Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Inside Cursor (Part 3)
4 Plays8 days ago
Episode 2
AI in the context of Platform Engineering, On-Call and SREs.
How RubixKube plugin with Cursor gives users an exponential boost in productivity.
The Root Cause - A podcast by RubixKube
Learn More
Recommended
Transcript
User Personas and Cursor Usage
00:00:00
Priyank Upadhyay
So Mohit, we were talking about various kind of user personas that have been using Cursor, not just developers, but people from other aspects of engineering too.
00:00:16
Priyank Upadhyay
How does it affect a support engineer or technical staff, like what you are doing, an SRE or a DevOps person, how how they are interacting with it?
00:00:33
Mohit
So this is just to say that, okay, so there's definitely some cursor usage on my end despite not being a coder. So how I personally have been using it or like a lot of people in our team and across different companies, different support systems can potentially use it in the future is that, for example, if I'm getting a query from like a particular enterprise customer and then I see that query, my we have a lot of rules, skills internally to like kind of put that in cursor. And then thanks to all these crazy MCPs out there and plugins out there, even like RubixKube helps me like make a story out of whatever the investigation kind of comes up with. to give like a walkthrough of how it will look like is that Priyank reached out to me saying that, hey, I have been experiencing increased lag today on my cursor. What's what's happening, guys?
00:01:23
Mohit
And then I see this query. I'm like, OK, interesting. Then I put this Priyank query to my cursor, and I'm like, we have all these amazing MCPs and plugins right like already in place. Rules are already in place. Now we start the investigation that, OK.
00:01:36
Mohit
What does, what's Priyank's role, right? Like is he part of a team subscription? What, where does he stand on that? Okay. Now let's go to, okay, what do we know about his enterprise network configurations, right? Then we check that.
00:01:47
Mohit
Okay. Then it goes to a layer of like, let me check all these RubixKube plugins. Let me check all these different MCPs we have connected to see, has there been some sort of a
Enhancing Support Efficiency with Cursor
00:01:54
Mohit
network outage? Has there been some sort of like reports of, people actually showing that it's a repetitive bit pattern or not right so that kind of gives me enough and more context to give you a very personalized answer right like which is really relevant to priyank and it's not really relevant to another person who's working in another company right because i can just easily get all these sort of like
00:02:18
Mohit
It's again at the end of the are getting number and text strings from these tools, but Cursor is the orchestrator who creates a story out of it. And for me that is incredibly helpful because that helps me tell you, hey Priyank, we noticed that you stay in this region and we noticed that sometimes users on this particular network can get like a slow response. Can you try just checking your mobile hotspot once to let us see if that is exactly the cause or not, right? like If I were not using cursor, this would take me like a lot of back and forth with you just to understand, hey Priyank, where are you staying? like what are What is your like status? And all of those things. right like So I think that process has definitely gone from, let's say, three back and forth between us to like the first response I give you is automatically going to address that. right
00:03:02
Mohit
So I think that's, I really still think it's underrated how much people are sitting on not using cursor for their support. right like I sometimes interact with all these like companies, I am having some sort of support problem with this. And they would ask me like the most basic thing, like, hey, could you please tell us like if this was the product you're talking about?
00:03:22
Mohit
And I'm like, I wish they had cursor, right? like they could go They would put my phone number at the back end, it would automatically fetch why when does this person ordered. And it would just give them, like the first response would be literally answering my query. But instead, now here we are you are asking me about like which order it is, like what so yeah i think support is something i still think like people are sitting on and not using cursor or like just incredible mcpn plugin again like rubik's cube there's like now databricks datadog stripe all of these like incredible mcpn plugins out there which just makes your support system so efficient
00:03:59
Priyank Upadhyay
Makes sense. I think cursor...
Innovations in Cursor: RubixKube Integration
00:04:02
Priyank Upadhyay
So from inside cursor you are seeing this. I come from an outside cursor perspective, right?
00:04:10
Priyank Upadhyay
um We launched RubixKube last year and one of the most called out features from our users was I don't want to switch from my IDE to something else.
