Gaurav Sen · Gaurav Sen · Transcript & summary by WhipScribe
It gives working engineers a short, practical answer to whether AI makes their design and coding experience obsolete, and what to learn next.
Tap a chapter to open that moment in the transcript; the link to it is copied for you.
“Focus should shift on the design and the decisioning elements.” Arvind · at 7:01 — open in the transcript →
Gaurav Sen hosts Arvind of the Gen Academy, a teacher of AI engineering who credits Sen's system design videos with helping him pass interviews, to answer the question engineers keep asking: how to stay relevant over the next decade of AI tools. Arvind's first advice is to tune out the doomsday narrative and learn the basics, from what a language model does to terms like context window, temperature, top p and RAG. Sen adds that most of the jargon hides simple ideas, such as context engineering meaning better input gives a better answer.
The core argument is that writing code is no longer the scarce skill. Companies that handed Cursor to every engineer now struggle to decide which generated code to trust in production, and Sen saw the same flattening at his India AI hackathon, where entries built with AI looked alike. Both say value moves upstream to design, testing and judgment, which depend on context and taste that experienced engineers carry.
For next steps, Arvind compares AI to an unmapped jungle, unlike software engineering, which he says took about 30 years to harden into known career paths. His two rules are to build without waiting for permission and to build in public so a record of your thinking exists. Sen closes with a Meta anecdote about engineers being asked to think like influencers.
An AI engineering instructor at the Gen Academy, which he runs with Aishwarya, with popular Maven courses on AI engineering, evaluations, system design and MCP. Gaurav Sen says his community has more than 25,000 engineers.
Yes. Arvind of the Gen Academy argues that because AI now generates code quickly, an engineer's focus should move upstream to design, architecture and decision-making. He says the context for those choices lives in experienced engineers' heads, so tenure and taste matter more, not less. Gaurav Sen agrees that AI's gains are mostly in coding, while architecture and foresight remain the engineer's expertise.
Arvind of the Gen Academy recommends basic AI literacy before anything else: what a large language model is, how it handles language, and the terms that get thrown around, such as context window, temperature, top p and RAG. He says playing with chatbots gives only a surface sense of their abilities, while understanding what drives them lets engineers get meaningful output and judge what AI can and cannot do.
Gaurav Sen says a vector database is just a database with its own kind of indexing and storage, holding a different kind of data. Context engineering, in his telling, means the quality of an LLM's response rises with the context you feed into it. He describes loop engineering, a newer term, as working out how to orchestrate those calls the right way.
Arvind argues that engineers who already code gain the most from AI because, beyond writing code faster, they know how systems are built and can tell good code from bad. He says many companies that gave Cursor to all their junior and senior engineers now face so much generated code that they struggle to decide what to trust and promote to production, and few people can judge that output.
Gaurav Sen ran an India AI hackathon for tier two and tier three cities and villages, asking teams to use AI to save doctors' time where there is about one doctor per 11,000 patients. He says the submissions all seemed the same: AI helped the teams but also cut down what made each one unique, which he sees as a missing sense of taste and evaluation.
Arvind of the Gen Academy says AI careers have no mapped roads yet, unlike software engineering, which took about 30 years to harden. His two rules are to stop waiting for permission and build on your own strong point of view, and to build in public. Because everyone from individual engineers to enterprises is asking the same questions, public work creates a track record for your thinking.
Our own transcript of the public episode, with a timestamp about every 90 seconds. Tap a timestamp or the link button on a paragraph to copy a link that opens this page there.
0:00We have a special guest with us. It's Arvind, who's from the Gen Academy. If you're a software engineer who's looking to move to a role like forward deployed engineer or a solutions architect, or just looking to upskill so that you can perform better at work, I think this podcast will be useful. So welcome, Arvind. Thanks, Gaurav. And, you know, pleasure to be talking to you. I wouldn't be lying if I said that you helped me crack a few interviews with your system design videos back in the day. So Arvind, the most common question that we get is, I'm a software engineer. I know how to build systems. I know how to write code. But in the age of AI, there's a lot of new tools which are coming in.
