is AI replacing developers now data suggests otherwise all right so two things are probably going to happen in this article one this article is written by GitHub which makes me just assume that this is going to be an advertisement for co-pilot but I just want you to think about something which is why if you're wondering why I have not been freaked out about the AI Revolution is there's never been a time in history in which humans can do more and afterwards there's just less availability for jobs when it comes to programming other things yes but when it comes to programming it seems like every time we increase our velocity you just increase the amount of bugs and nonsense and difficulty and all of a sudden this number keeps going up and we just keep on going up and up and up and up and up and so I'm not worried at all and I'm not shocked by this title at all but I also think that GitHub really would like to sell more co-pilots let's see what it has to say uh remember when people feared AI would create fewer developers over time do you think they wrote this article before or after 03 was announced no the reality as revealed by the 2024's Octo octoverse report is quite the opposite AI isn't here to take your job it's here to amplify your impact this newsletter will explore the results of the report showing how AI is shaping the future of software development whether you're a seasoned coder or just starting out there's one clear takeaway here AI is making developers more efficient creative and indispensable than ever before let's go oh look at this the report starts off with the same the pretty much the same thing right here here all right so let's I'm going to look at the report a little bit more cuz I think the report probably is going to be a bit more interesting though the article might be a little bit more to the point let's hop back and forth between the two in 2024 python overtook JavaScript as the most popular language on GitHub while Jupiter notebooks skyrocketed to be fair so I don't know if you saw this graphic I believe they had typescript and JavaScript split I think they probably just deserve to be under the same family honestly like if you're going to measure most popular language you should probably just put them under the same category uh and if that's the case then I believe typescript and python are or typescript and JavaScript is more popular than than python it's kind of like it's kind of like splitting python on like python 2 versus Python 3 it it'd be a little bit weird to do that uh both of which underscore the uh surge in data science and machine learning on GitHub we've also we also see increased interest in AI agents and smaller models that require less computational power reflecting a shift across the industry as more people focus on new use cases for AI word calculator is fun to play with our data shows a lot more people are joining the global developer community in the past year more developers joined GitHub engaged with open source and public projects in some cases empowered by AI empowered by AI advertisement and since tools like GitHub co-pilot started going mainstream in early 2023 the number of developers on GitHub has rapidly grown with significant gains in the global South wow we see signals that AI is driving interest in software development we can't fully explain the surge in global growth our data reflects I think that's probably true like imagine you're just starting out programming and with a few calls you could make a product that does so much with like AI even though it may not be accurate to what it does it just does a lot of things right and I think that that is probably very exciting because I remember when I was a kid the thing that got me exciting was going on and being able to make like small changes and seeing it happen in a game right a game that allowed programming was so exciting to me and I loved it so I would imagine the same thing's probably true with AI where not only when you enter you don't have to figure it out all by yourself you have something that can spit out all the tutorials you've ever wanted but B it actually just does funny things right like you can just like show a picture of a chessboard and be like what's the next move and maybe it makes it up 40% of the time but nonetheless it still can like do something which just feels wild like that's just a crazy experience to have right you still need to know what you're doing to some oh no I would argue that you need to know what you're doing to a full extent but toy applications you can get by that's why you know like when they revealed 03 this big thinking model what did they do did they show you a massive application where is able to fix something in a giant code repository and able to actually be a very productive citizen or B did it actually just simply produce yet another tutorial project It produced a yet another tutorial project because it knows it can Ace that they know that if you want the AI to do the simple stuff it's going to do it well they literally said start me a project that makes a call out to an API and then writes the output to a file that's not a crazy project right I'm not over here being like holy cow a wizard has entered the building right hold on everybody Wizards are entering the building now right that's like 10 minutes of work it's like not crazy worth of work right so the more you know I think the better honestly the better you'll be okay at GitHub we know the critical role open source plays in bridging early experimentation and widespread adoption in this year's octoverse report we'll explore how AI is and a rapidly growing Global developer Community are coming together with compounding results all right 5 point let's see five uh 518 million total projects on GitHub with 25% year-over-year growth I I don't think that's surprising in the sense that every like git has obviously reached like a critical mass adoption where almost everybody uses git right like it's probably the most ubiquitous piece of software that anybody uses CU I don't use vs code some of you don't use Vim other people use you know Zed some people use intellig but almost every developer has some sort of interaction SL intersection with GitHub at some time and so it's not surprising that you still see huge amount of project growth on GitHub because if any of us start a project plus all the new people coming in it's just like it's just such a huge number and by the way starting a project this is kind of a silly measurement because I mean how many projects do you abandon the year quite a few contributions to all projects on GitHub 5. 