00:04:20
Priyank Upadhyay
So can you bring in RubixKube within my workflow? So we released a plugin for cursor and I think it's one of the biggest change users are seeing and they are really liking it.
00:04:36
Priyank Upadhyay
you As you said, right you can investigate infra issues or any SRE related tasks directly from sitting on your cursor and maybe there there is a root cause that got generated and now you also have your editor so that you can apply those fixes directly.
00:04:51
Priyank Upadhyay
So makes it makes the life much easier.
00:04:54
Mohit
Absolutely. Absolutely. I think that's what, right. I guess how it what used to be a lot of just engineering waiting to hear back from engineering now kind of has come down. Right. And that's not to say that we, uh, there are a lot of time when we actually still reach out to engineers, but it's just that, okay.
00:05:03
Priyank Upadhyay
Yes.
00:05:09
Mohit
Maybe in in the past it was like 9 out of 10. Now it has come down to 3 out of 10. And even if we are going to the engineers, it's already with like very strong investigative data, very strong diagnostic data that, hey, we already figured this much out.
00:05:22
Mohit
We just need your help to figure like the missing part or deploy the fixed part. right like So I think that is what personally we have noticed or like I have noticed using cursor, add cursor for the support.
AI's Economic Impact and Educational Divide
00:05:37
Priyank Upadhyay
You seem very optimistic on the AI impact that is happening. Do you see any down downsides of AI in the industry?
00:05:53
Priyank Upadhyay
There are like economic ones that everybody is aware of, but other than that, do you see anything AI is impacting?
00:06:03
Mohit
I would say like this is from like this is because of AI which might happen but also I don't genuinely think it's just only the AI to blame right like I think there's a like like this is a multi-decade ongoing issue of like education system not able to keep up with the in trans like transformations the world is going through. And I think with AI, LLMs, all of these things in the past five years, even with my cousins and like even with people I know, I have been like, oh, just because you probably are not from the best college or not in the best place in life, you don't have access to this tools or like, knowledge and that's gonna keep you so away from like the top people who are now at the forefront of this right and for me like as much as i'm optimistic like that part still like makes me a bit like empathetic towards a lot of people that uh unfortunately like this is sort of leading to a divide of like
00:06:56
Mohit
okay who can get access to these models who can get access to these knowledge and there's like a huge subset of people who are probably not getting access to these knowledge and tools uh despite there being like so many good free tools also despite being there like enough information on twitter i'd still think like so much of at least in our country right like india is still dependent on college and college are still probably teaching like i don't know a very one-on-one which kind of puts these people at disadvantage from everyone else who's been like either self-learning has access to better resources. So I feel like that part I always like think about and do feel like a lot of empathy for these users.
00:07:34
Mohit
But yeah, I don't know if I'm like, I have any negative emotions towards again, the air AI itself, right? Like at some point, Again, it's being like cautiously optimistic and also kind of feeling like we have to just go with these things. right like we can't really ah We should not be hindering progress or technological innovation just because we think it's going to go in a different direction. I think it's important, like there are like enough people out there who are raising some very good concerns about the safety, about like, okay, what happens to the economic part of things, as you said. And I think those conversations are definitely important.
00:08:11
Mohit
But at the same time, I wouldn't be the one to say that, yeah, let's hinder the technological progression just because we don't know where this is going, right?
Balancing AI Progress with Safety Concerns
00:08:19
Mohit
Again, do as much scrutiny, get more government people on board, have more case studies, more research studies. But at the same time, like,
00:08:27
Mohit
we should not be hindering progress is what I think. I think life truly finds a way. Again, coming from like, oh, you will know a Harari book. Life always has found a way. I'm pretty sure we will. And historically, we have been just doing better and better. And I would like to believe we continue to do do that, right? Like maybe it will... um help me, I don't know, explore my love for chess again after the AGI is here and we are all like probably in a good state. So maybe that, right? like That's why I'm cautiously optimistic and I would put a lot more emphasis on like not the AI companies themselves and they do try to do their best at like safeguarding things and everything, but also a lot more on like the educational system, a lot more on like just government trying to actually work with these people and then trying to come up with like policies at a much faster rate. like
00:09:12
Mohit
Don't ask people to stop progressing. You should be progressing much faster maybe. So yeah.