0:37There's a lot of new things that are being expected of me from my management or even from myself. What do I do so that I can stay relevant for the next five to ten years? First off, it's about disengaging from the AI doomsday narrative, right? Because that tends to evoke a lot of anxiety in people where it's not necessary. In reality, what you really need to be doing is, as you said, right, if you haven't ever, like, you know, sort of interfaced with one of those tools that are very, let's say, to build something meaningful or you just have fundamental questions or what is an L and what is an AI and all that you need to get into that space and learn those fundamental concepts right so essentially what I'm trying to say is
1:13put on more of a learner's hat and just spend a little bit of time just understanding the space understand the foundational building blocks once you have that in place then you're able to sort of see the next steps. And as you learn, I think it becomes very essential for you to think about how to integrate that or augment your current workflows with those learnings. Yeah. I think one of the interesting things that we are having a conversation before we turn on the camera and the way Arvind broke it down, I think is one of the best ways I've seen a person break down the problem. The theory isn't that hard it's not like you need to do research right before you get in what do you think? Yeah it's very applied right the theory is not that hard
1:57terminologies are usually very scary but it's all pretty intuitive and easy to understand by going and playing around with tools like co-work chat gpt and the likes of it you're going to get one level of understanding you get like a sense for the capabilities but for you to really dig deep and take advantage of these tools in a meaningful way you need to develop an intuition of what's driving it right like what's creating all that magic and you know that's that's why i say that go and learn those fundamentals basic ai literacy right what is an llm how does it comprehend language and how is it able to do these things familiarize yourselves with you know terminologies that get thrown around you mentioned a few context window temperature
2:40top p rag and all these things it's important for you to just like understand what they are and then you'd find yourself in a better position to leverage these tools and actually get meaningful output from them. If you look at it in terms of a manager or somebody who's looking to upskill their team, the way they come to us is please tell us exactly what we need to know so that our existing teams with smart engineers can make the most out of these new tools. And it's simple. It's like, okay, we start with AI literacy. What's a vector database? Well, it's a database which has its own kind of indexing and storage. That's it. The kind of data that's being stored is different. And then you go, oh, okay, what's context engineering?
3:20Well, the LLM is going to take an input and give you a response. The quality of the response will go up based on what kind of context you can pump in that's it that's pretty much context engineering and now there's this new term being thrown around which is loop engineering which if you look at it it's more around okay how do i orchestrate this the right way so like you said basic knowledge is out there now is it really true that with just this little basic knowledge they are going to be able to do well what is it that is helping engineers perform well at work? So we're talking about just engineers, right? Oh, yeah. Maybe even in general, yeah. Obviously, there are a lot of things that go into it. But at the fundamental level,
3:57it's always going to be about understanding the concepts and applying them, right? That's how learning works. And if you are in a place where you don't understand the foundations then it's imperative that you go and like you know take a stab at understanding them right a lot of free content a lot of like you know sort of in fact you can just talk to chat gpt and whatnot and get up to date on it and then try and find ways to apply it right and applying it could mean building a small like you know demo project a prototype or whatnot or it could also mean integrating if you're working on a live project that you know has the scope for all of these things maybe start thinking about what are different ways to do it especially for engineers
4:32today a lot of people are talking about coding uh being sort of democratized and you know the software engineering job profile and threat and so on but then in fact i think these guys the engineers who are coding today benefit the most from these tools because not only can they obviously accelerate the coding that almost anybody can today, but they uniquely have a view into how systems are built, what good code looks like. They're able to tell what good code looks like versus bad code. In fact, most of the companies we speak to have this problem. The senior engineers are asking, hey, okay, we've given our engineers, all of our engineers, junior engineers, senior engineers, all of them have access to Cursor.
5:10They're all generating so much code that now we don't know what to trust. How do we promote code to production? And there's still very few people who can do that, who are able to tell what is good enough, who are able to sort of, who's going to judge the output from these agents? Yeah, this happened recently. There was a hackathon that we from these agents. Yeah. I mean, yeah, this, this happened recently. There was a hackathon that we were conducting. I think I made a video also on this, inviting people for India AI hackathon, which is for tier two and tier three cities, villages. You have doctors to patients in a ratio of one is to 11,000, which is 11, 11 X more than the WHO recommended ratio in the cities.
5:41It's fine. I think it's like eight 50 or something or a thousand or something. So can you use AI to help save doctors time? And I'll be honest the submissions all seem the same. I'm not blaming the teams I think they have used AI and it's helped them but it's also hurt them because their uniqueness in bringing this forward has been cut down. So the point that you are making about can we evaluate whether this is a good solution or not? The preference, the taste, that is something which is really important. If you create a project, can you guarantee that, yes, what I claim is actually what is happening? Can I measure that? And that's missing. It should be in our DNA as engineers that, hey, I have written tests.
6:20It's working out. I mean, that's the basic idea of an engineer, right? Making a system reliable. The bridge can't fall. If the bridge falls, then you're a scientist if you are okay with that. But it's still not reached on the AI side. Right, yeah. I mean, just taking that thought process forward, right? It's not that it hasn't reached. It's that I just feel like people are focusing on the wrong things, right? Even traditionally speaking, pre-AI, engineers were meant to think about design. They were meant to think about how to architect a system and so on. And then I felt like there was a lot of emphasis on the coding part and coding quickly and efficiently because that took up the bulk of their time.
