 36:06and my tools and my pipeline and
- 36:09everything is 30% more faster faster,
- 36:1140% more accurate.
- 36:13>> The model that just came out that I just
- 36:15used in Claude is now 40 times more
- 36:18smarter, you know, 40% more smarter. Um,
- 36:22this new inferencing layer that came out
- 36:25or whatever, this inferencing tool that
- 36:27came out is 20% more smarter or faster.
- 36:30This rooting mechanism that came out
- 36:32that does everything for me, I just
- 36:33write one prompt is now 10% more
- 36:36smarter. What about you?
- 36:39Are you smarter than what what you you
- 36:41know than you what than what you were
- 36:43yesterday? What are you doing to improve
- 36:46your own personal self? So using these
- 36:48AI tools as a helping hand is I think
- 36:50the key. Uh don't worry about
- 36:55don't worry about having n number of
- 36:57things to learn. The industry will
- 36:59always change. There's always going to
- 37:00be something that you don't know and
- 37:02that's fine. So it makes us human. Do
- 37:04you know do you think I know everything
- 37:05about Kubernetes? I don't. and I was
- 37:07field CTO at Civo building data centers.
- 37:09I don't know everything about
- 37:10Kubernetes. I don't think many people
- 37:12do. This is such a big project. I don't
- 37:14know everything about Linux. I don't
- 37:16know everything about networking,
- 37:18>> but I know a lot from my hands-on
- 37:20experiences, right? Um, so if you're
- 37:24looking for if you're scared about
- 37:25losing your job to AI, if AI can replace
- 37:30any programming task that you're doing
- 37:32right now, 100% you [clears throat] will
- 37:34get replaced. No question about it. If
- 37:36you're just good at C++ or Java and AI
- 37:39can do that for you, you don't bring any
- 37:41sort of other problem solving skills to
- 37:43the table. Any sort of like uh uh you
- 37:46know teamwork skills, empathy is
- 37:49obviously a crucial one. How do you
- 37:51interact well with developers, your
- 37:53users, UI, UX and whatnot? Just the
- 37:56whole sort of like shenanigan.
- 37:58Yeah, if if AI can replace you, it will.
- 38:02And um if you use AI as a helping hand
- 38:06>> to sort of like um
- 38:09empower stuff that you're already good
- 38:11at,
- 38:13I think it's great. So taking back to
- 38:14the example of that 12-year-old kid, he
- 38:16used claw or cordex or something pretty
- 38:18heavily
- 38:19>> but the thought process behind it, the
- 38:21infrastructure behind it, the
- 38:22architecture behind it. Um
- 38:25a lot of questions lot of uh
- 38:27conversations recently around
- 38:28tokconomics people spending because you
- 38:31know when cubernetes came around 2014
- 38:33and then there was the cloud world
- 38:34people money sending money in the cloud
- 38:37bills and whatnot and now even today
- 38:39there are so many companies active
- 38:41there's a whole phops foundation the
- 38:43whole idea of a phops foundation is
- 38:44reducing your cloud bills there's so
- 38:46many companies
- 38:47>> you worked in that space before right
- 38:49>> yeah I was at casti I was leading
- 38:50developer relations and um business
- 38:53doing some amazing so many companies
- 38:55doing similar stuff.
- 38:57>> So now we're seeing that with AI as
- 38:58well. They have big budgets. Do you see
- 39:00that? Did you see that news? 500
- 39:02billion.
- 39:03>> 500 million or 500 billion on cloud
- 39:06bill. Half a bill something like that.
- 39:09>> That would be like half of their
- 39:10>> No, no, no. 500 million. 500 million.
- 39:13>> I always think like cuz they Yeah. They
- 39:14only have like somewhat billion of
- 39:16revenue right now, which is crazy
- 39:18already. But uh
- 39:18>> Yeah. It can't be half a trillion
- 39:20obviously.
- 39:20>> Yeah. Yeah. [laughter] uh half a half a
- 39:23billion dollars on claude bill and stuff
- 39:26>> and even Sam Alman all these people are
- 39:28now like hey um
- 39:31>> people need to talk about you know
- 39:33reducing their token costs
- 39:35>> um a lot of uh agent hardnesses having
- 39:38the narrative around that so AI won't
- 39:41replace you personally a person using AI
- 39:43will I don't know who said it
- 39:45>> but it's a very good term but also
- 39:47something becomes easier to do a lot
- 39:49more people do it
- 39:50>> you can spend spend more time in problem
- 39:51solving. There are so many problems in
- 39:53the world. Solve those problems. Use AI
- 39:55to
- 39:55>> Yeah.
- 39:55>> as a helping hand.
- 39:56>> I I want to dig a little bit deeper into
- 39:58that because the I think the the topic
- 39:59is really really important and you made
- 40:02some great points, but I want to be a
- 40:05bit more practical. What can I do if
- 40:07I've been, you know, you are an
- 40:08experienced engineer, you're not a
- 40:10junior.
- 40:11>> Um, for a junior, it seems somewhat
- 40:13almost easier because you are not
- 40:15constrained to the same ways of thinking
- 40:17that that 12-year-old has no
- 40:21restrain in the his head of like oh this
- 40:23is the way I need to build this fast API
- 40:25server first and then I can think about
- 40:27how I scale out of this no he's just
- 40:29like I want to I he almost has a natural
- 40:31product sense which I think is what is
- 40:33needed also for what you said in terms
- 40:35of problem solving and building
- 40:36something people want but if I'm in my
- 40:3940s and I built my career on being very
- 40:41knowledgeable in Kubernetes for example
- 40:44what did I do to remain relevant do you
- 40:47have a practical advice
- 40:49>> if First of all, I'm 27.
- 40:51>> Yeah.
- 40:51>> Uh, so if if I was in my 40s,
- 40:56>> again, it's it's about how you
- 40:58communicate with the tools.
- 41:00>> So code is just communication.
- 41:03You has to write code, Java, C, C++, and
- 41:06it just interacts with the systems. Now
- 41:08that's transitioning into natural
- 41:10languages. everything else behind that,
- 41:13the architecture, how do you want to
- 41:15system your, you know, architect your
- 41:16system, how do you want to maintain it
- 41:18and things like that, that's all still
- 41:21there. There's always going to be a need
- 41:22for a human to do that. So
- 41:25my my answer is just in two parts. First
- 41:29one being don't let go of the basics.
- 41:31>> Yeah,
- 41:32>> that's first.
- 41:34Yeah, critical thinking. You you get to
- 41:36escape the hype cycle. You get to know
- 41:38the basic foundations and fundamentals.
- 41:40Mhm.
- 41:40>> And the second thing is like use AI to
- 41:43automate the mundane tasks that you
- 41:45don't like. Um and double down on
- 41:50problem solving is what I would
- 41:51recommend. So if you're 40 years old,
- 41:53good news for you because uh you're so
- 41:56into your career now. You have more time
- 41:59to do stuff that was on your backlog.