00:09:19
Priyank Upadhyay
One of the points you mentioned was around safety and guardrails. Do you feel there is a limit? So we have been putting code in production for around a year now, vibe coding and putting it.
00:09:34
Priyank Upadhyay
um Recently, I was listening to Google launching an identity awareness for, I think it was called SynthID.
00:09:48
Priyank Upadhyay
if I'm not wrong. So any AI-generated images are being flagged and identified with Synth ID. And so you know this is an AI-generated versus a real image, right?
00:10:02
Priyank Upadhyay
Now, it totally has a lot of concerns and things that we need to take care of. is there a way to flag that fingerprint? It is an AI-generated code versus you know human has written it.
00:10:23
Mohit
ah i think like again i'm not the best person to speak here on like the technical nitty-gritties of how does like uh like i might not be the best person because i have not like personally been like a safety researcher in that regard but i would still like to believe that like That's where the whole conversation of like taste and like understanding comes into place. right like If I'm asking some like let's your very veteran professional engineer and asking them to make me a calculator app, they'll probably do it in like a good 400 lines, something along those lines. like I'm not sure. right like But it will be a very neat and optimized version of a calculator app right and without anything extra. like There will be no ah no confetti if my answer is 2 plus 2 equal to 4, no confetti.
00:11:09
Mohit
But it's possible that when I ask LLMs to do it, or like one of these AI coding tools to use it, that there are like some additional things added, which I probably did not ask for, and now the code is like 2000 lines, right?
00:11:22
Mohit
So I think there are like some human ways of like definitely seeing some websites and calling it like, okay, this is probably a a gen website, right? At the same time, I don't know from like a technical point of view, what will be the best case scenario.
00:11:35
Mohit
Also, at some point I really do feel like the conversation goes to like, do we really need to know if it was AI generator or not? right like We are already seeing companies who are like clocking more than 80% of their code committed on main just using AI-assisted coding. So where does this line kind of end? like Why does the distinction matter? right like What I would probably emphasize more on like is and just making sure there are enough safety and guard drills internally from a company point of view even like a lot of humans in the loop systems even if like there are like auto-generated code and everything going out so that there is some sort of like hygiene, cleanliness, safety all maintained at quality so that at the end of the day even if anything goes wrong the assumption is that okay of course it was AI-generated code but okay who was the one who managed this right who built the uh
00:12:26
Mohit
automation which led to this right like and then we can go from there to like kind of diagnose the root cause right but i still feel like i don't know how much important it is to actually see ai code versus non-ai code like
00:12:37
Priyank Upadhyay
But see, the the point I was making is you see a lot of news coming up. AI deleted my database or something or the other.
00:12:50
Priyank Upadhyay
We still have to root cause it, which, let's say, just comparing the AI tools itself, which AI tool I should not be using. The harnessing is not proper. These kind of things should be diagnosed and at least audited well.
00:13:05
Priyank Upadhyay
Don't you think that is important?
00:13:07
Mohit
Yeah, again, absolutely. like I think we still have to be like as careful. like i think a lot of I think the point here which you're mentioning does sound really well to me in the sense that ah just because the shipping of code has become so easy does not mean you should also ship that carelessly.
00:13:25
Mohit
It's not a guarantee of safety, it's a guarantee of like implementation. Now, is this implementation best or not? That's for you to decide. I think like we are definitely not super close to that future right away that yeah let's not even like review this thing which AI generator or let's not even review our practices let's not design our better workflows I think the good companies which we are seeing or like the good teams who are working with AI do this very meticulously right like of course like we have got this very powerful tool in our hands but again the
00:13:33
Priyank Upadhyay
Yeah.
