Claude Code or Github Copilot: A Practical Comparison!

本站添加: 2026-05-11

[00:00:00]Hello everyone. In this video, we are going to talk about Cloud Code and GitHub Copilot, comparing their features and how they are working, different options they have. I have um GitHub Copilot subscription, as well as Cloud Code. So, it's a good chance to just sit together and see uh what they are offering to us. Here, you can see the CLI interface for both of them. Once you're running the cloud, you can see such a this uh interface for both of them. I run the Cloud Code uh with this option, dangerously escape permissions. Usually, if you know what you're doing, to be honest, you can enable this option, which uh just prevent Cloud Code to asking many questions during your your session, and then you let uh cloud to do whatever it wants, right? So, just be careful when you're enabling this one.

[00:01:03]And usually, I use that one, but I have some kind of restriction for it. Put some instruction, you cannot delete file outside of the current directory or changing any operating system settings, something like that, to just guard it with um security issues. All right? So, I have this directory, to-do app cloud, and in the Copilot as well, I run it the same option, which uh it has allow all. So, basically, you just run Copilot {dash} {dash} allow all, and then it will do the same as this dangerously escape permissions. All right? So, these are empty uh directories. So, we can uh we are going to uh work on them and create a to-do app. But, before that, let's consider um some commands. So, here it is the the CLI interface, as I said. But, later go to the cloud we go to the VS Code and then also checking their extension as well. All right? So, the first thing you can check is models. What they offer for LLM models to generate code, right? So, in the cloud code because it is a explicitly for the Anthropic company, it only offers their own models, which is Sonnet, Opus, and Haiku. I am using Sonnet here.

[00:02:39]Uh for Opus, if you want to do some complicated task, such as refactoring a big code base, or planning a uh features that is very big, Opus is the best options you can use. But, as you can see, here it is saying two times more costlier than Sonnet, right? So, it is a trade-off for you. If you want to get a better job, you have to pay more. Haiku also is just for conversation, not good for coding. So, here I'm using Sonnet. Let's see what we have in the Copilot.

[00:03:20]So, Copilot, as it is a kind of interface for different models, because it's it is for Microsoft, they don't have their own model.

[00:03:33]So, they are offering different actually models that you can use, such as ChatGPT or Claude.

[00:03:42]And because I am going to just compare both of them, I'm going to use the same model with the same effort, which is here is high. You can have if for example if you want to use this the Sonnet 4.5 it is high, right? So let's just back to that Sonnet 4.6 and it's asking is it low, medium or high which is more regarding to the how you want them to resonate about the the problem about the task they are doing.

[00:04:17]So because here in the plot it is uh putting on high effort. Here also I am using the same to have a better comparison. All right.

[00:04:28]So the next command that I want to use so first of all here as you can see there is lots of commands both in the cloud code and the copilot. You can use whatever you want but some of them are really outstanding and I want to more focus on them. I don't want to go through all of the the commands which is like taking so much time. One of the the good features that both of them actually have is remote. So this the remote control you can use this remote control that allows you to run some task on like your cloud code or your mobile application if you have the cloud mobile application as well then you can manage your sessions.

[00:05:22]The same thing we have also on the copilot as well. So you can see your statuses from GitHub or the mobile app for GitHub, right? So it's it's a good good command. Sometimes maybe you have something else to do so you just put remote and then it you can check your session status outside. I don't know. You you want to go out buying something or doing something else. Then you can check your phone to see the session is done or no, cancel it or manage the the sessions you have. Right, one of the the other things that I want to highlight here is plan. Right, the plan mode also you as you can see the commands are similar together, which is a good one to be honest because once you're switching between these tools, it's good to have similarity between them. Otherwise, it is very difficult to switch and it's a learning curve as well. So, having these options similar is really helpful. So, the plan mode, if you want to uh just plan for some features into your application. Like you don't know what you want to do or you don't know what exactly based on your code base the features that you want to implement, how it will be. Then if you put on the plan mode, it will start asking you what you want to do. Right, so later on we can quickly check. Cloud also introduced a new feature called ultra plan, which kind of you can as based on this description, you can have different Well, actually, ultra plan, let's say like that. I don't know what other words I can use. So, you can like on your website on the cloud web, you can see the plans, edit it or other stuff you can use. But usually I I just for now I use the the plan one. And same we have also in the copilot. Right, so you can plan, put your command prompt and then start planning for that.