2 billion that's crazy open source m maintainers verified students and teachers have used GitHub co-pilot at no cost greater than 1 million I I get it for free I haven't used it in like 6 months but I still get it for free contributions to public and open source projects in 2024 1 billion Pro let's see publicly generated AI projects with 98% year-over-year growth waight public generative AI projects with 98% year-over-year growth also I mean not shocking python overtakes JavaScript as number one language cool okay cool not a shocker either yeah it shouldn't be a shocker just because python has access to all the AI stuff and AI is so hot right now ai is the handle of programming all right surgeon Global generative AI activity AI is growing and evolving fast and developers globally are going far beyond code generation with uh with today's tools and models while the United States leads the in contributions to generative AI projects on GitHub we see more absolute activity outside the US the let's see in 2024 there was a 59% surge in the number of contributions to generative AI project on get up a 98% increase in the number of projects overall and many of those contributions came from places like India Germany Japan and Singapore I think one thing that's really interesting about AI is that any of the like The Cutting Edge models uh or any of The Cutting Edge programs like Devon you've effectively priced out a huge amount of locations $500 a month in some locations around the earth that's near a a salary for an engineer so it's like you've already priced out that group and for Devon to become useful for someone to be able to use in a lot of part of the world they can't have 62 and a half hours a month for $500 right like for me to want to use Devon in a realistic way I would not only need to be able to allow like my company would not only allow me to use it which they're probably not going to allow me to use it but if they did allow me to use it I'd have to be able to use it for like what 800 to, 1600 hours a month for it to be useful because I need to be able to have like four devons running when I go to bed I would expect all four of them to be able to be doing something right like can I actually make it run and actually be useful so it's like you have a huge problem in my country 500 is a good salary exactly it's pricing people out the new chat jippy stuff the new 01 access and the Sora access is $200 a month that's a lot like that's a lot of money in a lot of places and so it actually causes real price outs so is it surprising that the US leads in the projects not at all and that's just because of the economics the community ities are growing but I think there's there's definitely like a there's an income uh distribution problem when it comes to that $200 is what I pay for rent well you can also pay for rent on the internet if you'd like to I've already seen this stuff I've already seen all these things don't worry I I understand that 01 can't count fingers remember it's it's an approximation model right like I mean this is what generative AI does is its function approximation and so if it mostly if it's like pretty much a hand it's a hand how many pictures are drawn where there's four fingers on a hand that's like like most cartoons so why wouldn't six fingers also be a hand it's not a big deal all right rapidly growing number of developers worldwide especially in Africa Latin America and Asia notable growth is occurring in India which is expected to have the world's largest developer population on GitHub in 2028 wow that's a cool number also a crazy number well hey hats off to you India as well as across uh Africa and Latin America we also see Brazil's developer Community growing fasted I just see a Brazil mentioned can we get a Brazil mentioned uh some of the attributable let's see some of this is attributable to students the GitHub education program for instance has more than 7 million verified participants by the way I do think that the GitHub education when it comes to co-pilot access I think is really dangerous by the way Brazil mentioned people love Brazil we've also seen 100% year year-over-year growth among students teachers and open source maintainers adopting GitHub co-pilot as part of part of our complim complimentary access program I really think that if you're a student please I'm begging you I know it doesn't sound I I'm sure there's AI maximalist that will say I'm incorrect on these kind of things I beg you do not use co-pilot I beg you use chat GPT and ask it questions about what you're doing wrong and go look up the answers and go read always ask for links and go read from the source itself learn to learn learn to fish for yourself and then if you just choose that you're not a big fan of fishing turn on co-pilot okay some people they just don't want to fish and hey you know what you don't need to produce artisinal code you don't you do not have to produce artisanal handcrafted freerange grass Fred vegan friendly keto meat diet only code you just don't have to but you should know how to do that and if you don't know how to do that you need to take some time to go learn how to do that if you don't I'm telling you it's just going to cause all these slowdowns and all these problems there's like this classic phrase that a lot of people don't know or they do know they just don't apply to the most obvious things in life which is that shortcuts make long delays I believe Mary from Lord of the Rings Mary or pipp and I always forget which one I want to say it's Mary say this and of course they take a shortcut cause a long delay you don't need to shortcut your education just learn how to do it there you go and move on uh this suggests AI isn't just helping more people learn to write code or build software faster it's also attracting and helping more people become developers firsttime open source contributors continue to show widescale interest in AI projects but aren't seeing signs that AI has hurt open source with lowquality contributions I I think that's untrue I think they're uh I think they are sniffing their own farts on this one and what I mean by that is that there's that really interesting study