00:13:59
Mohit
like you know like if you give a hammer like if the only thing you give a person is a hammer then everything looks like a nail so there is like there's enough importance to be given to like okay instruct it guide it take care of it like understand what it's giving you like do enough test trials right like staging and like all those things are still like necessarily precautions we should take and Even when we're dealing with users, we see a lot of times they will be like, hey, the agent did this, right? But then we again have to re-emphasize, it's up to you, right? Like, are you sure you wanted to run the agent like completely auto-run on each, like permissionlessly, right? like
00:14:35
Mohit
There's a safeguard in place and cursor also, right? Like you can allow less certain permissions, you can ask it not to do certain things and it does a pretty fairly good job. But at the same
AI Marketing Myths and Realities
00:14:44
Mohit
time, it's not to say that, okay, now it's just magical, right?
00:14:47
Mohit
Like there will never be a problem with your company ever again.
00:14:50
Priyank Upadhyay
that's That's, I think, a problem of marketing. That's how a lot of industry is marketing their products, being magical and doing stuff which which will be fully autonomous.
00:15:06
Mohit
Yeah, i I mean, to be very honest, I think that's a wider problem of just like attention span and what it takes to like create distribution than customers in this market that if I now tell you that, Hey, my tool has like 93.1% accuracy.
00:15:21
Mohit
The person is like, Oh, okay. You missed out on 6.9%. That's like not nice. Everyone just wants like, so I do think like it's definitely overblown and like, there are like a lot of, unfortunately, uh,
00:15:36
Mohit
misleading claims which comes in the picture just because you want some initial traction but again I think in free markets everything is a matter of like both user demand and supply so right if the user is not gonna listen to you or like work with you when you say 80% uptime and they want 100% then kind of it's like ah you know what let me just put like a 99.1% uptime and put a star there just for the marketing sake so I think it's just like a trade-off of the free markets we live in
00:16:05
Priyank Upadhyay
sense all Alright, um we'll move to our last segment Mohith and this is where I want to touch up on some things which are very relatable to people, anybody um like jobs, roles, what a human being should be.
00:16:31
Priyank Upadhyay
What's your thought process on that? Let's start with that.
00:16:34
Mohit
What a human being should be. Man, I really thought I was coming for a tech podcast. of What a human being should be.
00:16:42
Priyank Upadhyay
You are a non-tech person, right?
00:16:45
Mohit
I don't think we should be calling anyone non-tech at this point. like who if like
00:16:49
Priyank Upadhyay
Yeah, I think let's let's segue to that then. You brought up a very good point. Does a non-technical even a category anymore?
00:17:00
Mohit
It shouldn't be like i I have been like fairly advocating against it. Like, I mean, I do always call myself non-technical, but with the caveat from a point of degree and like background experience, right?
00:17:10
Priyank Upadhyay
Yeah.
00:17:11
Mohit
Like that is always a caveat I have put in. oh But am I non-technical? Because again, when I was in 9th, I was jailbreaking my phone. Oh shit, I probably should not admit it on call.
00:17:22
Mohit
But like I was jailbreaking my phone and then I was like going to all these sketchy websites, trying to download stuff. and like So was I non-technical? no right like that is still a technical part of me even now like now with the ai tools of course i have become like 10x technical right like i have started merging prs not in the main but like for our internal dashboards and everything so i've started like merging prs now and like raising all these things so I don't think I'm non-technical and this also ties into the whole job thing which you were mentioning about that at this point I think you should not pride yourself if you are calling yourself non-technical yeah yeah I like non-technical I'm creative as a person no you have to be technical
00:18:05
Priyank Upadhyay
That's basically limiting yourself because anyways, you are not non-technical. If you're consuming Instagram, YouTube every day, you are in that trend.
00:18:16
Priyank Upadhyay
You should try to think more
00:18:16
Mohit
yeah
The Evolving Demand for Technical Skills
00:18:19
Priyank Upadhyay
algorithmically, logically in your thought process.
00:18:21
Priyank Upadhyay
And then code is fungible now. You can generate whatever you want.