- 42:01And the last thing I would mention is I
- 42:03don't know at which point we are now
- 42:05where you can just I think we are there
- 42:07that you can just approve the code that
- 42:10AI wrote.
- 42:11>> Oh.
- 42:11>> Um
- 42:12>> okay that's an interesting time. Yeah.
- 42:14And
- 42:14>> I mean it depends for I mean with code
- 42:16review tools.
- 42:17>> Okay.
- 42:17>> I think they're quite smart enough to
- 42:19figure out
- 42:20>> there are any vulnerabilities and
- 42:21whatnot.
- 42:22>> Um but in the end there's always going
- 42:25to be a human who's approving that
- 42:27>> the PR being merged or whatever,
- 42:29>> right? So when you are using these AI
- 42:32tools to write code,
- 42:35just make sure it's checked,
- 42:38you know.
- 42:38>> Yeah. And
- 42:39>> apply your because you're the one who
- 42:41still owns the the responsibility,
- 42:43right, for whatever is shipped, not the
- 42:44agent. You cannot blame the agent, but
- 42:46you can also not fire the agent, but you
- 42:48can fire yourself.
- 42:48>> That's the other thing with AI like
- 42:50agent. How do you give access uh AI? How
- 42:53do you give identity access to AI
- 42:55agents?
- 42:56>> Mhm. You know, if agents mess some
- 42:57messes something up, there needs an
- 42:59identity attached to it that you can
- 43:01call someone real.
- 43:03>> Yeah. I mean, that's uh that's a whole
- 43:05different discussion. I did had the
- 43:06pleasure of recording an episode with
- 43:08work, the founder of work. They do
- 43:10authentication for agents. So, I
- 43:12recommend people checking that out. I'll
- 43:13happy to send it your way as well. He's
- 43:15an amazing founder. The company is used
- 43:17by companies like Anthropic, OpenAI, and
- 43:19so on. They use his product. Um, but I
- 43:22want to briefly stick around because I
- 43:24think the topic is very interesting. Um,
- 43:28let's talk about what you just said
- 43:30regarding code review and code review
- 43:32agents being good enough to review the
- 43:33code. So, you don't need to actually
- 43:35like you can just approve.
- 43:37>> I've been hearing a lot of talk about
- 43:39that not being the case. Recently, there
- 43:42was the AI engineering code summit here
- 43:45in San Francisco, World Fair, I think it
- 43:46was called. Yeah,
- 43:47>> I was there.
- 43:48>> You were there. Yeah,
- 43:49>> remember the talk from um shout out to
- 43:52Dex Hury. He uh he made an amazing talk
- 43:54about why agent software factories fail.
- 43:58And the reason he brought was very
- 44:00interesting because he said that
- 44:01basically the reward system how we do
- 44:05reinforcement learning today. It does
- 44:07not take into account the long feedback
- 44:09loop that you have when um agents write
- 44:13unmaintainable code because the fact
- 44:15that you have to do a refactor at some
- 44:16point because things like performance
- 44:18slows down or maybe you had some
- 44:19security vulnerabilities that your tests
- 44:21didn't catch and it deployed fine. All
- 44:23of that is not built into the loop
- 44:24because those issues arise three, four,
- 44:26five, six, seven months later and that
- 44:28hence they are not being downgraded in
- 44:30the reinforcement loop and that's why
- 44:32today you still have to look at the code
- 44:34and you cannot just um deploy you know
- 44:37lights off um software factories because
- 44:40as you and I can talk to that too uh we
- 44:43have the saying at code rabby the
- 44:45entropy in your codebase gets bigger and
- 44:47bigger as you deploy more and more
- 44:49background coding agents into your your
- 44:51codebase. I have a project that I've
- 44:52been working on with coding agents for
- 44:54about a year now and I can speak to that
- 44:55like at the start when Esther first
- 44:57built it out like I felt like I was
- 44:59really using agentic engineering and uh
- 45:02like putting the build it out very fast
- 45:04build out the back end very well build
- 45:05out the ETL pipeline that it needed to
- 45:07process some things very well and I
- 45:09really understood it now my agents are
- 45:11in there all the time while I sleep or
- 45:12while I go to my daytime job
- 45:14>> and I just feel like the entropy has
- 45:16increased so much so I don't I'm not I'm
- 45:18sometimes worried when I click on the
- 45:20even though all the flags are green I
- 45:21don't Do you feel the same for some?
- 45:23>> Let's take an example like Claude is
- 45:24running if let's say Claude is running
- 45:25your Slack and connected to your email.
- 45:28It's essentially the same thing. You
- 45:29could argue that hey, what if Claude
- 45:31sent sends something messed up?
- 45:33>> What if which it does,
- 45:34>> you know, and like Yeah. So that's the
- 45:36thing. It it does something very but
- 45:37people are still like there's one entity
- 45:39that's like I'm not going to give access
- 45:41to my services to AI.
- 45:42>> Mhm.
- 45:42>> Which local AI is getting so popular,
- 45:44you know,
- 45:45>> uh because of that
- 45:46>> and then the others are like okay I
- 45:48don't care. um with code reviews in
- 45:50itself I think it depends on the
- 45:54what is at stake so for example I think
- 45:58we will go into a era into an era where
- 46:03humans are not reviewing code at all I
- 46:05don't know when that will be 10 20 30
- 46:07years down the line we will definitely
- 46:08reach there that will happen
- 46:11>> mhm
- 46:12>> um because of how fast that things are
- 46:13moving there's a lot of you know you
- 46:15mentioned the context increasing and
- 46:16increasing and increasing the whole
- 46:18narrative around like company brains
- 46:20obviously a lot of talk now now around
- 46:23agent memory so that you don't have to
- 46:25give the whole context again and again
- 46:26in every conversation
- 46:28>> but
- 46:30human I could be wrong but human beings
- 46:32not being reviewing not reviewing code
- 46:34and just agent handling everything I
- 46:36don't know you're saying that won't
- 46:38happen in the next 50 years
- 46:39>> oh no I will I'm
- 46:40>> next 100 years
- 46:41>> no no not even I'd say like I don't
- 46:43think we're going to ship like that
- 46:45actually
- 46:47I work at a where we we build our
- 46:49foundation on the fact that that will
- 46:51not be the case anymore. Right. So we
- 46:52building a [clears throat] lot lot of
- 46:54tools. We just recently launched a tool
- 46:56called change stack which is a
- 46:57completely new way of how to review a
- 46:59PR.
- 47:00>> Yeah.
- 47:00>> Uh so we don't we want to strip away
- 47:02from you taking the the looking at the
- 47:04code line by line and we are obstructing
- 47:05multiple layers.
- 47:06>> You have to you you have to because
- 47:08these tools are so awesome. Why would
- 47:10you not use it? If a
- 47:13if an AI tool has that much potential
- 47:15and power to scan the vulnerabilities
- 47:17and check what happened in the industry
- 47:19and make sure that you're not making the
- 47:20same mistakes and stuff, no one's asking
- 47:22you not to review it. But my point is
- 47:25around how much of the what is at stake.