not too long ago that showed that the average change in a line used to be once every six months and is now like every 3 weeks since the Advent of generative AI so it was like every six months a line of code would change now it's like every 3 weeks and that's just simply because in the oldie days you know back when we used to have to artisanally handcraft our code you couldn't just like rewrite whole swads of it super fast now it's like oh I can just rewrite it like I don't have to learn how you did it instead I can just rewrite it and that causes coach turn yeah uh I believe they aren't seeing a worsening all the maintainers are seeing it though yeah yeah I have also heard a lot of people talk from a maintainers perspective there's been a lot of articles that I've read about low uh lowquality bug tickets lowquality contribution poor Communications they end up being llms um security I hear security issues have just been inundated with the most lowquality crap research that actually feels very like I feel genuinely bad and so we kind of coined up a term called um it's a denial a denial of attention attack where you're able to effectively steal and siphon away all the maintainers ability to actually pay attention to the issues and then it kind of creates lowquality software by accident right python is now the most used language on GitHub as a global open source activity continues to extend beyond traditional software development again I believe this is just a manufactured stat in the sense that uh they wanted to make it so that AI is actually growing that much that python becomes number one but they also separate JavaScript from typescript which just feels you know we saw python emerged for the first time as the most useful language on GitHub or the most used language on GitHub forgive my misread there more on that later python is used heavily across machine learning data science scientific Computing hobbyist and home automation Fields among others the rise in Python usage correlates with large communities of people joining the open source Community from across the stem world rather than traditional community of software developers this year we also saw 92% I wonder how they get that this has to be from education where they're just getting like okay you're a student what are you studying oh you're a nurse therefore you're part of the stem you're not that okay I mean that must be what they're saying saying uh this because I don't know how you do that outside of the outside of the student area are they counting Forks I'm not sure that's a good question they don't tell us how they're counting things I assume Forks do count this could indicate people in data science AI machine learning and Academia increasingly use GitHub systems programming languages like Russ are also on the rise even as python JavaScript typescript and Java Remain the most widely used languages on GitHub also none of that should be surprising all right let's read some more uh Global community of developers are growing F uh 5. 2 billion contributions didn't they already say that yeah they already said it right there restating one uh fast let's see has the fastest growing developer population India new repositories in 2024 108 million wow it's a lot of that's a lot of dead code and I also am curious what they mean by new repositories I wonder if they also count Forks because Forks are a new repository regarding the Indian Dev population being on let's see being large on GitHub isn't this also due to the decision a few months ago when we learned that Indian developers create GitHub accounts and poll requests because teachers are telling them that it increases their chances of getting a job which is a bit artificial yeah there was that I'm not sure if that plays I don't know if that plays a role into this but that was a thing where there was some teaching videos on YouTube showing to make issues on real GitHub accounts and all there's a lot of that right I I don't somehow I don't think that their impact is that big all right geographical heat map of developer population around the world oh interesting okay so it is really just India United States secondary you start seeing China tertiary you start seeing Russia and what appears I I assume we're looking at Germany France Spain Poland maybe kind of hard from that one and then what do we got Brazil mentioned are they are they doing Brazil mentioned yeah dude also Sri Lanka totally not allowed in this group rip Sri Lanka not not being a part of the studying group here h h h all right there's both stability and change among the top 20 countries with the highest number of developers on GitHub while India continues to approach the number one spot we now proect by 2028 based on the updated projections more below the United States continues to have the most developers worldwide on GitHub despite this we have seen greater growth outside the US every year since 2013 and that trend has sped up over time okay so India is now number two China oh China not number one oh my gosh it's kind of interesting seeing Brazil also climb up really fast that's pretty cool cries in Austria yeah I'm sorry China third place China number three uh globally we see developer communities growing significantly Brazil India and Nigeria are especially growing fast which is notable given that they are the most populous regions in their respective continents with linguistical with linguistically diverse populations I'm also curious about uh AI usage right because this whole thing has been this AI usage business by the way I love that France just stopped being number one France died in 2020 in Indonesia kicked him out indes kicked out France but because you know this whole article is kind of gearing things towards AI but they're also saying um these countries right here and I'm curious because I assume that $500 a month is a decent amount of money in Brazil considering minimum wage is right around 200 United States Federal Reserve notes so like that would be a pretty decent amount of of money one would assume like six sa yeah it's it's like a it's it's a pretty de it's a pretty decent amount of money overall the top 20 country uh developer communities have largely maintained their positions with a few notable exceptions these include pH Philippines overtaking Australia and Pakistan overtaking Poland stay smart the rise of these non-english high population regions is notable given that it is happening at the same time as their proliferation of generative AI tools which are increasingly enabling developers to engage with code in their natural language oh interesting yeah uh I have a quick question what is the does anyone know the English speaking proportion of India I always thought that India was pretty well versed in English that could be just just like my ignorant center of the universe people are saying like 60s 7s 0420 safe bet very high English is taught in every state school right 69.