00:18:27
Mohit
Absolutely. I think like I agree to your point. like We stopped becoming non-technical the day when we got this laptop and mobile in our hand. right like In that sense, like it goes way back. But now with the advent of all of these extremely incredible tools, I think the leap like the gap has just been like cemented super close now. right like if i want like we we are also like seeing on twitter like people are like non-technical conventional people are now pushing products making good mrr and like good ars and like they are maintaining code right like so uh yeah and tying this back into the job market thing like yeah i think be technical right like you have to be technical that's kind of become like one of our like you can be as idealistic as
00:19:09
Priyank Upadhyay
Default skills basically like you know writing, reading, walking. This should be something like an evident skill that you must possess if you want to grow in any field or role.
00:19:22
Mohit
Absolutely, absolutely. Right. Like, uh, and this is like, if I just broadly think across different departments, right? Like for example, marketing, I can imagine like maybe four years back, you could get hired in marketing just because you were very good at like the copywriting.
00:19:35
Mohit
Right. I'm pretty sure there are jobs out there, which is probably still hiring for that, but. are they hiring as much as like 2000 people to now like probably 20 people now right so what i'm coming at is like okay if you are a marketer and you think you are non-technical no you still need to understand like okay how can i ship like a good growth campaign make a website for it how can i use ai generator like videos and assets to like make my everyday tweeting better or like posts better how can i work with like a email newsletter which we are giving to better identify from the data that this one got more clicks and put it in imcp put in something and like try to get a story like what really worked for you versus did not right like so i think calling yourself non-technical at this point of time would be just like doing yourself a bit of injustice and also just like is not what market really wants right like so gotta just like keep up
00:20:27
Priyank Upadhyay
And a lot of things like even in sales, you you could possibly automate your whole workflow with you know AI tools like Cursor.
00:20:38
Priyank Upadhyay
You can do a lot of prospecting, lead generation, a lot of things that you would have usually done manually can be automated by self.
00:20:47
Mohit
Absolutely, yeah.
00:20:48
Priyank Upadhyay
So in 2027, what does a five-person company look like? I think it's possible. So... Deal?
00:20:57
Mohit
Oh yeah, I think it's definitely possible. i think like the first I would also like just be like, okay, what company are we talking about? we talking about a company which is like good enough to sustain this five people and like making good enough and like, you know, like what is like a company, right? Because right now if I incorporate LLC and then that's a company too, and I might not even look at it for a year, right? Like, so I think one question is that, local what kind of company are we talking about?
00:21:22
Mohit
I personally feel like, yes, you can be a five-person team, build a product, get like good revenue on it, and just keep yourself like working on it. And you probably wouldn't need to expand the team and a lot of those things. And you'll probably be okay to do.
00:21:36
Mohit
But at the same time, like if you want to be like the i don't know the next unicorn and like you want to achieve, like you know give to enterprises, you want to be like the...
00:21:47
Mohit
a product leader of a particular category i personally don't think a five person a extremely good five person team might be able to do it what i'm coming at is like just with ai i don't think it fixes that problem right you still need enough smart people in the room the the world is too complicated to just have like
00:21:59
Priyank Upadhyay
yeah
00:22:05
Mohit
a problem statement written in a prompt, right?
Scaling Teams with AI: Opportunities and Limits
00:22:08
Mohit
Like there can't be a possible prompt which is so detailed and has enough world understanding, enough your company understanding, context understanding that this prompt gives you an answer which accounts for every single detail, right? Like the world is still incredibly complex. So, right.
00:22:25
Mohit
uh we still need a lot more people like and the whole fact that we see the frontier labs we see cursor we see all of these ai tools have a careers page which has active listings says something right like these companies are also like hey we still need you it's not that we don't need you it's again like a different requirement in a different perspective we need a different kind of like outlook to the problem but it's not to say that now we don't need a person right like uh maybe a very distant point in the future that could be a possibility but i don't think that's like i wouldn't have gotten hired like three months back if like they did not need a human person right like similarly i think uh that's what i think like it's possible but not if you want to like grow extremely big and you want to like maintain everything as healthy
00:23:15
Priyank Upadhyay
Makes sense.