- 47:28So I'll give you an example.
- 47:29>> Mhm.
- 47:30>> It's Black Friday.
- 47:31>> Okay. You are a billion dollar company.
- 47:33You're $20 billion company. You have a
- 47:35big sale on Black Friday. Okay. Or
- 47:37something like that. and you get paged
- 47:39at 2 a.m. get paged at 2 a.m. that
- 47:41something's messed up. You have to fix
- 47:43it otherwise customers will get
- 47:44affected. Let's say guaranteed let's
- 47:47just say guaranteed you will lose uh
- 47:48$500 million in sales guaranteed.
- 47:51>> If you don't fix this now you have two
- 47:53options. You can just ask AI hey
- 47:56>> please fix this you know
- 47:58>> make no mistakes
- 47:59>> make no mistakes please fix it. You are
- 48:01awesome very smart you are now working
- 48:03as a senior debugging on call engineer
- 48:06and whatn not. I don't think when a
- 48:09billion dollar sale is on the line even
- 48:12though the AI can do everything
- 48:13perfectly the that human will double
- 48:16check
- 48:17>> there's so much on the line
- 48:19>> you know so it's more about the
- 48:20emotional aspect of it as well let's say
- 48:2310 years down the line you get into the
- 48:25same issue
- 48:26>> oh yeah there's there's there's the CEO
- 48:28of Versel he made this post Giamo uh
- 48:31there's like if you're not looking at
- 48:33code anymore the one of these five
- 48:34things is true and he like listed about
- 48:37either you don't have customers or you
- 48:39don't have revenue or you are still
- 48:41experimenting or something.
- 48:42>> So even if 10 years down the line we
- 48:44reach a point where
- 48:46>> you know you don't have to look into the
- 48:48AI code review which we probably will.
- 48:50So that was my earlier point like right
- 48:51now we are heading in a direction where
- 48:54you don't have to review the code
- 48:56>> but the thing is that if something
- 48:57breaks
- 48:59the human is going to be in charge
- 49:01>> but don't you think that we have to
- 49:03understand the change still
- 49:06>> you do but it again depends on the
- 49:08context if I'm changing a button's color
- 49:10>> I'm not reviewing much
- 49:12>> okay so we're building a lot of tools
- 49:14here around helping you understand the
- 49:16blast radius of your changes like the
- 49:18obstruction layer of what you touch.
- 49:20Maybe you like you're doing a back end
- 49:22change and trying to surface that maybe
- 49:24your agent changed the front end file.
- 49:25So bring that to your intention instead
- 49:27of you having to dig through it or find
- 49:28out in production later.
- 49:30>> These type of things that's the way I
- 49:31see personally the the world going on
- 49:33that pace. And then next to that is also
- 49:36then of course the security part which
- 49:38>> obviously
- 49:39>> a lot of the number of security
- 49:41vulnerabilities that have increased in
- 49:43the age of AI is just so much so many
- 49:45secrets sprawls and whatnot. So yeah,
- 49:48depends on the blast radius. Um, but
- 49:51obviously in the end, whoever is
- 49:52approving the PR is a human.
- 49:54>> Yeah.
- 49:54>> So that person will be responsible when
- 49:56something, you know, the light LLM
- 49:58attack that happened as well. So
- 50:01>> that was incredible.
- 50:02>> That was kind of that was kind of crazy.
- 50:04>> Yeah.
- 50:04>> So I'm very pro AI. Um but am I
- 50:08comfortable right now just for
- 50:10it to run my whole systems without me
- 50:13being or my engineers being like okay
- 50:15hey
- 50:16>> we are the ones who are doing the final
- 50:17approval
- 50:19>> um no but it works hand in hand the
- 50:23review that the code has given
- 50:26>> read by human checks everything is fine
- 50:28merge the PR and if you're not using
- 50:31like AI code reviews to review code I
- 50:34mean why not cuz they're super fast and
- 50:35stuff you know why not use it. But in
- 50:37the end, it's a human taking the check.
- 50:40And to answer the point,
- 50:43um, 10 years down the line, it's a Black
- 50:46Friday sale, agent can solve the bug,
- 50:49but the person will go on the laptop and
- 50:51see themselves. You know, there's so
- 50:53much on the line. So, if you don't have
- 50:55paying customers, you don't care.
- 50:57>> Yeah. No, that's f well put. Uh, moving
- 51:00on the lines, I think another point that
- 51:02I had in my mental notes of things that
- 51:04I really wanted to get your talk on is
- 51:06you're running so many events all over
- 51:08the world, Europe, London, Paris, um,
- 51:12Berlin, San Francisco all the time,
- 51:14India, global events. So, you're very
- 51:17close to um, a lot of incredible
- 51:19builders that you just u were telling us
- 51:21about earlier as well. How do you see
- 51:24them building and how have you seen the
- 51:26shift of them using the tools like you
- 51:29know cloud using cloud was very
- 51:31different today than it is from a year
- 51:33ago and uh same goes with like some
- 51:36tools that we don't we stopped using
- 51:37maybe uh I remember for example windsurf
- 51:40having huge popularity last summer and
- 51:42then now it's kind of died down and I
- 51:44guess Devon has taken over but then as
- 51:46well like the approach of how we use
- 51:48tools it used to be
- 51:51taboo to kind of open massive PRs and
- 51:53people be like, "Oh, you should break
- 51:55that down and da and distribute it so
- 51:57people can understand." Now it's like
- 51:58you see on Twitter, you see big Ctos
- 52:02posting 400,000 lines of PRs, whatever.
- 52:05So, uh, tell me what how do you see the
- 52:08shift?
- 52:09>> Yeah, I mean in Europe,
- 52:11I don't think in US many people care
- 52:13about like data sovereignty as such. The
- 52:16US has a big uh argument around that.
- 52:19Even India with their um data production
- 52:21laws now um at our hackathons we mainly
- 52:25see people have now enough time because
- 52:27they have just one day to build in the
- 52:29in the event space you know so the the
- 52:33idea is how do you well how how well do
- 52:35you integrate the sponsor tools
- 52:37>> um in your project
- 52:38>> and I don't think many people are
- 52:40spending too much time on UI
- 52:42>> as much because they can just wipe code
- 52:44UI and UX every single UI we see is like
- 52:48incredible
- 52:49Um and uh something that would take like
- 52:53coding from scratch like I know you
- 52:54didn't code it from scratch. I know you
- 52:56used some sort of like AI tools for
- 52:58that. Um one of the cool thing that
- 53:00people do is like if they're working in
- 53:02a team they would divide the teams. They
- 53:04would like you use you work on the
- 53:06testing part. You you work on the review
- 53:08part. I'll work on the infrastructure.
- 53:11>> Um but one other one other aspect is is
- 53:14>> Are you saying they're not doing that
- 53:15anymore? They're doing that. They are
- 53:17doing that anymore. No, they are doing
- 53:19that
- 53:19>> at our I mean that's how I used to go to
- 53:21hackathons as a student all the time as
- 53:23well. Like I would build out the back
- 53:24end and then this other guy would build
- 53:26out the front end.
