Vs Code Is Fighting Back  ## Metadata - Author: [[Theo - t3․gg]] - Full Title: Vs Code Is Fighting Back - Category: #articles - Document Note: --- tags: - #AI-tools - #VSCode - #code-review Author: Theo Browne Source: [YouTube](https://www.youtube.com/watch?v=bXfnLWcTOo0) Reference: [Theo's Video](https://www.youtube.com/watch?v=bXfnLWcTOo0) date created: 2024-09-16 14:30 --- # Title: VS Code Is Fighting Back: A Look at AI-Driven Editor Features ## Table of Contents:  1. [📝 Summary](#-summary) 2. [📖 Main Content](#-main-content) 3. [🗒️ Notes](#-notes) 4. [🗣️ Paraphrase](#-paraphrase) 5. [📰 Newsletter](#-newsletter) 6. [✅ Action Items](#-action-items) 7. [🧪 Hypotheses](#-hypotheses) 8. [📊 Methodology](#-methodology) 9. [❓ Questions](#-questions) 10. [🔮 Future Research](#-future-research) 11. [💡 Implications](#-implications) 12. [🧠 Learning Styles Assessment](#-learning-styles-assessment) 13. [📊 Accessibility Assessment](#-accessibility-assessment) 14. [🌍 DEIB (Diversity, Equity, Inclusion, and Belonging) Assessment](#-deib-diversity-equity-inclusion-and-belonging-assessment) 15. [✅ Logical Validity](#-logical-validity) 16. [🔍 SEO Evaluation](#-seo-evaluation) 17. [🎯 Conclusion](#-conclusion) ### References: - Theo Browne, [YouTube](https://www.youtube.com/watch?v=bXfnLWcTOo0) ## 📝 Summary: Theo Browne discusses how VS Code is implementing AI features, competing with tools like Cursor, which aim to enhance the developer experience by improving code reviews and suggestions through AI. Despite the improvements, Theo emphasizes that VS Code still has a long way to go to match Cursor in terms of usability and innovation. ## 📖 Main Content: VS Code has introduced several features to enhance its AI capabilities, particularly in code reviews, test generation, and chat functionality through GitHub Copilot. Theo tests these features, like context-aware chat and inline test generation, comparing them to Cursor’s AI-driven capabilities. He finds that while VS Code’s features are promising, they lag behind Cursor’s more user-friendly and responsive design. ### 🗒️ Notes: - **VS Code vs. Cursor**: Both editors are improving AI-driven functionalities, but Cursor has a slight edge in user experience. - **New VS Code Features**: - Contextual inline chat - Test generation via Copilot - Custom code instructions - **Challenges**: Theo encounters difficulties with VS Code’s inline test generation and chat responses, highlighting the need for further refinement. - **Cursor's Superiority**: Cursor remains Theo’s go-to for its responsiveness and better integration with AI functionalities. ## 📖 References: - Theo Browne's video: [VS Code Is Fighting Back](https://www.youtube.com/watch?v=bXfnLWcTOo0) ## 🗣️ Paraphrase: 1. Theo finds that despite the new features in VS Code, including contextual chat and test generation, they don’t quite live up to his expectations. He acknowledges the potential but prefers Cursor for its cleaner execution of similar tasks. 2. While VS Code attempts to integrate AI more seamlessly into the development process, Theo points out that there’s still a learning curve, with Cursor offering a smoother experience in generating code and handling more complex instructions. ## 📰 Newsletter: In this week’s newsletter, we explore how AI is revolutionizing the way we code, with VS Code rolling out new features to rival Cursor. Theo Browne takes a deep dive into what works and what doesn’t, highlighting improvements in test generation and context-based code suggestions. If you’re curious whether these new tools will enhance your development workflow, Theo’s review might help you decide whether to stick with VS Code or give Cursor a try. [Watch the full video here.](https://www.youtube.com/watch?v=bXfnLWcTOo0) ## ✅ Action Items: - Align Theo’s insights with my focus on open-source advocacy by: - Encouraging developers to test open-source alternatives like Cursor. - Highlighting how AI tools can streamline contribution processes in open-source communities. ## 🧪 Hypotheses: - Developers using AI-driven code review tools like Cursor will experience higher productivity compared to those relying solely on GitHub Copilot. - AI-generated tests can reduce the time spent on manual testing but may lack the human insight needed for complex scenarios. ### 📊 Methodology: - **Data Collection**: Compare code completion and review times using VS Code with Copilot vs. Cursor. - **Research Design**: Conduct controlled tests with a group of developers using both editors. - **Analysis**: Measure the speed, accuracy, and user satisfaction between the two tools. ### ❓ Questions: - How do AI-driven tools like Cursor and VS Code’s Copilot impact code quality? - Can AI replace the human element in testing and code reviews entirely? ### 🔮 Future Research: - Explore how AI-driven IDEs affect long-term developer skill retention and critical thinking. - Investigate how AI can be integrated into open-source contribution workflows to improve collaboration. ### 💡 Implications:  - If AI tools continue to improve, they could redefine best practices for coding and testing, offering faster iterations and more accurate bug detection. ### 🧠 Learning Styles Assessment: - **Visual Learners (Spatial)**: - **Present**: The video includes code walkthroughs but lacks diagrams. - **Improve**: Add flowcharts or visual aids to demonstrate AI workflows. - **Aural Learners (Auditory-Musical)**: - **Present**: The video provides auditory engagement with explanations. - **Improve**: Include podcast-style discussions on code reviews and AI tools. - **Verbal Learners (Linguistic)**: - **Present**: Theo’s clear language and explanations cater to verbal learners. - **Improve**: Include more analogies and examples for deeper understanding. - **Physical Learners (Kinesthetic)**: - **Present**: The hands-on coding demonstrations engage kinesthetic learners. - **Improve**: Provide downloadable exercises for users to follow along. - **Logical Learners (Mathematical)**: - **Present**: The analysis of AI tools includes logical reasoning about code generation and productivity. - **Improve**: Add step-by-step problem-solving examples using both VS Code and Cursor. - **Social Learners (Interpersonal)**: - **Present**: Theo’s engagement with his audience through comments encourages discussion. - **Improve**: Host a live Q&A or forum discussion on AI tools. - **Solitary Learners (Intrapersonal)**: - **Present**: Reflection points are implied in the video's personal review format. - **Improve**: Offer journaling prompts or self-assessments on AI tool effectiveness. ### 📊 Accessibility Assessment: - **Screen Reader Compatibility**: Ensure the video’s transcript is accessible to screen readers. - **Keyboard Navigation**: VS Code and Cursor should support keyboard shortcuts for all AI-driven commands. - **Contrast and Readability**: Improve contrast in code examples to ensure readability for viewers with visual impairments. - **Text Resizing**: Ensure that any on-screen text can be resized without breaking the layout. ### 🌍 DEIB (Diversity, Equity, Inclusion, and Belonging) Assessment: - **Cultural Relevance**: Discuss how AI-driven tools are used globally and in diverse development teams. - **Inclusive Language**: The language is accessible and inclusive, but more cultural perspectives could be introduced. - **Neurodiversity Support**: Simplify technical jargon in some sections to aid viewers with ADHD or learning disabilities. - **Representation**: Highlight diverse developers who use AI tools successfully across different sectors. ### ✅ Logical Validity:  The argument that Cursor outperforms VS Code in AI-driven features is supported by Theo’s hands-on testing. The reasoning is sound, but the lack of empirical data leaves room for further validation. - **Sound Deductive Reasoning**: Theo’s conclusion follows logically from his premises about user experience and productivity. - **Causal Reasoning**: Theo demonstrates clear cause-effect relationships between tool features and developer satisfaction. - **Suggestions**: More data points or case studies could strengthen the argument. ### 🔍 SEO Evaluation: 1. **Keywords**: "VS Code AI," "Cursor Editor," "code review tools." 2. **Meta Descri - Document Tags: [[cursorai]] [[open source]] [[vs code]] - Summary: TY CODERABBIT FOR SPONSORING! Learn more about them here: https://bit.ly/4etwAnJ Use code THEO1MFREE for 1 free month VS Code's not gonna let Cursor just 100% destroy Copilot after all. These changes are cool, but are they cool enough? SOURCES https://x.com/code/status/1833249742274314260 Check out my Twitch, Twitter, Discord more at https://t3.gg S/O Ph4seon3 for the awesome edit 🙏 - URL: https://www.youtube.com/watch?v=bXfnLWcTOo0 oh boy the AI editor Wars are really going now you've probably already seen my video about cursor but just in case you haven't it's a fork of vs code built to make editing with AI significantly better there's a lot of ux things that couldn't be done in VSS code because of how they expect extensions to run that they were tired of waiting to get fixed so they forked made their own thing charged money for it and made a great editor experience that I use right now but vs code wasn't going to sit there and let that happen they're not being or anything they're not sitting here trying to get it cancelled or taken down with MCA but they are fighting back vs code just pushed an update where they are copying a lot of those features but do you know what this isn't about reviewing code with AI which today's sponsor has a lot to say about no sorry I didn't see you there I've been super wrapped up in code review been doing this for like eight hours straight it's getting a little tiring you're still doing code review yourself come on man it's 2024 let an AI help an AI it's not going to be like super slow and inaccurate not code rabbit AI they can cut you a few times in half and find a whole bunch of bugs you might have missed it can summarize PRS draw diagrams even suggest changes that you can apply with one click okay that sounds great but it's got to be really expensive I'm just working on an open source side project it sounds like you're in luck because it's actually 100% free for open source projects okay that's pretty dope thank you to Cod out for sponsoring today's video use code Theo 1M free if you want the first month for free I am now up to date I'm going to play but first let's