[00:07:39]So, let's put one prompt here. What I want to do, I have a prompt to create a to-do app using pure HTML, CSS, and vanilla JavaScript. And asking to add, edit, delete, complete the task we have. Use separate index, style, and script. Right? So, I will put the same um prompt here into the Copilot, and let's see. Let's see how they are working.

[00:08:09]Remember that the both of them are using Sonnet as the LLM model, and let's see what they are doing for us. I will, like, I start um the the tasking in the You can see different uh reasoning for the models.

[00:08:26]Here in the Copilot, you can see how they are thinking here. User wants me to create a responsive to-do app, blah blah blah. And then here in the Claude, it is actually showing you the code that it is generating.

[00:08:41]Uh so, different uh I would say interface, but at the end, they are doing the same thing. But, uh let's see the the output. Uh we will uh run the app to see how they are doing. And also, we can now see the performance.

[00:08:59]Uh which of them are faster, which of them are like putting a more nicer, let's say, app that we ask uh to generate.

[00:09:11]Yeah, so here it uh writing the HTML, style CSS, and JavaScript script one. Uh it is all um the It's all uh pure ja- JavaScript. So, uh both of them are the the the performance similar. Uh just took 1 minute and 10 seconds to uh to generate, and here the same.

[00:09:37]Uh it is like listing all the features it has listing the um the HTML and the file generates different wording let's say like that all right.

[00:09:49]So what I want to do here is let's I will add well no not this one I will add another [clears throat] window here and let's go to source temp and then yeah.

[00:10:11]Well actually this is the to-do plot okay and same I want to actually open the code so here I will go to co-pilot one right let's open this one and then also explorer so as you can see both of them created these three files as we asked co-pilot and Claude right so let's open the Claude one index here it is the kind of the interface it has how about the the co-pilot all right so co-pilot let let me put them together here as you can see co-pilot to-do app Claude my tasks hmm like I little more like the co-pilot generated app right let's add some task here task one task two, task three.

[00:11:37]And then, how about this one? Task one, task two, task three.

[00:11:44]Okay.

[00:11:46]And um let's make this one done. Here, this is the one. How about this one?

[00:11:53]Hmm, I would say Cloud is a little like, I don't know, simpler um UI generated.

[00:12:04]Co-pilot added more um like, I don't know, nice UI to it. That is interesting. Both of them has edit and delete.

[00:12:16]You can see.

[00:12:18]Let's edit this one. Oh, for edit, it open up a a model.

[00:12:24]So, let's just ask it to 20. How about this one? Oh, okay, nice. So, as you can see, like, they are approaching um the same thing um for doing the the same task because they are using the same model, right?

[00:12:44]Uh how about delete one? Okay, deletes and this one. All right.

[00:12:50]One task left here. It's saying one task remaining here.

[00:12:55]And I asked to local storage. So, if I just uh refresh this one, okay, it it um kind of keeping the the status of the task.

[00:13:08]All right, not that bad, right? Co-pilot is a little wins here.

[00:13:13]Uh let's go to actually the checking the VS Code, all right? So, I will open uh the VS Code for this one, and also uh code for Claude one.

[00:13:29]So, I will put them together as well.

[00:13:32]Okay.

[00:13:34]So, here actually, if I just trying to uh, do them if it is a little bigger. So, in the the Copilot, you have chat. Also, you can open it with this toggle one, right?

[00:13:55]And here you can see you we have Claude on it. And if I switch to the Claude, it's a little different the the extension in the VS code, but we can see a index.html, a script, style.

[00:14:12]And um, here in the uh, Copilot, you can see the um, the sessions it has.

[00:14:20]Uh, well, actually here also you can see in the Claude the the previous sessions um, you were working with them. So, what I want to do uh, here because um, the you can you have different options here, changing the model and probably configure tools, the agent mode, I don't know, ask, plan, different stuff here. So, what I want to do is asking uh, from here, not from the CLI, another prompt here.

[00:14:55]Just to add dark light theme.

[00:14:59]And then here also, I will add dark add um, dark light theme.

[00:15:13]Okay, let's do this one here as well.

[00:15:18]This window is for Copilot. This one is Claude.

[00:15:26]Let's see how they are doing.

[00:15:31]First reading all the files and trying to actually adding the team here for um the having dark or light theme.

[00:15:45]All right. It seems Copilot is a little faster.