- 53:26>> Yeah. Cuz usually in the big inerson
- 53:28events they have many sponsors.
- 53:29>> Yeah.
- 53:30>> Right. So someone's doing this one
- 53:31thing, someone's doing monitoring,
- 53:33someone's setting up something else.
- 53:34>> Yeah.
- 53:34>> Um and then there's infrastructure. So
- 53:36where do I host it?
- 53:38>> And then somebody doesn't come through
- 53:39with their partner, you're screwed.
- 53:40>> Yeah. Someone's [laughter]
- 53:41someone's just uh doing uh you've seen
- 53:44this many times. Someone's just the
- 53:46marketing person. They're the ones
- 53:48>> you need them too though. They're very
- 53:49important. Very important then because
- 53:51then if can nobody can speak you don't
- 53:52win.
- 53:52>> They're they're telling you about the
- 53:54content about how the direction should
- 53:56be and about the
- 53:57>> video when they do the uh recording for
- 54:00the demo.
- 54:00>> Yeah.
- 54:01>> Um so we see also we also see cracked
- 54:03engineers. One person will come alone.
- 54:05>> Yeah.
- 54:06>> And they will win.
- 54:08>> So that's I think that's the main point.
- 54:10uh for that for that person to be
- 54:12working as a designer for that person to
- 54:14be working as a infrastructure person as
- 54:17a developer as a I don't think they
- 54:19write test during the hackathon really
- 54:22>> um
- 54:23>> but that one person is now doing all
- 54:24these things and that's made possible
- 54:26via via AI in 8 hours being able to
- 54:29delegate all these tasks so you can
- 54:31focus on what my theme is what my
- 54:33problem solving is what my
- 54:34>> more solo developers know more solo
- 54:36participants
- 54:37>> 50%
- 54:39That's crazy. It used to be like you
- 54:41people would get anxiety because they go
- 54:43to an event to a hackathon, they knew
- 54:44they had to find a teammate otherwise
- 54:46they would have no chance.
- 54:47>> In VM devs, uh 50 something% come by
- 54:51themselves.
- 54:53>> Wow.
- 54:53>> Yeah.
- 54:54>> So that's a that's a huge shift.
- 54:55>> Yeah. And that alone person can can do
- 54:59all these things in eight hours because
- 55:00of AI.
- 55:01>> So
- 55:03>> their main question is I need to well
- 55:05obviously win. I need to make cool
- 55:06projects. I need to integrate all the
- 55:08sponsor tools. I need to work on my idea
- 55:11and um presentation and whatnot and um
- 55:16and obviously how these integrate with
- 55:19each other. So it's not like they're
- 55:20delegating the entire thing to claude. I
- 55:22don't think that's going to work. It's
- 55:23just that the way they communicate has
- 55:25changed. They were writing code before.
- 55:26They're still writing code, reviewing
- 55:28code, but
- 55:30coding tools are doing most of the heavy
- 55:31lifting.
- 55:32>> Okay. Interesting. Uh what is maybe give
- 55:36me some some idea like what what type of
- 55:37tools do you see most popularly used on
- 55:39the market? Is everybody on claude or
- 55:41>> Yeah. A lot of people on claude. Codex
- 55:43is up there.
- 55:44>> Yeah.
- 55:44>> Um
- 55:45>> cursor. Any cursup users you see on your
- 55:47events?
- 55:47>> Cursor.
- 55:48>> Yeah.
- 55:49>> Not as much.
- 55:50>> No. Claude definitely and Codex.
- 55:53>> Yeah.
- 55:53>> In our events. Um and these are also not
- 55:57like sponsors like they just bring their
- 55:58own claw and codic setup.
- 56:01>> Um
- 56:02>> yeah.
- 56:02>> Interesting. Any review tools?
- 56:06>> Uh any review tools? [laughter] We did a
- 56:08hack.
- 56:09>> We did a hackathon with code rabbit.
- 56:11>> Yeah.
- 56:11>> I think um during hackathons people
- 56:14don't care much about code quality.
- 56:16>> Yeah. No, you just got to ship it.
- 56:19That's fine.
- 56:19>> You just got to ship it in 8 hours. Make
- 56:21sure it works.
- 56:22>> Make sure make sure it works for the
- 56:23demo. That's it. Breaks afterwards.
- 56:25That's fine. Then you onboard the users
- 56:27and everything crashes. But yeah, it's
- 56:28all right.
- 56:28>> They're not they don't want they don't
- 56:30really care that much about code
- 56:31quality. They just want to ship ship
- 56:33fast. Care about UI a lot because that's
- 56:35you know
- 56:36>> it looks cool so the judges are happy.
- 56:38They care about deploying so people can
- 56:40use it easily. Reversal is popular.
- 56:42Versel is quite popular. Yeah.
- 56:43>> No that's cool. Um how do you see I
- 56:47guess in the in the trend of all of
- 56:48these people that are participating?
- 56:51Are there any patterns that you see and
- 56:54folks that come to your events where
- 56:55like okay these people they might be
- 56:57coming quite successful in what they're
- 56:58going to do?
- 57:00You can definitely see seasoned uh
- 57:03hackathon people at events. One trait
- 57:06that I have seen uh is um when you have
- 57:09uh let's say three or four sponsors in
- 57:11hackathon think about online or in
- 57:13person and you have a limited amount of
- 57:14time.
- 57:14>> Yeah.
- 57:15>> Number one mistake I see people make who
- 57:16don't win is they would have a lot of
- 57:18features in line.
- 57:19>> Yeah.
- 57:20>> And they would be like I'm going to imp
- 57:21implement all these and whatnot. The
- 57:24person who wins
- 57:27build a minimum viable product first.
- 57:30Very simple UI, simple project that
- 57:32narrate that narrates the idea they had
- 57:34on initially.
- 57:35>> Mhm.
- 57:36>> Has a working feature, the main feature,
- 57:39simple UI, integrates all the tools,
- 57:42save that as version one for backup, and
- 57:44then they build on top of that. This
- 57:46helps a lot with stress because if
- 57:48you're building something for 8 hours,
- 57:51it you will get stressed a little bit
- 57:53because the time is ticking.
- 57:54>> This helps with stress quite a bit.
- 57:56>> Yeah.
- 57:57>> Um and the other thing is make it easy
- 57:59for people to judge your project.
- 58:01>> The some people will have like a signup
- 58:03page or they would ask the people to
- 58:05like pull up a Docker image. The judge
- 58:07is not pulling a Docker image to run
- 58:08your project.
- 58:09>> You know, [laughter] keep the demo
- 58:11open-ended.
- 58:11>> Yeah. And the last thing would be um
- 58:16the idea does not matter that much.
- 58:18>> Oh really?
- 58:19>> Yeah. I have because I have seen people
- 58:22with the simplest ideas
- 58:24>> and perfect execution win hackathons.
- 58:27>> Okay.