00:23:17
Mohit
there anything which can backfire here?
00:23:18
Priyank Upadhyay
Basically, like these are the roles that are staying for longer. um Engineering, product management, marketing, these are going to buy like reduce in terms of human count, but more quality output.
00:23:24
Mohit
Yeah.
00:23:37
Mohit
Yeah, again, i even like with the ones which I said the jobs will be there, I feel like, again, AI assisted, right? like I think AI assisted is the key tagline for every single job title.
00:23:45
Priyank Upadhyay
But still, let's be fair about it. I think a person who was doing like 10 people team was doing something. Now, single person can do the 10 people job.
00:24:00
Mohit
Yeah, I think I agree with that. like As I said, like I'm not saying that oh AI has not impacted anything in our industry, right? Like that would be super like ah not reading the room. It would be super unfair, right? Like all of the, all I'm saying is like, it is just changing the kind of qualities expected from people. it is changing the kind of hiring, which has been taking, it has been changing a lot of like expectations between a hiring manager and the employee now.
Career Adaptability in the AI Era
00:24:29
Mohit
But at at the same time, like, a it goes back to the book I was mentioning that we do we are living in a time where I will probably have to undergo a career pivot multiple times in my lifetime and that is now something which I have to make peace with and is it gonna be by relearning a new set of skill is it gonna be like making something of my own is it gonna be writing a book and learning how to paint could be right but I think a lot of mental shift has to take place in just being okay with that this is not the time where I can work in the same title for 20 years and just grow the career ladder probably not I'll have to readapt multiple times during my life and I should just make peace with it just so that it helps me sleep better at night and be okay with something which might come tomorrow which I don't know about
00:25:06
Priyank Upadhyay
Makes sense.
00:25:11
Priyank Upadhyay
Makes sense. if ah We do have a lot of young audience just coming out of college. For a 20-year-old, if they come and ask you whether they should be learning to code, what would you tell them?
00:25:26
Mohit
hmm
00:25:28
Mohit
Yeah, I think I would still be ah asking them to learn how to code. I think it's like a lot of learning is not just like based on output. It's also like just fundamental understanding of how something works, right? Like so any ah learning underwear I would personally think is not like a waste, right? Like I could ask them like, hey, if you are in like an urgent need of money and you need to get a job in one month, then probably I would say, okay, try not to learn coding, right? Like it might not like get you to a level where you can get a job in a month. Then my advice would differ. But in general, like just because we have AI toolings, we have AI coding, all of those things does not mean we should stay away from learning and going back to the basics and learning the fundamentals, right?
00:26:08
Mohit
Like it always like can be possible that hopefully, like if you are able to like go back to the basics, then you also have the opportunity to rebuild or rework something from the basics. right like You can be a 10x engineer further down the line when you know that ah AI told me like to do 10 things, but I know like this can be done in two things. right like so i think no learning is bad in that sense.
00:26:32
Mohit
At the same time, it just depends on like what is the objective you are looking for. a And just to take a bit of time here, like if the objective is to just get a job and you don't have a particular loyalty to a particular department, title, all of that, then I would ask you probably that, okay, if you're just that sort of a driven person, then probably don't spend time learning how to code. Try picking up skills on the different AI coding tools. Try building out like side projects. Try...
00:26:58
Mohit
you're not getting ah attention from these people on different platforms. And then once you find you have enough knowledge and skill, make like the best pitch for a company, do work for them without getting paid.
00:27:09
Mohit
And then so that they wonder that, oh, if I give this person some money, how much will they do if they're doing it without getting paid? Right. like So I think that is the advice. Like I still am like very, I personally read a bit still.
00:27:22
Mohit
i personally try to like understand the basics again. Maybe it's through GPT that I understand. Okay. How does a plane work until a long
00:27:29
Priyank Upadhyay
Yeah, the mode has changed.
00:27:31
Mohit
the mode has changed but I think the learning still has to continue it's not to replace as you said right like it's not to replace your brain itself it's just supposed to be a good uh like a power up in a game