420 per depends on the region yeah uh so it is interesting though uh projecting the top 10 developer communities on GitHub through 2030 to identify the communities with the highest growth potential over the next 5 years we created the predictions based on the current trends and our predictions from the last year have changed India is now on track uh to suppress suppress the United States and number of developers by 2028 interesting I I I guess the real big question is going to be is AI actually killing jobs or not because if AI is killing jobs I assume that in the next few years this chart changes drastically I also think it's a it's an interesting kind of measurement just generally speaking to say this is where the people are from who are creating stuff on GitHub it's kind of an interesting measurement I want okay I'm going to switch back to the other article cuz I want to see where they're getting this idea that uh AI replacing develop where are they getting this cuz right now the only thing I'm seeing is that they're saying the octoverse is saying that there's a lot of new developers that doesn't mean that AI is not killing jobs or replacing developers that just simply says that there's a lot of people getting involved in programming so I'm curious what they mean by that all right so let's look at this thing AI is accelerating the developer Community AI sparked significant grow in the uh Global developer Community this past year with more contributors than ever joining GitHub yep okay see this is again this is just suggesting more people are joining GitHub AI isn't replacing us instead this year's octoport reverse shows evidence that it's nurturing a dynamic diverse Community transforming how we learn build and collaborate I think that they're taking the wrong statistic out of here like you know me I'm all for AI not taking our jobs but I'm just saying that I don't think that an increased amount of people using GitHub is proof positive that AI is making a dent either direction I think the promise of economic freedom in a world where we have been economically depressed over the last five four years with just all the stuff that's happened I think that makes way more sense as an argument than it does like ah AI was created right it's just like I don't I'm not really purchasing that as the argument I think the opportunity I think the opportunity of of economic freedom is more is more appealing uh AI needs more use cases in other domains than just coding uh AI has a lot of use cases it's just they're all just 80% accurate right let's see they're trying to say that they're being empowered not being reduced but they're only looking at the first part yeah they're only looking at the first part I well again I do not think you can even make that suggestion perhaps AI is going to make it easier for people to create an account and maybe that's just what happened maybe a whole bunch of people created an account played with AI for a moment and then that was that and then they haven't gone back like we actually don't even know the long-term ramifications by any of this all right acable insights to uh to take away one leverage AI for routine task so you can focus on bigger creative stuff this is the classic one AI tools like GitHub co-pilot reduce time SP uh spent on repetitive tasks according to the report tools like GitHub co-pilot crazy that GitHub co-pilot is mentioned so much in a report I wonder why that would happen geez it's almost like this was slightly an advertisement get up can slash vulnerab vulnerability fixing time from 1. 5 hours to just 28 minutes do anytime you see these by the way anytime you see this I just want you to just remember that when you see someone put an exact amount of minutes on how long it does it takes to do something I just want you to just just pretend that it doesn't exist okay that's it the most healthy thing for you to do is to just believe that you never even read it and just walk away CU that is crazy take because you know what happens right for those that don't know what's happening right here what they're going to do is they're going to probably take 200 developers or some small amount they're going to take a very small or medium-sized code base of something like some isolated thing they're going to give a bunch of developers who don't know the code base who aren't a part of anything and then they're going to hand them this this code base and they're going to say okay half of you get co-pilot and those so that's probably half of them that already use co-pilot so they're like oh I can move quickly and then they're going to take the other half and say you can't use co-pilot or any AIS and they're like well we already use AI so we have to stop and they're like yes you have to stop and now go and so they're going to be in this whole new environment in which they're not used to operating in because they've become dependent on the AIS then you have a bunch of people who who are dependent on the AIS being able to use AIS and then they measure the differences between these two people on a brand new brand new code base with some sort of problem and then they're like oh look at that these 200 developers that's it that's all we measured clearly that means Co a co-pilot will make you three times more efficient because that's how this works that's how programming works that's how studies work it's a completely Fair experiment to me shut up makes sense yep uh integrate smaller flexible AI models into your projects developers are increasingly working with smaller llm models like select Lama model from meta a tailored for specific taxs that require lower compute explore projects like oama and learn how to incorporate compact uh adaptable AI models into your own work I actually think this is good advice by the way I think there's probably something here that could be