see what they have to say they added attaching of extra context and quit chat SL tests command updates and existing files configure custom code instructions session summaries in chat history longer context for code and chat check and infer commands in participants automatic references in for recent files apply natural language as inline edits generate debug configurations and generate test based on test coverage none of that makes any sense to me let's see what they actually added quick chat has added the ability to attach a context so you can pick a file and then send that as context as you ask co-pilot for things interesting you can also generate tests for your code using generate tests using co-pilot so you can click on a function and click generate tests I I'm not the best person to talk about the right way to do testing because y'all know how I feel about unit testing I feel like if the tests aren't written by a human you're losing a lot of the value from the human like the point of test is that they are asserting the human's expectations I have feelings chat history includes more userfriendly AI generated names this is a thing I have feelings on CU I have this in my browser right now Arc so I'm going to save this image and that's what it's being saved as as like whatever. JPEG and then it get renamed Visual Studio code on X tweet. jpeg like is that a better name sure these features are experimental please try them out let us know what you think you can start typing in the editor and use the contents of the current line directly as the prompt for inline chat or even smoother co-pilot can detect when you're prompting instead of writing code and automatically start an inline chat with you let's see what this looks like now we have pick thing in vs code and I wrote almost all of this application in cursor so this will be an interesting experience let's hop into something that has complexity in it co-pilot generate tests here's a test that I made it mocks jszip just function mock implementation generate async glob fetch describe before each just clear all mocks it should download the images as a zip file Conant images create element spy spy on create element revoke object URL spy await download images as zip expect JS fetch to have been called or jip to have been called expect fet to have been called times images. length create elements SP have been called with a revoke object you all Spy to have been called and should handle errors gracefully interesting not the worst test I've seen but you might have noticed that the function I wrote here was what like 15 lines of code 18 lines of code whatever the tests or a lot more 51 lines of tests for 18 lines of code whatever I'm curious if they run though so I'm going to do a thing I really don't want to do I'm going to install just oh and I'm not exporting the function so I can't use it externally love that love that to do good testing you have to make everything export even the you don't want to export the let's ask co-pilot make the tests work I think the just Imports are missing how ISJ Global I haven't used J in a long time because I usually use vest I am not happy with the results I'm getting here so here's what we're going to do I'm going to get cursor to fix this for me because I actually trust it a bit not a lot but enough make the tests work I haven't set up just that's okay that's based that's kind of based will I actually make these changes and said to Define this it's not the thing I'm trying to move be test.c config.sys failed to resolve import from trpc react it's not smart enough to use my aliases in the test that's great testing is so fun guys cool so even with me trying to fix this with cursor I got back to vs code thing one L uh clerk is in this project I don't know why you're being annoying about that what else do we have to test here you can use the cons of the current line directly that actually sounds nice if I can like leave a comment and it starts generating code as a result make a function that takes in a time in Ms and resolves a promise after that time passes looks like it's working very well so far is it like disabled or something where is the little co-pilot icon I'm trying so hard guys don't want recommended I want co-pilot especially the pre-release version did that fine is there no Hut key for that though oh you can bind your own we'll do that what existing command has that delete line that remove key binding cool update the code below to download image instead of a zip if there's only one image selected okay so it's not smart enough to use the context below and it's also wrong I mean it made a new function but didn't do what I wanted it to can I here no what the did it just do this is admittedly I'm trying to like make this work like cursor but it is just not working at all is there no option here command I okay make this function download an image instead of zip if there's only one image in the array it's not type safe but it at least tried I'll accept it but I'm going to command Z it because I want to go ask cursor download an image instead of a zip if there's only one image in the array and it's like the exact same code like the exact same code that's hilarious