GitHub Copilot X Explained | A big step forward...

what's up everybody it's Travis here from Travis dot media so a few days ago Microsoft announced GitHub copilot X in amidst all of the chat GPT talk and the gpt4 talk I feel like this development is super relevant for developers so in this video I want to do two things one I want to talk about how GitHub copilot has been doing so far with developers how it's been ranking with developers do they like it do they not like it and two I want to answer what is GitHub copilot X and what is all of the amazingness that comes with it if you're looking for answers like will AI take my job or should I even learn to code with all of the AI stuff coming out this year then check out my last video I talk about all of that there in this video I want to stay focused on GitHub copilot X in case you didn't know it Microsoft is completely dominating AI this year from having a large share in open AI to GitHub co-pilot to all of the Bing AI tools and even Microsoft 365 copilot have you seen this yet imagine an Excel spreadsheet of data in being able to ask it to analyze it and give you three key trends or telling it to apply some color coding to the table to help it stand out or to even project the future growth of a graph you got the pass data the present data tell me where this trend is going to go or think about this prompt in PowerPoint create a five slide presentation based on a Word document and include relevant stock photos or how about Outlook summarize the emails I missed while I was out last week flag any important items for me or how about with word create a one-page draft based on this rough outline that I made or make the third paragraph more concise for me I think that's pretty amazing and I think that's rolled out to a select audience it's not out yet but I think that's going to have a big impact this year but we're not here to talk about Microsoft 365 copilot we're here to talk about GitHub copilot so GitHub co-pilot itself has been out for a while it's marketed as your AI pair programmer and it's pretty accurate if you check out the page here you'll see how it works so you can type some comments like get average runtime of successful runs in seconds and it'll actually generate the function for you with pretty good accuracy or you can just name the method like Max sum slice and it'll give you suggestions on what you should type and then here's some more examples so fetch tweets and it generates the code for you draw a scatter plot get a rating from Goodreads see how much that generates and there's JavaScript there's python here's Ruby here's typescript here's go but have developers been using this are developers happy are developers getting value out of it well GitHub did some research quantifying GitHub co-pilot's impact on developer productivity and happiness so just to summarize this I don't want to stay here too long but just to summarize this they surveyed more than 2 000 developers and here's what they found finding one developer productivity goes beyond speed developers are satisfied so improving developer satisfaction and they're able to conserve mental energy by using copilot and then you'll see here I am more productive I'm less frustrating when coding I can focus on more satisfying work things like that now finding too is that they are actually faster so they took 95 developers put them in two groups one using copilot one not using copilot in the group using copilot finished 55 faster than the other group and here's kind of a summary of the whole article we're here to support developers while they build software that's the point of these AI tools they're here to support you they're here is like an assistant to help you out to help you from getting frustrated and burned out and things like that so in our research we saw that GitHub copilot supports faster completion times conserves developers mental energy helps them focus on more satisfying work and ultimately find more fun in the coding they do so I think developers are really finding value in this and as a developer I think it's important to dig into this copilot ecosystem because chat gbt that's for like the bigger masses you can use it for your coding people do and they do successfully but entrepreneurs can use it and Technical writers can use it bloggers can use it mothers can use it kids can use it there's lots of use cases it's a wide use case for chat GPT GitHub co-pilot is specific for developers so with all of that success why GitHub copilot X so like I said they announced this a couple of days ago and the point of it is kind of found in the name GitHub copilot X this x is a placeholder for different things that copilot is going to be able to do so think GitHub co-pilot docs GitHub copilot PR GitHub copilot blocks GitHub copilot CLI that X is a placeholder for all of its capabilities so you'll see here with chat in terminal interfaces support for pull requests in early adoption of openai's gpt4 GitHub copilot X is their vision for the future of AI powered software development so that's another thing it uses the gpt4 model and not only is that more accurate and faster but the big Improvement is that it gives you 25 000 tokens or word count as opposed to the older model that gives you three thousand so when you write a prompt you can't go over 3000 with the old model that's where chat GPT is you got a 3 000 word count limit well gpt4 on top of its improvements gives you a 25 000 word count limit so that in itself was amazing to me but as I started looking more into it this is a huge Improvement in the direction that Microsoft is going is going to be massive for the developer community so look at some of these features here so first you have this chat aware conversations so in your vs code or Visual Studio you're going to have this chat window so highlight some code here's some python here highlight that and then over here you can ask it to write unit test functions for your code or you can tell it to explain your code or you can say what is this variable doing so here it says write a set of unit test functions for the selected code it writes the unit test and then it explains what it does down here it says if you're stuck solving a problem ask GitHub copilot to explain a piece of code bump it to an error have GitHub copilot fix it it'll even generate unit tests you can get back to building what's next next there's the tailored docs so instead of having to visit documentation