Chinmay Bharti, Cod: Building with 5 AI agents instead of software engineers

welcome to another founder wisdom pod YC Edition we have chimme biti with us he's co-founder at code an AI he part of y combinators winter 24 uh Edition he's uh been a Quant developer it's not his first uh startup as well he's been a founded resident so interesting guy all together going to have fun uh talking of about a bunch of topics from AI to startups he's in SF I in Waka Mexico this podcast is brought to you by pod.com if you want to start your scale of podcast go to pod.com chin may welcome to pod tell me a bit more about yourself and what you're up to nowadays yeah hi man nice to meet you Charles uh nice to meet you guys so uh I'm chin I'm the CD and co-founder of cod uh Cod helps you write best quality code in production so I'll tell you why this idea came from in my previous ttin I observed one thing that bad code is being shipped every day code that doesn't have documentation doesn't have unit tests uh not readable not maintainable and this had an extreme impact on my company uh because everything slowed down so because of this this this was the problem that I I was passionate to solve and I started Cod anti to solve this specific problem how did you get accepted into YC you might you must be exceptional yeah yeah so actually we got rejected at first so in the first why combinator call uh we got rejected because we were not able to convey the idea properly but uh we had this thing that hey like if uh so we got rejected to all of the VCS from India before that and it was kind of end of the company so we were like hey if it don't get accepted here then uh it's not going to fly so we wrote a mail again that hey can you give me one chance and this never happens like they don't give another chance but uh Tom actually so Tom is the group partner uh he actually gave gave us another chance and uh he scheduled an interview with Michael Cel and uh like we explained the idea properly to Michael and that's why we get we got accepted and here we are right now so what did Michael see in you and your co-founder right you're two or three yeah two okay two what did he see in both of you and what kind of feedback did he give to Yeah man so uh it's just a 10 minutes uh meeting so uh first of all he was impressed by the never give up attitude so like if Founders are this much committed that they can get done so uh that was the first thing that impressed him and uh second is we were able to put the idea in words very properly because we practiced a lot for that we messed up once so we exactly knew what to say and Michael was like Hey this actually makes sense and uh you guys are very well suited to solve this problem so yeah like we'll give you a chance yeah that's what happened right then tell me about your background because you're a smart guy you've been a Quant developer I think it's the first time I see that title on LinkedIn yeah yeah so uh I graduated from IIT Bombay so it's uh it's one of the top engineering institutes in India uh then I uh did a Quant internship Quant developer internship at Blue analytics so that's a German trading firm where I worked there so it was a it was just a bunch of mathematicians tried to hack the market and they wanted a developer to develop all the algorithms for them so I worked there for six months buil their whole infrastructure from ground up and uh I always wanted to do startups man like uh in in my college Years also I tried to start a lot of stuff but nothing uh really took off so I decided to join uh another startup as a founding engineer so I joined zevi uh so we were building search search engines for d2c Brands so consider Amazon like search interface for your Shopify store or something like that so that was the idea and we were building that and I saw the V 0er to1 Journey there like I joined the team when they didn't have uh any Revenue any funding so build the whole product uh uh and building a surge product is like especially difficult uh because you have to manage a lot of scale and stuff so yeah man it's been a rough ride but after college I was I had a placement at I got placed at Jaguar Land Rover uh into their autonomous car division but I was like okay I mean that's that that was not something that I wanted to do I like I resigned before even joining the job and I like let's see what happens and uh I just started working on COD and a and fortunately I got into y compan right what was the challenges of coding for mathematicians oh uh man th whenever you are trading the markets the number one most important thing is uh latency you have to be your code should be extremely optimized to a millisecond like that is the kind of latency that they expect so like writing a lot of lowlevel code and uh optimizing the analysis so I mean it's it's more like writing C++ very low level C++ and working with fpgas and stuff so it's that's where I actually learned the importance of good code and this this importance of good code is not really known to a lot of people outside the trading world so that's why I was like yeah man this is something that can be useful for everyone and we can actually create a lot of value with this so that's where the good code comes from do you see a mathematical worlds when you look at the laws of physics and what's surrounding you do you see a world that's coded oh God it's kind of a matric stuff but uh ah I mean I don't have a very very strong thoughts about it it's like this question is very similar to do we are we living in a simulation so maybe yeah maybe not but uh I mean so one thing that I'm seeing uh by living in this industry is because of AI everything is transforming so fast and I'll tell you one thing man uh incumbents are very scared because now so I I actually have an AI agent setup so I have five developers uh five AI developers coding for me and man like you can achieve such a scale with these agents and incumbents are really scared because incumbents have built the system by hiring like 200 engineers and now we can disrupt them with like we have infinite