6:55Now that you're able to do this really fast, just generate code and whatnot, I think the focus should shift more upstream. Focus should shift on the design and the decisioning elements. And to be able to do that, the context still lives in the engineer's head. It's about on the design and the decisioning elements and to be able to do that the context still lives in the engineers head. It's about people with tenure and taste. So are you saying this is another question which comes a lot. I have 10 years of experience, I have 20 years of experience, I have worked in various systems on various systems and various fields and now I feel like AI is doing things better than me. Is this true or do you
7:24think AI is doing things better which is. Is this true or do you think AI is doing things better which is limited to coding but the subject matter expertise for an engineer would be I think in software would be architecture would be you know that's where taste or foresight comes in. So that's do you think that's what's happening? I have seen this with product managers definitely they say that yeah I can help but it's also limited like my yeah my understanding is yeah exactly right that's that seems to be the most practical and logical conclusion but how are you going to tell or how are you going to develop a good sense for what ai could probably do well versus what it cannot and this ties back to the
8:00initial conversation we have it right that's why it's important to sort of upskill or learn and understand the foundations of this new technology that's coming together is changing by the moment a minute but it's important to understand how it's set up so you also get a sense for its capabilities it would be like someone who is writing code using mainframes and you know writing code for applications before 2000 and then 2000 comes in the internet comes in it's not like everything that you learned doesn't matter it does but it's just that the way you're thinking maybe or the capabilities of this new technology is going to change the way in which you think and work think and work exactly yep all of the stuff that i said is basically just taking whatever
8:39arvin's thoughts were and then putting it back again so okay now i am let's say i'm a person who's graduated 10 years ago which is more than 10 years ago now actually and then once i go through the ai literacy terms what would you suggest in terms of things i should do like should i go to my manager say that this is a project i want to work on or should i search for open source projects or should i create my side projects how do i gain the confidence through implementation that yeah this is this is this is it yeah there actually a very difficult question simply because this is the analogy that I keep falling back to right is that software engineering as a profession as a domain took about 30 odd years to harden so people who've been thinking about the
9:22software engineering and who've been strategizing for strong software engineering job profiles at least in the last decade pretty much knew the lay of the land the roads were already laid and then you know there were definite strategies you could go after because you also knew what the interview process looked like and so on but with AI it's almost like you know you're in the middle of a jungle and you're slowly trying to like you know sort of hack your way out yeah hack your way out and you're trying to like you know you're in the middle of a jungle and you're slowly trying to like you know sort of hack your way out yeah hack your way out and you're trying to like you know sort of organize
9:48that space and develop and so on so it comes down to two things at least that we believe in Aishwarya and I at the Gen Academy is that it comes down to not looking waiting for permission not waiting for permission in the sense just try and understand what AI can do and then you've already like you know developed so much intuition about whatever domain you're in go out there if you have a point of view have an opinion have a strong one and go out there and build it and the second one that we think is really important at least in today's world just because everybody has the same question in their head right whether you're an IC or a manager or even like you know an enterprise all of them are trying
10:20to answer the same question and the only way you can like you know get in front of them are trying to answer the same question. And the only way you can like, you know, get in front of them is by building in public. It creates a track record for your thoughts and your opinion. And that matters today more than ever. Yeah. I have a friend, Hemant Pandey. He was saying that he used to work in meta and he was saying that every software engineer is asked to think like an influencer also. Because whatever you do, you publish the work and you let everyone know about it because we are an influencer company and the engineers also influencers. And it's much more important than many of us give, you know, some people try to write it off as, oh, they just show off their
10:52work. But if they don't show off, who's going to show off their work? So yeah, makes sense. So thanks, Arvind. These were the most important questions. I think it's on all of our minds. And there's really the biggest question, which is how do I navigate in this new space? Thank you for coming folks. If you haven't seen Arvind's website, you should, it's new. And it's the Gen Academy, right? Arvind is very popular on Maven. He's taken courses on AI engineering on in AI engineering, there's evaluations, there's a system design also, MCP, lots of stuff. It's very nice to see you, Arvind. And I think your community has more than 25,000 engineers. It's nice to be in the same space. So thanks, Arvind, for coming.
11:32No, absolutely, Gaurav. Thanks for having me. Really enjoyed our conversation. And yeah, with learning especially, right? We were chatting about this. It's about you should try and learn from any source that makes sense for you if you are looking to upskill in AI. And ultimately, it comes down to a voice that resonates with you. For the longest time it was Gaurav for me especially during my software engineering days, his system design courses did help me a lot. So yeah just pick any resource that makes sense. Now it's Aishwarya or both the Aishwaryas probably. A lot of people probably don't know about this. My wife is also called Aishwarya and so is my co-founder. That's good stuff. Thank you for coming.
12:15This shake hand, half of it will be in this camera, Cheers guys.