think react or mdn or AWS you can just ask and have it generate answers based on the integrated documentation it will answer your questions based on the documentation so here it says spend less time searching and more time learning by getting personalized answers that are grounded in maintainer written documentation and we can see better examples of this here so co-pilot for docs if you look at this video here is copilot for docs the dock sets they're working with now is GitHub react.js mdn and Azure I'm sure this will be expanded but if you watch here this user chooses react.js and it asks a question my component re-renders too often and it reaches out to the react documentation and pulls in some answers there are several possible reasons why to avoid this blah blah and then over here to the right you'll get the reference to the documentation so that'll come up and you can click on it and go straight to that point in the documentation that's pretty neat in the example here how do I vertically Center a div they're pulling from the mdn docs which is a website full of documentation for so many things web development and it explains there are different ways to vertically Center a div here's some code on how to do it and some other pointers so I think that's great instead of having to visit the docs just ask GitHub copilot it'll pull from there give you the answers and the link to that specific part in the documentation now the next feature and a really big one I think is pull requests so GitHub keeps track of your work it suggests descriptions and helps reviewers reason about your changes with the Code walkthrough so when you create a PR you usually have to write up what you did what does the code do well GitHub copilot will do all of that for you it'll analyze your code and build out your PR for you and let's look at this in more detail so co-pilot for pull requests check out these examples so suggestions for your pull request description there's a couple of markers you can use so there's copilot summary which expands to a one paragraph summary of the changes in a pull request and then there's copilot walkthrough which expands to a detailed list of changes including links to the relevant pieces of code so if you look at this example here this is really really good so here's a PR you have the description you have the related issue in the explanation of changes normally you're going to describe your PR you're going to link to the related issue and then you're going to have to explain all of your changes and no developers really like to do this so what GitHub co-pilot does if you watch this you can use these markers so up here on the description they're going to put copilot summary and then under explanation of changes they'll put copilot walkthrough and so they're using these markers to tell copilot to fill in this information for them so once they process this it processes and the pr is written for them so here is the description that's going to read the code and describe what's going on and then there's the explanation of changes which is going to describe all of the changes I think that's huge and I know I'll use that personally 100 another neat feature is that if you add some code that doesn't have testing to back it it will automatically generate some tests for your code it says I found two changes that may need tests without a corresponding test Suite so there's some changes nobody wrote tests for them let me do it for you it sees your changes in the file like right here water heater.pi and it realizes hey with these changes you might need some tests and so it's going to suggest some tests for you there's a button here that says suggests tests and writes the tests for you next up ghost text this is just like Auto suggestions it just gets better at Auto suggesting you'll see it pop up here so that's neat but I think this one right here is Big resolving issues with AI so imagine somebody creates an issue in your repo they say hey you should replace tensorflow with pi torch and you get it and you're like oh that's gonna be a big task how do I do that well when someone creates the issue GitHub co-pilot can automatically describe how to solve that issue and even automatically suggest the changes you need to make so check this one out someone's creating an issue replace tensorflow with pi torch submit new issue and then you're going to type slash PR bot how how do I do this run that in GitHub copilot gives you steps on how to actually do this in the issue and then other people can say that's a good idea or I would switch this I would change this type of thing then there's another feature called PR bot suggest it says consider using PR bot suggest to see some possible code changes so they're going to type this and it's actually going to search through your code files and tell you what should be changed to implement this change now it's not going to be perfect but I think that's a big step in advancement next there's reviewing pull requests with AI so in this video we use AI to describe the changes in the pull request and to review the code so let's see if we can describe this pull request somebody send you a pull request and you're like I have no clue no context on this you can ask it to describe it to you so search through these two files and it gave a description down here this pull request makes the following changes pretty helpful then there's AI powered PR completion PR repair and review responses so I think the pr part of GitHub copilot is going to be really helpful the next feature is the GitHub copilot CLI so forget how to delete a tag ask GitHub copilot for assistance right in your terminal so what this does is you can ask it how to do something in the terminal it will tell you the command it'll build out the command for you it'll explain what the command does and then it gives you a prompt do you want to run this do you want to revise the query or cancel it so you'll see all of that right here but looking at it more in depth copilot for CLI and it's basically just assistance in your terminal examples here there are three modes of interaction there's the question mark question mark which is meant as a general purpose go-to for arbitrary shell commands so you want to know how to list JS files and it generates you this find command here with the grep addition to it and then there's git question mark which is used for git invocations so here it's saying git list all commits and it tells you