Engineers on our hand so this is this actually presents a very nice opportunity for disruption of all the incumbants that are there and a lot of IC companies are trying that so someone is building uh building an alternative to Salesforce like AI enabled alternative to Salesforce we are building an AI enabled alternative for sonar Q so is the industry leader in code code quality so a lot of Transformations are happening so this actually we can actually change the Dynamics of the whole world you can shift the powers in the tech industry by just uh having AI work for us man dude I love that um is it custom built why don't you offer that as a service yeah so uh actually uh we have taken it from so there's this open source GitHub repository GPD engineer so we build some custom layers on top of that and uh people people are also developing so if if you know Devon came out so Devon is your AI software engineer within a week people have open we people have built open source de uh people are building open source perplexity man like perplexity is like I don't know $200 million funded and they're building the open source alternative for that so and it's and the and the product is itself like open source perplexities itself is very good so it's not like it's hard to build it's just uh building is so easy right now that uh everyone is uh I mean this is this is similar to Industrial Revolution that happened the AI Revolution and within the next five years a lot of things are going to get disrupted and how are they automated are they testing your platforms as potential customers reporting bug to fixing or executing on new features okay uh okay so I uh we personally use use it to develop new features and once the features are developed uh we also pass it through a testing pipeline so end to end development and testing obviously there is uh so I'll tell you the vision of code and in few years you will have tens and hundreds of AI agents writing code for you so there has to be a super agent that oversees whatever output is being generated so Cod and AI will evolve into that super agent so what I'm doing right now is I'm looking what is what are all the agents producing but my job will be replaced by cant Ai and cant will actually review all the code that is being generated refactor The Code test the code so that's how I'm seeing code anti to evolve so that's the long-term Vision I love that then what will you do as the CEO uh I mean uh in what context your company's context because you said it would replace you yeah man so yeah obviously like there is always uh so there is always room for iteration and optimization so once you develop a feature you get new ideas that hey we can implement this also we can Implement that also and uh through my journey I've understood one thing that to survive uh in the market like customers are so difficult to predict like it's so random that you will never know you never know what is uh what is the thing that they really want so you have to build multiple verticals at the same time and see how they react to it and a lot of startups don't do that and that's why they die they just spend building one vertical completely for a year and after that they get to understand that hey this is not something people want so that is the Trap that uh I guess a lot of startups are following into totally agree with it but I wouldn't say that customers are not predictable I think they're very predictable I think that the market itself is quite hard to predict but customers it's quite easy to predict what they want they don't know how to explain in CLE in clear Words which is why you need data points inflections and AIS to sort of make sense of all that but there's some whatat predictable I want to ask you though um AI do you think it can be creative and if so when exactly that would be the new frontier H yeah so did you watch that hum AI demo YP bro you should definitely watch that demo. hum. so it actually has emotions built into it so as you talk to the AI it actually tells you what is your emotion and it reacts uh in contrast to that uh emotion so if you see the demo what they have done is uh so they have actually converted the emotions into a numerical format or or a categorical format and they have built a model I mean this is what I'm assuming they have done they have built a model that will actually predict the category of the text that you're speaking and it also has the intensity uh it also judges the intensity and stuff but uh creativity is something that can definitely be uh taught to AI so so uh for for example currently whenever you are making any call any LM call there is this this parameter temperature so if you set the temperature to high let's say you set the temper temperature to one then AI gets more creative and if you set that temperature to zero then AI gives more deterministic results so creativity like people have already started fine tuning on the creativity uh creativity parameter temperature parameter so that is definitely possible man right I find that the AI creativity is quite poor my definition of creativity as a Founder obsessed with product Market fit is an idea that makes sense right it's a great idea I have a very strong feeling that it's going to have a product Market fit um AI I don't often get impressed by that whether it's the data points in my brain and so forth and from listening to your definition it seems that the stronger you go on the creative sides the riskier you also go so and and then I ask myself is the AI for at creation because it can't really predict the future or is it because it lack it's lacking a bunch of data points that I have like what do I have that it doesn't okay [Music] ah so I think I think uh so I think this can be I mean I'm not sure but this can be soled to some extent via the uh assistants model so I'll tell you an interesting use case so one of the one of my batchmates in YC so they took all the lectures uh of s M man and all the startup lectures of Sam mman and feeded it like fine tuned the AI on top of it and now people are talking to that Ai and