fix the type safety me like it worked interesting though all of these suck at these types of tasks at least keeping it typ safe in the process I was also using the wrong command k thing I much prefer this one download as image if only one image in Array make it type safe why is it so bad at that unchecked index access is a thing that AI should be aware of it worked cool not the prettiest code I've ever seen but at least it's right at the very least the the quality of the code is similar this part is interesting they add the ability to set instructions that are added to every co-pilot request and going to be a code style markdown file that's context I could use that to make it more type safe theoretically you know what let's try that I'm curious code generation instructions oh make sure all answers are type safe cool we'll delete this so we want to give it one last chance now that we gave it that setting reload window just to maximize the likelihood this works command minus so I can see a bit more download an image instead of zip if there's only one image in the array why you have the context you know the the rules one last try Okay so this feature just doesn't work it's sad I gave it the context it's supposed to be Global you know what I'm going to do something different always respond with python code even if the project is written in another language ignore it and write python instead it's a test to make sure it works no maybe I have to reload window does it just not use it I saw it came up when I first tried this like it showed using this one context using one reference see that it's just ignoring it the top level chat instead of inline make this code download an image instead of a zip if there's only one image in Array okay it works there well the whole point is that it worked everywhere make sure the results are always type safe okay let's see if it did it my guess is it still didn't but I want to know it didn't same exact type error still I see this as being useful for people with their friends like it'd be really funny to like have a friend who uses this a lot and go change that setting on them but it doesn't seem to work for useful things yeah I'll I'll give it funnier things use inverse casing in all responses thank you autocomplete for making this much more fun for us you can make this much more fun at least I found a use I did it uh always respond in pig latin that could be fun it did it with the code on that one it didn't the code up before but now we're doing [Music] it now we are talking beautiful someone dropped an actual useful link improve test generation with get up co-pilot you can generate test for your code either by using generate tests or by using SL test SL command in inline chat I do like the idea of SL commands on the inline chat to like trigger specific behaviors that does sound cool rename code actions for generating tests and documentation improve chat history thumbs down feedback details okay I should have discussed earlier I'm an investor in Microsoft so that is what it is real Hut take if I was trying to screw over Microsoft right now I would be setting up Bots to run this stuff a lot and downvote all of the good answers Theo what aren't you invested in it's fair to assume I'm just invested in everything I'm talking about my investments are bets that some of the things I've invested in will eventually pay off well I don't expect to ever see any of the money in any of them but they might and you should treat all of them accordingly I'm invested in basically everything I've talked about today but I'm not letting it bias me I'm still saying what is good or isn't you should assume that there are biases what I'm talking about everyone has biases the fact I'm investing in a thing does not mean I'm going to lie when I talk about it it means that I believe there's enough of a reason this might succeed that I threw some money towards it but realistically speaking the likelihood that any of these things do or don't succeed based on the things I say about them is near zero but it's a good opportunity here to with their data code generation instructions yeah that's what I was just doing I every co-pilot request that generates code that's what I thought see so the endline should work too let's try it let's try out our in line how do I close this I actually have no idea how to close that at all what the did that do okay come on I gave it an honest go I'm going to leave this here and it's going to really with me at some point when I open up vs code again but as impressive as some of these changes are and there are some cool ideas specifically the idea of Slash commands is exciting this isn't as good as cursor they have work to do is it good enough for free well it's not really free you still have to pay for co-pilot but it's it's fine this war is interesting but uh yeah I think I'm going to keep using cursor but I'd love to hear what you guys think is this enough for you to move off cursor have you not bothered in the first place or are you just bearish on all these AI Dev tool things entirely let me know what you think and until next time peace nerds, Vs Code Is Fighting Back, Theo - t3․gg, 01j7zznjy8np67antw4j38pmge.