- 58:27>> So it's not like
- 58:28>> what's perfect execution look like?
- 58:30>> It just works and it's seamless and it's
- 58:32good user interface and UI UI and stuff.
- 58:35There was one hackathon where it was
- 58:36like
- 58:37>> there was just a lot of Chrome, you know
- 58:39how there are a lot of Chrome tabs and
- 58:40stuff. One project person built like
- 58:42some extension to manage all those
- 58:44external code.
- 58:45>> Oh, I have a friend of mine. He needs
- 58:47that package. Man, this sounds crazy.
- 58:49>> Simple simple tool, simple execution,
- 58:51won the won the project.
- 58:53>> Uh I don't know which hackathon was
- 58:54this.
- 58:55>> Yeah.
- 58:55>> So idea matters.
- 58:57>> Uhhuh.
- 58:58>> It's not like a crazy thing that's going
- 59:00to let that's going to make you win. I
- 59:02mean in my idea like in my world and
- 59:04perception and I've I've done a fair
- 59:06amount of hackathons to to to say that
- 59:08but I felt like the idea was getting
- 59:11more and more important because I feel
- 59:13like the the problem solving skill the
- 59:16um idea of focusing your attention on
- 59:18something that brings value to people
- 59:21and is kind of somewhat novel or breaks
- 59:22the pattern even um is something that is
- 59:25more and more important as the execution
- 59:27gets easier and easier because of the AI
- 59:29coding agent. Don't you agree? Yeah, but
- 59:32if you want to win the judges at these
- 59:36if you go to like a sponsored event
- 59:38>> and [clears throat] you want to win a
- 59:38sponsored track prize, they want to see
- 59:41how well you integrated the sponsor
- 59:42tools.
- 59:43>> Oh,
- 59:43>> and these seasoned hackathon people know
- 59:45that.
- 59:46>> Yeah.
- 59:46>> So, the idea does matter. It's just that
- 59:48if it's purely just the idea
- 59:51>> and the execution is not there.
- 59:53>> Yeah.
- 59:53>> Then um you're not going to win.
- 59:56>> Okay.
- 59:57>> But the main thing is about if you're if
- 59:58you're learning I'm going to say
- 59:59something cringe. If you're learning
- 1:00:00something new and you're having fun,
- 1:00:02you're a winner.
- 1:00:03>> I mean, that's true.
- 1:00:04>> Um, but these people are seasoned
- 1:00:06hackathon people. You'll see at big
- 1:00:07prizes.
- 1:00:08>> If a hackathon has like 100k in prizes,
- 1:00:10you'll see a lot of people who are
- 1:00:11seasoned hackathon professionals.
- 1:00:12>> Okay.
- 1:00:13>> They have their own setup, whole
- 1:00:14templates and everything.
- 1:00:15>> Sharks are these guys that come with
- 1:00:17their own keyboard to the event with
- 1:00:18their little stand or bring a second
- 1:00:20monitor.
- 1:00:20>> They bring a second monitor and uh
- 1:00:23>> their ideas are always simple.
- 1:00:25>> Yeah.
- 1:00:26uh nothing too crazy and uh we don't
- 1:00:29have themes in our hackathons as such we
- 1:00:31keep it open-ended we can do anything
- 1:00:33>> but obviously if it's a themed hackathon
- 1:00:34like hey build the best uh project for
- 1:00:37city infrastructure
- 1:00:39>> obviously then how unique your idea is
- 1:00:40matters
- 1:00:41>> but idea alone won't do anything if you
- 1:00:44are not executing it well someone with a
- 1:00:47boring idea and perfect execution will
- 1:00:48win over your great idea and lack of
- 1:00:51execution [clears throat]
- 1:00:52>> well I guess there's a tangent that you
- 1:00:54can to the professional world as well.
- 1:00:56If your boss gives you a task and you
- 1:00:58you do like uh the craziest idea of how
- 1:01:02to solve it and the craziest angle that
- 1:01:03has super much potential but you don't
- 1:01:05execute well
- 1:01:06>> ship fast. This is what Zuckerberg also
- 1:01:08did with the meta
- 1:01:09>> the meta glasses thing. The whole AI
- 1:01:12universe
- 1:01:15he was doing and people were like oh no
- 1:01:17his demo failed
- 1:01:18>> and whatn not you know Zuckerberg's demo
- 1:01:20failed and should have waited and this
- 1:01:23and that. No, it's fine.
- 1:01:26>> Yeah,
- 1:01:26>> build, ship, iterate on it. It's good.
- 1:01:30>> Well, that's true. I mean, it's also a
- 1:01:31fair statement to make when you have the
- 1:01:33entire distribution engine of the world
- 1:01:34on your product. But
- 1:01:35>> Exactly.
- 1:01:36>> But yeah, actually that's that's true.
- 1:01:38>> If it's like a smaller company, that
- 1:01:40would backfire a little bit.
- 1:01:42>> Yeah.
- 1:01:42>> Which is why minimum vi minimum viable
- 1:01:44product that you know will work.
- 1:01:46>> Mhm.
- 1:01:46>> And any other fancy other features you
- 1:01:48want to add, just do it later.
- 1:01:50>> Yeah.
- 1:01:50>> But yeah, key is moving fast. Ship fast,
- 1:01:52move fast, break things.
- 1:01:54>> Yeah.
- 1:01:54>> Learn from your users.
- 1:01:56>> Great. Um, yeah, maybe starting to to
- 1:01:59wrap up a little bit. I I'd love to
- 1:02:01finish this wonderful conversation on a
- 1:02:03couple of encouraging notes and
- 1:02:05thoughts. Um, you just mentioned a few
- 1:02:07of them. Ship fast, learn fast. You
- 1:02:09know, this entire we hear that a lot.
- 1:02:12Uh, do you have some practical advice
- 1:02:13for people who might be listening
- 1:02:14thinking like, hey, um, you know, some
- 1:02:16of the things you just said really
- 1:02:18resonated with me. I'm in that spot
- 1:02:19where I'm somewhat in my career but not
- 1:02:22truly AI native yet and I don't feel
- 1:02:24like I need to get somewhere. Where
- 1:02:25where do people pick up? What do I do?
- 1:02:28>> Yeah, if you're looking for your next
- 1:02:30role, I would say list down what kind of
- 1:02:33role do you want?
- 1:02:34>> It's your ideal company look like what
- 1:02:36does that role look like? Like what does
- 1:02:37the person need to do? And then see what
- 1:02:39kind of AI tools are being used in that
- 1:02:41role.
- 1:02:42>> If you're just starting out, yeah, the
- 1:02:44advice for you is very simple. It does
- 1:02:46not matter. Go get started. Get a job at
- 1:02:48the Apple store, learn everything about
- 1:02:50the Apple ecosystem, build Apple apps,
- 1:02:53then go to your next job, right? Does
- 1:02:56not matter. The question around which
- 1:02:57tool I get started with, language do I
- 1:02:59get started, does not matter.
- 1:03:01>> Just literally doesn't. Just get started
- 1:03:03with building.