asking for advice and it is actually it's it's very similar to Sam Alman giving you advice on your startup that's super powerful man like how like let's say let's say we make an AI of you like you are the expert of lead generation and stuff so we can actually just get all of your podcast all of your YouTube content and make an AI of you and then people can ask you questions that hey is my email correct is my uh this correct that correct man that's super powerful you can make AI clones that can actually do uh to some extent what you can do but obviously like it's impossible to completely replace the human right now right no that makes a lot of sense um say that we take scoutman Paul Graham and all the Y seat uh people pair it with a bunch of Silicon Valley top gurus and we as them for product Market fit idea I'm pretty sure that it would be kickass idea especially if it's real time imagine if you can have like real time advice that they give yesterday on a podcast and then you can Branch it out you can go and in the branch of these ideas and and suggest similar stuff because AIS are good at that too um do you see GPT being deployed this year what are your AI predictions uh so I I've I went to a mixer yesterday and uh uh so it was a core generation Founders Meetup so uh there uh uh I saw the I mean there was an an engineer from opena was also there so we asked him the same question when is it coming out so he said that there might not be G GPT 5 so there might be GPT 4.5 so uh and it's coming out very soon uh I'm not sure whether he was lying or not but he had a poker face so but I believe that so Claud Claud released Opus and uh a lot of uh lot of people are already replacing gp4 with Opus in production and uh this is is all of the founders that I'm talking to 70% of the founders have done that so if they don't release something right now it's difficult for them so they will definitely release something in the next three months yeah I went from gp4 to Gemini ala and then yeah Opus is it's so exciting man yeah I felt like we had like a lack of excitement for a couple of months but yeah it's it's really impressive the the kind of startup war that we're witnessing do you think uh this war to and this arms race will finish in a good way or do you think we're gonna finish in Armon man that's why I like anthropic because their Moto is building safe AGI so the so if you look at the job market right now there are very very Less jobs in the market cuz I I'll tell you one thing I was this close to hiring a front-end engineer and a front-end designer then I got to know this tool subframe that offers everything in $50 per month like you can actually design components they have beautiful templates you can design components and then you directly get the react JS code so I was actually going to have two Engineers but I ended up getting the tool for $50 per month man it's insane Roi like why would I hire people now so I mean because of this advancement it's getting scary I mean I'm not sure uh like how will people end up like where will people end up if they don't have jobs and uh everyone is opting for CS but I mean it's difficult man like CS there are very few kind of jobs that will still be relevant after like let's say five to 10 years and it's scary dude like I'm not sure how this will play out like time will tell but uh already the job market is so I'm scared scared of that yeah the answer would be like if they're highly creative you might want to hire them but then if they're highly creative and gritty why don't they start their own startups you know so I think that at least 70 to 80% of the market will be self-employed in a decade or two from now yeah yeah yeah man that definitely makes sense Charles uh I mean Indie hackers Indie like Creator economy is something that is like developing rapidly and uh yeah like these are the only two two domains that remain uh so apart from software I guess Creator economy is something that will Thrive and uh people building Hardware will also Thrive because you need Hardware to run Ai and everyone is is like running air but there is not enough hardware for that so I don't know man how it will play out but yeah Creator economy definitely has potential and then what will become of Corporation will they let go of employees by the the 10,000s uh in terms of the future if everyone uh is using AIS now what will become of them yeah corporations are doing to fail uh I believe because I'll tell you why uh people are already building what so let's say one Corporation took 10 years to build everything from scratch like they have been developing something from 2009 and now people are building the same thing within like a year or two and uh there are lot of Open Source uh components available so it's very easy to get started so I think ultimately uh at least in software there will be uh a very flat uh hierarchy uh so it will be like everyone I mean it will be like there will be no there will there won't be any uh big players in software because everything will be commoditized you see Google man like Google was developing search uh since uh like last 20 years but now how many Google searches do you do you just go on yeah exactly just go on chat there's this website stack Overflow that is used for debugging uh you should see the graph so the searches have like exponentially dropped because everyone now ask CH GP stack such a big company man like I don't know crazy how many Industries chat GPT disrupted and like Alman is walking like almost oblivious to what he just I'm prettyy sure he's not oblivious but yeah he sh shifted time in space and the the future of humanity by aggregating top talent in in one little room and making miracles happen um yeah thanks so much for today chin mate uh we're already out of time we'll do round two you're an interesting Guy where can people find out more about you yeah so you can connect me on LinkedIn and Twitter sorry X you can connect me uh you can just search inti and or you can also visit cod.

to know more about what we are doing and get connected with us thank you