the command to do that with Git and then there's GH question mark which combines the GitHub CLI query interface with the convenience of having AI generate the complicated Flags in JQ expressions for you so here you're asking list all closed PRS and it gives you this ghpr list with all of its arguments so all this is is just copilot for CLI you're going to use your CLI you forgot some commands you can just ask GitHub copilot to generate those commands for you and describe to you what it's doing and these are the four features listed on this page but there's more check this out so if you go to githubnext.com you can see what's out there what's coming next and all of that good stuff with GitHub and if you scroll down through projects you'll see copilot for docs which we looked at copilot for pull requests we looked at copilot for CLI then there's copilot voice check this out so write code without the keyboard difficulty typing use your voice to code without spelling things out and there's an example here you're not going to be able to hear this because there's a guy talking telling GitHub what to type but you can see it pop up here maybe I can bring these headphones up to my microphone get tight fair with average column values drop duplicates from the frame Titanic data a GitHub new line plot line graph of age versus Fair comp change to scatter plot show plot hey GitHub exit code mode hey GitHub run program so this guy just told GitHub copilot to type all this code for him he said import pandas import this Library get this Titanic data he didn't have to say from where he just said get the Titanic data from the web then he's like clean up the values that have null for age find the fares put this information together and plot a graph all in Python and you shouldn't have to type anything now I think this is neat because many of us talk to ourselves why we code and so if I could just talk instead of type that might be good might not be good I don't know but I think it's neat so that's copilot voice then there's code brushes so if we look at that we've added a toolbox of brushes to our Visual Studio code extension that can modify your code so check this out here's some use cases so make code more readable you'll see this long convoluted function here if you brush over it it turns into an if else statement add types so we have this code here we have a b how about we add the type of number to a b does that for you fix simple bugs they misspell this word low so how do you fix that GitHub copilot fixes it add debugging statements to console log out some important stuff make your code more robust so if you need some more browser support you can do that and more you can Define your own custom brushes as well I don't know how this works I haven't used this but it's actually out there already for you to use if you have GitHub copilot and you have the extension there's another extension called GitHub co-pilot Labs so a vs code extension for experimental applications of GitHub copilot so there's the copilot extension and then there's this extra extension for experimental applications you can sign up for it and it gives you that feature that brush brush feature actually it has a number of features it has a test generation where you can generate tests it can explain the code to you it can translate the code so you can take some python you can change it to like JavaScript and some other things so that's already out there and with that you can get this code brushing feature and then finally there's a number of things here that you can look at but the last one I want to talk about is GitHub blocks so reimagine repositories with custom interactive blocks build Rich documentation enhance your workflows and bring your code base to life so here's GitHub blocks and what it does basically is it makes your repository more lively with like charts and graphs and things like that so you'll see as I scroll down there's an image beside this code so this code actually executes into a graph here beside it here's some other charts create searchable reference documentations turn your repository structure into your table of contents if you have a bunch of Json here and want to make it nice it will put it in a chart for you like a spreadsheet chart that's kind of neat and some other things you can see here if you go to this page blocks.github next.com you can see all of the examples here it says understand the health of your community and the people who are part of it understand the structure of your code at a glance so here's like a plot graph of your code and that's GitHub blocks and that's it that's all I have to show you and that excites me I don't know about you but I find all of this exciting AI is something that developers need to embrace if you don't embrace it you're going to fall behind and you're going to be replaced all of this here is to enable us to be more efficient it's to keep us from burning out is to enable us to do a better job and I think GitHub copilot is doing a good job of moving all of this forward and that's all really I wanted to demo today if you look at this list most of it's still on the wait list so docs pull requests CLI voice there's still in a wait list code brushes there's that extension for that that I mentioned GitHub blocks has a wait list GitHub co-pilot radar incremental Code ql test pilot GitHub copilot Labs these are like usable prototypes but I think most of this you got to get on the wait list and you have to be a GitHub co-pilot paying customer it's like 10 bucks a month which I think is very reasonable for the value that you get so I hope this was helpful I hope you know more about GitHub copilot X now and how you might start using this in the coming weeks and months and if you have any questions leave a comment down below if you found this helpful thumbs up consider subscribing and I'll see you in the next video [Music]

Channel: Travis Media

Views: 45,742

Keywords: github copilot x, github copilot tutorial, github copilot, github copilot vscode, github copilot demo, copilot x, new github copilot, chatgpt, gpt-4, openai gpt 4, openai plugins, chatgpt 4, gpt-4 ai, github copilot x vscode, github copilot visual studio, github copilot CLI, github copilot PR, github copilot blocks, github copilot docs, what is github copilot, ai programming python, github copilot react, github copilot python, travis media

Id: 8_0DJ9FOlOM

Length: 18min 59sec (1139 seconds)

Published: Sun Mar 26 2023