- 1:03:04>> Um, that's my recommendation for juniors
- 1:03:07like students.
- 1:03:08>> Yeah,
- 1:03:09>> just get started with building and as
- 1:03:10you move forward, just experiment with
- 1:03:12different tools, expand your horizon.
- 1:03:15>> Yeah. No one expects you to be an expert
- 1:03:17when you're a student 17 18 years old.
- 1:03:19Um but what they do expect you to do is
- 1:03:23have
- 1:03:24enough drive and sort of like they want
- 1:03:26to see what has this person done based
- 1:03:29on the resources that they already have
- 1:03:30available. So I go to hackathons meetups
- 1:03:34speak at conferences learning in public
- 1:03:36which we mentioned earlier.
- 1:03:37>> It can be a little bit awkward speaking
- 1:03:39at conferences
- 1:03:40>> intimidating
- 1:03:41>> intimidating but it's going to help you.
- 1:03:43Yeah,
- 1:03:43>> you know, and these conferences have
- 1:03:45special tracks for first- time speakers.
- 1:03:47>> So, that's working in your favor.
- 1:03:50>> And if you're a senior engineer, um, mid
- 1:03:53to level engineer,
- 1:03:54>> I mean, you're already there. You're
- 1:03:55already using AI in your sort of like
- 1:03:58workflows.
- 1:03:59>> Mhm.
- 1:03:59>> But now you have more time for problem
- 1:04:02solving and now you can shift faster.
- 1:04:04>> So, it's truly about focusing on this
- 1:04:06product mindset, I think, is what I'm
- 1:04:07hearing.
- 1:04:08>> Product mindset, shift fast, use AI to
- 1:04:10achieve all the tasks and goals.
- 1:04:12>> Yeah. But make sure you are
- 1:04:15>> you're not completely relying on it.
- 1:04:17>> Yeah. um using it as a helping hand and
- 1:04:19when you're using AI to review code
- 1:04:21tools and whatever
- 1:04:22>> um yeah do a final check because if
- 1:04:25something messes up you're the one who
- 1:04:27is responsible and one one thing I
- 1:04:29mentioned earlier in the three points
- 1:04:31that I was mentioning when AI is writing
- 1:04:33code this is also true for junior
- 1:04:35engineers
- 1:04:37um when AI is writing code especially if
- 1:04:39you're a student and AI can build the
- 1:04:40whole website for you that is great I
- 1:04:43think you should do it but whatever the
- 1:04:46AI I has written make sure you read
- 1:04:48through that and understand and learn
- 1:04:50from that
- 1:04:50>> so that you are not just
- 1:04:52>> a code monkey.
- 1:04:53>> Yeah,
- 1:04:54>> that's the word. So if you're a student,
- 1:04:56use AI, use clot to code
- 1:04:58>> 100%. But what it's generating after
- 1:05:00it's done doing it, you can ask it
- 1:05:02followup questions like hey explain to
- 1:05:04me how this works, what function you
- 1:05:06wrote. That's an interesting segue into
- 1:05:09uh I mean also what you talked about the
- 1:05:11next frontier of a engineering not being
- 1:05:13in the code writing anymore but in the
- 1:05:15architecture level and truly
- 1:05:16understanding as a software engineer
- 1:05:18like what what is my next big thing that
- 1:05:21I should learn study where do I bring
- 1:05:23the most value today.
- 1:05:24>> Yeah. Um
- 1:05:28how to
- 1:05:29if there's one thing I would say is
- 1:05:32learn how to run LLMs locally.
- 1:05:35Oh wow, that's a that's a good one.
- 1:05:37That's a that's so
- 1:05:38>> that's one thing that learn how to run
- 1:05:40LLMs locally,
- 1:05:41>> learn how to fine-tune models locally.
- 1:05:44>> Um and uh how do you transition that
- 1:05:48pipeline
- 1:05:49>> uh at scale?
- 1:05:51>> What kind of because that's where the
- 1:05:52people are headed. you know a lot of
- 1:05:54talk around uh uh running AI locally in
- 1:05:58secure environments and there are any
- 1:06:01number of things you can do to run that
- 1:06:04to run LLMs on like infrastructure so
- 1:06:06for example when Kubernetes came out
- 1:06:08there's so many ways by which you can do
- 1:06:09it can have your own virtual machines
- 1:06:11and Kubernetes on top of that you can
- 1:06:13use managed Kubernetes platforms like uh
- 1:06:15EKS or whatever you can run it on that
- 1:06:17you can use all these neo clouds and
- 1:06:19smaller providers you can do that so
- 1:06:20when you go to a job and the boss asks
- 1:06:22you like you're just starting out and
- 1:06:24hey we want to run LLMs we want to use
- 1:06:26influencering tools we have 100 options
- 1:06:28which one is the best so now if you have
- 1:06:30the whole industry landscape idea
- 1:06:32>> like hey you can be like we can just use
- 1:06:33an API no we won't use an API we'll
- 1:06:36we'll run it on our own hardware like
- 1:06:38this no we won't use that we're going to
- 1:06:39use Cerebras we're not going to use
- 1:06:41Cerebras which is going to pay uh codecs
- 1:06:43a monthly bill or whatnot
- 1:06:45>> and um
- 1:06:47>> yeah I mean you know about it like
- 1:06:49running cubernetes there are so many
- 1:06:50ways to do it the same thing with AI I
- 1:06:52mean
- 1:06:52run learn how to run AI models locally.
- 1:06:55The other thing is because this is the
- 1:06:57most basic thing the fundamental it's
- 1:06:59like running a model on a machine is a
- 1:07:01fundamental.
- 1:07:02>> So it it brings us back to your earlier
- 1:07:05point
- 1:07:06>> the basics.
- 1:07:07>> Yeah.
- 1:07:07>> So it will teach you a lot
- 1:07:10>> um which in itself is a lot of uh
- 1:07:15>> valuable
- 1:07:15>> there's you know the book uh not the
- 1:07:17book the repository Cubanet is the hard
- 1:07:18way by Kelsey High Totower.
- 1:07:20>> Mhm. So if you're running Kubernetes
- 1:07:22obviously in production you're let's say
- 1:07:25billion dollar company you're most
- 1:07:26probably using AWS
- 1:07:28um
- 1:07:30like Netflix uses AWS but those people
- 1:07:33know how to run Kubernetes like on mini
- 1:07:36cube or locally or in their own little
- 1:07:38cluster right so that learning journey
- 1:07:42that you get you get to see things
- 1:07:44inside
- 1:07:44>> and then you just use a managed service
- 1:07:46to run that at scale
- 1:07:47>> okay
- 1:07:48>> so learn how to run LLMs locally It's a
- 1:07:50lot of fun and you'll learn a lot.
- 1:07:53[laughter]
- 1:07:53>> You're going to see how it how it uh how
- 1:07:55the networking works, how it interacts
- 1:07:57with the
- 1:07:58>> with other agents, how you run multiple
- 1:08:00agents.