really useful in the sense that you know like when I was at Netflix having to work with a lot of internal tools there were pages of documentation that said a lot of things without saying a lot of things and so it could be very useful to be able to train these like specific models that are offline on some internal documents and so that way you kind of have this like really hyp specific thing to be able to help you with just navigating the internals of a company that's really large I could see that being actually a really good idea and then you don't have to get Buy in because you're not like handing all your data out to some cloud provider instead of it's some offline items so that means the doctor bad it's not that the doctor are bad they're just very verbose and when you have a bunch of veros of every kind of Doc and you have 900 different projects you have to go through you eventually have like 10 ,000 pages to be able to do something very small it's like it's very difficult right uh when working with AI uh when working with AI try python yeah that seems that SS that seems reasonable you probably you should probably just use probably should just use Python We're All in This Together oh here we go inspirational moments everyone hold hands the octe reports highlights a powerful truth the AI revolution has space for everyone from driving open source innovation in Nigeria to scaling AI startups in Brazil or spearheading genitive AI adoption in South Korea developers worldwide are reshaping the future of technology together we are redefining what it means to be a developer I'm not sure what that means and Building Technology that powers the world to be fair it's never been easier to be a developer it's never been easier to get a website up it's never been easier today to get a website up than ever it's absolutely true I don't you know it doesn't mean it's great it just means that that that's what the world is but it also becomes exceptionally difficult to do something beyond that right we have we have fully optimized the the entry curve because that's where everybody's coming in at and you can see it you can kind of see it um all right what's our next steps AI has opened the doors for developers to push the limits of what's possible the question is now how will you take advantage of this new superpower I think that this I think this article though I would love it to be true I think the data that they're showing is just not a good it's just not a good indication whether that's true or not honestly I just don't think that I don't I do not think you can drive this this is this is one of those things I think people really need to learn how to do is you really need to learn how to argue about what data means and to be able to Simply see that stuff is growing and it's growing in AI therefore it's AI is not going to replace us like you can't even make that leap right that's like that's not a leap one can make with that data the data just simply suggests that AI has sparked a bunch of people to be curious about it and want to interact with it now what's the ending thing that's going to happen will will it phizz out will it become fantastic is it a big cultural moment or is it going to just be a something that kind of goes back down the hype cycle ends and then it fits into a niche we don't actually know yet and this doesn't help us at all making one decision or the other other than it shows that a lot of people have heard about it and a lot of people are excited about it we're in the hype cycle is what it tells me we're just hype cycling right now now maybe in 6 months we can come back to this and see what has happened and maybe that will be a bit better maybe it's a big hype cycle maybe it's not a big hype cycle maybe it's actually really amazing I just don't know and I don't think anyone else can know and I do not think that these stats give you any reason to know they they look at something and assume reasons basically yeah I mean this is just classic Jacob Sullivan what does he have to say by the time llms can replace developers they will also be able to replace managers accountants HR Executives entire companies product lines they're going to be tons of code bases out there that need fixing in the next couple of decades due to some guy thinking he could replace a senior developer with an LM this is really what's going to happen by far honestly like to be completely fair I think things like scrum scrum managers scrum Masters scrum maners uh I think some level of product I think uh definitely some level of of managerial stuff some level of HR will definitely get replaced by AIS much sooner than developers I think tribal knowledge and the ability and the ability to actually work with large Legacy code bases is going to be a very valuable skill to have and I can kind of give you really good reasons why and a lot of people are going to be like well haven't you seen 3 and we can talk about the reasons with O3 I'm going to make a specific video on 03 still I haven't made that video yet but that's that's going to come out later the the short and skinny of it is it's effectively untenable so here check this out 03 um energy cost this is this is just like crazy um uh 03 uh open AI I know it was it was on LinkedIn uh someone broke down the their guess as to what it cost it was approximately five it was a approximately 5 gallons of gasoline per task solved on the arc test so solving this thing our attempting to solve this thing right here took five gallons of gasoline for O3 like that's a that's a lot of energy like when you think about it that's a lot of energy oh there yeah there it is I it looks like I missed it yeah so someone is saying I don't know this may not be true by the way I I they may be overinflating their numbers they might be under inflating their numbers you don't actually know because there's also I mean think about all the inefficiencies when you're only calculating something with respect to to a single like data point it may not be that great but effectively just something like this that can't be right oh well it can it actually can be right because if you looked at uh if you look at this top part right up here you'll notice that uh for open AI High compute model it took 5. 7 billion tokens to compute something that's like a lot of processing just whether you like it or not that's just a whole ass amount of processing 5.