- 1:08:01>> Um and then you just sort of like see
- 1:08:02okay, how do I scale this now? And uh
- 1:08:04yeah, but it's uh always comes down to
- 1:08:07learn the basics.
- 1:08:08>> Yeah. how things are working and when
- 1:08:09you work with more complex tools um
- 1:08:12whatever the complex tool gives you ask
- 1:08:15it look at it study from it especially
- 1:08:19if you're a student study from it
- 1:08:20>> and then implement that
- 1:08:22>> yeah an interesting quote comes to mind
- 1:08:25here I had Simon the founder founder of
- 1:08:27assistant UI here not too long ago and
- 1:08:29he said cloud can get you 80% there
- 1:08:32today but all the alpha is in the
- 1:08:34remaining 20%
- 1:08:35>> and those that 20% determines who gets
- 1:08:38hired.
- 1:08:39>> Yeah. [laughter]
- 1:08:40>> Wonderful. Well, Canal, it's been an
- 1:08:42amazing conversation. Thank you so much
- 1:08:44for taking the time.
- 1:08:45>> This was a lot of fun.
- 1:08:46>> Yeah. I have a a little bit of a a
- 1:08:48ritual that I do with every guest. So, I
- 1:08:49have a couple of rapid fire questions uh
- 1:08:51that I'm going to ask you uh now if
- 1:08:53you're okay with it.
- 1:08:54>> Yeah. Rapid fire.
- 1:08:55>> Yeah. And you can answer them uh either
- 1:08:57short or if you want to expand that's
- 1:08:59also fine. Okay.
- 1:09:00>> Yeah, sure.
- 1:09:01>> All right. Favorite programming
- 1:09:02language?
- 1:09:04>> Golang.
- 1:09:05>> Golang.
- 1:09:05>> Yeah. I just love Golang. Let's just
- 1:09:07keep it there. I don't want to start a
- 1:09:08debate.
- 1:09:08>> Okay.
- 1:09:09>> Actually, shout out to Rust.
- 1:09:10>> Yeah.
- 1:09:11>> Um, shout out to Java also because my
- 1:09:13career started with Java, so it's
- 1:09:14emotional.
- 1:09:15>> Okay. All right. Good. We'll keep it
- 1:09:16there. Uh, favorite AI model.
- 1:09:20>> Fable five.
- 1:09:21>> Favorite five. [laughter]
- 1:09:23>> So good. It's very expensive.
- 1:09:24>> Yeah. But it's good. Um, cool. Uh,
- 1:09:27favorite coding harness?
- 1:09:30>> Cloud.
- 1:09:31>> Cloud. Yeah. Okay. Cloud code.
- 1:09:33>> Cloud code.
- 1:09:34>> Yeah. Like like that. Okay. All right.
- 1:09:36No open code, no nothing like that.
- 1:09:37>> I like cloud code.
- 1:09:38>> Okay, that's fair. I guess if I feel
- 1:09:41like also now that they like obviously
- 1:09:43they train they RL the model on the
- 1:09:45harness so it just gets better and
- 1:09:46better.
- 1:09:47>> I like the cloud community. I think
- 1:09:48that's another thing like the community
- 1:09:50is quite there's a lot lot less noise. I
- 1:09:54would say they just these people are
- 1:09:55just building and shipping.
- 1:09:56>> Where do you connect with the cloud
- 1:09:58community?
- 1:09:59>> Reddit.
- 1:09:59>> Mhm. and uh LinkedIn and there's a lot
- 1:10:03of pe we have seen claude very popular
- 1:10:04in vm make devs so a lot of our users
- 1:10:07when they get stuck somewhere they have
- 1:10:09a question
- 1:10:10>> probably I'm like what are you using
- 1:10:12claude codex is up there as well those
- 1:10:14are the top two
- 1:10:15>> okay what do you do when you don't code
- 1:10:17or or run your business
- 1:10:19>> um
- 1:10:20I'm a foodie
- 1:10:22>> love trying out food
- 1:10:24>> um I'm full-time on desk now so I'm not
- 1:10:26I'm working for myself
- 1:10:28>> um
- 1:10:29>> left my job in June. June was my last.
- 1:10:31>> Okay.
- 1:10:31>> So, I'm like, okay, we are scaling. We
- 1:10:33have close to two 250,000 people now.
- 1:10:36So, I'm like, I'm just going to do my
- 1:10:37own thing.
- 1:10:38>> Yeah.
- 1:10:38>> So, I thought if for someone who thinks
- 1:10:41like, okay, you're doing your own thing,
- 1:10:42you have all the time in the world. Now,
- 1:10:43it's like a full-time. I'm now doing I
- 1:10:45think I'm working more now
- 1:10:46>> than I was before.
- 1:10:48>> Yeah.
- 1:10:48>> Um, but yeah, I am a foodie, so I love
- 1:10:50trying new restaurants,
- 1:10:52>> love shopping.
- 1:10:53>> Yeah.
- 1:10:53>> For silly stuff. Um, just like toys and
- 1:10:56whatnot. Um,
- 1:10:59TV shows, love TV shows, movies. I don't
- 1:11:03go to cinema, but I watch it at home.
- 1:11:05>> Uh, spending time with friends and
- 1:11:07family is something I love quite a lot.
- 1:11:09>> Yeah.
- 1:11:09>> And, um, because I'm traveling a lot,
- 1:11:11anytime I get, I just relax. Uh, I live
- 1:11:14in London, so there's a lot of things to
- 1:11:16do like musicals and I don't drink, so
- 1:11:18not really pub kind of a I know you're
- 1:11:21from Germany. I don't know if you drink,
- 1:11:22but I have a I have German friends and
- 1:11:24>> Yeah. No, we it's part of the culture.
- 1:11:26drink the whole picture and he would not
- 1:11:27get drunk. I'm like, "How are you not,
- 1:11:29dude? How are you not drunk?" [laughter]
- 1:11:30He's like, "It's just beer."
- 1:11:32>> Yeah. Yeah.
- 1:11:32>> Beer doesn't make me drunk, right?
- 1:11:34>> It's water for us at this point.
- 1:11:35>> And I'm like, so I'm like, "If you want
- 1:11:36to get drunk, you have to drink hard
- 1:11:38liquor." He's like, "Yeah, beer I can
- 1:11:39drink as much as I want. I won't get
- 1:11:41drunk." I'm like, "Wow." Super
- 1:11:42>> that. I don't know about that. I don't
- 1:11:43want
- 1:11:44>> Yeah. But he's like, he drank this whole
- 1:11:45picture of beer.
- 1:11:46>> Yeah.
- 1:11:47>> He [laughter] was fine.
- 1:11:48>> Shout out to Daniel.
- 1:11:49>> Yeah.
- 1:11:50>> Uh,
- 1:11:51yeah. Know I'm a foodie and I love uh I
- 1:11:55think My life revolves around my family
- 1:11:57and friends.
- 1:11:58>> Yeah.
- 1:11:58>> So, just spending time with them no
- 1:11:59matter what we do
- 1:12:01>> is what I love. And uh yeah, foodie,
- 1:12:05movies, video games, Call of Duty, um
- 1:12:10running,
- 1:12:11badminton.
- 1:12:12>> Yeah,
- 1:12:13>> shopping.
- 1:12:14>> Nice. That's a lot. Yeah,
- 1:12:16>> that's it. Keeps me busy.
- 1:12:17>> Thanks for sharing that. Uh I think this
- 1:12:19is kind of the last point where I wanted
- 1:12:20to give you the opportunity if you want
- 1:12:22to say anything to the audience here
- 1:12:23that you got coming up. um anything
- 1:12:25people where can people follow you?
- 1:12:27Obviously you got your YouTube channels
- 1:12:28and so on and um yeah that's basically
- 1:12:31it.
- 1:12:32>> Yeah I'll just say two things. The first
- 1:12:34thing I would say is uh go to
- 1:12:35vmakedevs.org
- 1:12:37and uh check out all our events
- 1:12:39hackathons you can sign up and create an
- 1:12:41account. We have built a cool platform.
- 1:12:43we're going to do more cool stuff and if
- 1:12:44you're a company who wants to get your
- 1:12:46tool in the hands of like 200,000
- 1:12:49engineers,
- 1:12:50we have an incredible community and they
- 1:12:52would be more than happy to you know uh
- 1:12:55give you feedback, build with it. Uh the
- 1:12:58third last point I would say is don't
- 1:13:00worry it's going to be fine. That's what
- 1:13:02I would say cuz in the in the end life
- 1:13:05just works out and uh
- 1:13:08don't worry about things that are not in
- 1:13:10your control. AI is coming. AI is
- 1:13:12replacing jobs. AI will replace some
- 1:13:14fields. It's not in your control. Then
- 1:13:16why are you worrying?
- 1:13:17>> Mhm.
- 1:13:18>> You know, I mean it sounds silly.
- 1:13:21>> Yeah.
- 1:13:22>> Someone is losing their job and I'm like
- 1:13:23don't worry,
- 1:13:24>> you know. Can't you can't say that.
- 1:13:26>> But the thing is if something is not in
- 1:13:28your control, what can you do?
- 1:13:30>> But wouldn't you say that like let's say
- 1:13:33I'm I'm we just talked about this and uh
- 1:13:35I feel like we wanted to wrap up and now
- 1:13:37I need to go back into it. Um
- 1:13:39>> go for it.
- 1:13:40>> Yeah. You
- 1:13:43There's always something you can do.
- 1:13:45>> Exactly.
- 1:13:46So if I'm about to like if I'm on the
- 1:13:49verge and I feel like oh my relationship
- 1:13:51at work with my manager is not going so
- 1:13:53well like my performance is apparently
- 1:13:55not looking up so much like
- 1:13:57there need there should be something I
- 1:13:59can do about this like you know go watch
- 1:14:00your tutorials go come to one of your
- 1:14:02hackathons go like get code rabbit to
- 1:14:05review your code so you don't get any
- 1:14:06burn like all these things right so um I
- 1:14:10think that that that there's almost for
- 1:14:12anyone who's trying to who's not
- 1:14:14completely strangled and something or
- 1:14:16maybe works in a huge tech firm where
- 1:14:18they just get cut off. Even then,
- 1:14:19there's something they can do which is
- 1:14:21train yourself up and go look for the
- 1:14:23next opportunity. In fact, if your
- 1:14:25company if you feel like that right now,
- 1:14:26I think I should argue you should be
- 1:14:27proactive and try to get something
- 1:14:29somewhere else already, right?
- 1:14:30>> Yeah. It's it's it's like so life is not
- 1:14:33fair to everyone.
- 1:14:34>> Yeah.
- 1:14:34>> There's always hardships, you get laid
- 1:14:36off
- 1:14:37>> and um sometimes, you know, things are
- 1:14:40not in your control. This is a very
- 1:14:41personal advice. I understand it's very
- 1:14:44hard to follow for a lot of people.
- 1:14:46>> Human beings are wired differently. But
- 1:14:48this is how I am wired.
- 1:14:49>> Okay.
- 1:14:50>> If something is not in my control, it
- 1:14:52does not make me sad.
- 1:14:53>> Mhm.
- 1:14:54>> Worse things could happen to me in my
- 1:14:55life. But if it's not in my control,
- 1:14:57there's no point of worrying.
- 1:14:58>> Yeah.
- 1:14:58>> You know, uh but that's it has taken me
- 1:15:01some time to sort of like but there's
- 1:15:04always something you can do. There's
- 1:15:05always uh people are always looking for
- 1:15:07cool people. So people are always
- 1:15:09looking for good people.
- 1:15:11>> You can always do freelancing. You can
- 1:15:12always upskill yourself. There are
- 1:15:14people getting hired out there. There
- 1:15:17are people getting jobs out there. The
- 1:15:20question is just why are they getting
- 1:15:21hired? Why am I not getting hired? So
- 1:15:23obviously there's one good thing is
- 1:15:25because of all these AI resources, the
- 1:15:28barrier to entry has shrked quite a bit.
- 1:15:30So for you to study, for you to learn,
- 1:15:32for you to research, for you to upskill
- 1:15:33yourself, it's very easy now. I mean you
- 1:15:35have so many options.
- 1:15:36>> So silver lining I would say that's a
- 1:15:39silver lining. But uh if you want my
- 1:15:41personal advice, things that are not in
- 1:15:43your control, don't worry about it. Make
- 1:15:45sure you give your best. I'll give one
- 1:15:46one last example. Let's say John Wick
- 1:15:48has an interview and Spider-Man has an
- 1:15:50interview for the same job.
- 1:15:53>> John Wick is very focused. John Wick is
- 1:15:54spending 100% of his time
- 1:15:56>> preparing for the interview. Spider-Man
- 1:15:59is spending 50% of his time preparing
- 1:16:00for the interview, 50% worrying about,
- 1:16:02oh, what would I do if I don't clear it?
- 1:16:04And let's imagine they both have the
- 1:16:06same intellect.
- 1:16:08>> Who do you think has a higher chance of
- 1:16:09doing
- 1:16:10you know.
- 1:16:10>> Yeah. I mean, you can also o overprepare
- 1:16:13that also happens. But, uh, yeah.
- 1:16:15>> Yeah. Things that are not in your
- 1:16:16control, don't worry. And then, yeah,
- 1:16:17that's it.
- 1:16:18>> Okay. I think we'll leave it at that.
- 1:16:20Thank you so much for coming on the
- 1:16:21show, Canal.
- 1:16:22>> Thanks everyone for tuning in and um,
- 1:16:24yeah, won't be the last time that you
- 1:16:25see us. I think
- 1:16:26>> I will be back. Great.
About this transcript
This page contains the full transcript of Which Skills Will Matter Most for Software Engineers? | Kunal Kushwaha by CodeRabbit, generated from the public captions YouTube serves with the video. The transcript has 15,562 words across 2,400 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.