[00:15:50]Let's see um how they are acting here. It It's really beautiful to see uh what they are doing, right? So, as you can see um all right. So, Copilot is done.

[00:16:06]Yeah.

[00:16:07]And uh it's done the job.

[00:16:11]And Cloud Code is still doing um let's see how it is doing. I'm I'm not sure. Okay. It's It's done.

[00:16:21]So, uh let's let's see. I will keep this um Copilot extension and will show you uh what's changed. I will keep all of them.

[00:16:32]And then, let's go back to um the the UI. All right. So, if I This is the four Cloud Code. If I refresh, okay, I can see. How about this one? All right.

[00:16:47]Um so, you can see Cloud used the the black one as a background. It's a little like I don't know uh not black.

[00:16:58]So, let's see. Okay. It's working.

[00:17:01]Yeah.

[00:17:03]Nice.

[00:17:06]Let's add another task.

[00:17:08]Here.

[00:17:10]Task.

[00:17:11]Okay. That's That's uh good, actually.

[00:17:15]Um so, yeah. Here you could see different options, different uh behaviors, I would say, from the tools, although they are using the same LLM models, Sonnet I used here.

[00:17:30]But still, you can see different functionalities, different um UI they generate. Um, so, still I like the Copilot, uh at least for the UI it generates here.

[00:17:45]But I am using Cloud Code mainly, to be honest. Uh but here, in this demo, yeah, you can see uh little Copilot wins.

[00:17:55]So, let's go back to um the CLI just to check uh different other commands.

[00:18:05]Another command that you can use, actually, is uh usage.

[00:18:11]So, here you can see in this uh so far, how much you used for the Cloud.

[00:18:20]Right? So, here it is current session, 12%, current week, 7% I used Cloud.

[00:18:28]And how about the uh the Copilot? Here you can see the changes, 60 60 uh plot changes, one premium request, and also token it used. Send 82K and get 6K uh token, right?

[00:18:49]Another one, which is related to the current session, is context.

[00:18:57]So, if you put the uh round the context here, this showing you uh based on the current model, as you can see the um the the token, the maximum token in one session is 200k token for the Claude tool and co-pilot it is actually 160 token. So, another thing, window token means the context of one session means you can do more interact with the same session more until two a thousand but it is a good practice to not expand it. I mean, once you hit the 100k let's say, you need to either run a new session or clear or compact. Right? So, as we can see it is 60% well actually 20% and here it is 13% different options here, system prompts, system tools, how they are the token used and also in the co-pilot little less options to tell you where they use those tokens.

[00:20:19]But, once you have a very long conversation with the the LLM, you have two options. One is clear right? To clear all of the context in the current session. It's just exactly like running a new session. And same you have here in the the Claude code as well.

[00:20:42]If you want to keep continue but your token is like reaching the maximum, you have another option called compact.

[00:20:50]Compact is actually summarize your your conversation history and the same you have in the co-pilot as well.

[00:21:00]But, I didn't find it useful to be honest.

[00:21:04]So, I try to like either clear or just run a new session after like I I did my job. So, I try to keep the sessions small rather than just I don't know trying to um conversation doing conversation small messages small prompts. It's better to be organized because as you can as you know that like the Copilot recently changed the way that they are billing it is based on the token usage Cloud still didn't but probably will be like that.

[00:21:43]So, it's not just like a messaging tool.

[00:21:48]You are a developer. You need to be organized. You need to be like even plan mode is is a good one. You first you plan what you want to implement and then start a new session and then just run implement that um session, right?

[00:22:08]One point that I wanted to do here. So, if you change like or like in middle of the conversation, you want to change the model.

[00:22:19]Be careful because if you now if I want to switch because I have like I don't know 30k token use. If I want to switch now, it will consume feeding all of those token to the new model as well.

[00:22:36]Which again unnecessary cost for you.

[00:22:41]If you want to change the model, just clear or run a new session and then switch to the new model, okay? Same for the um same for the Cloud as well. So, that is like a point that how you you can manage your um your model and the the tool you're using. Well, before we didn't have any code here, right? So, if I want if you like run your Copilot or Cloud Code in a code base existing code, it ask you to uh create a instruction file.

[00:23:23]In the Copilot, it called Copilot instruction. In the Cloud Code, it called plot.md, right? So, what how you can create that one? Simply just by running in it. Okay?

[00:23:37]So, if I run in it here and same in it here as well, it will like read all the code in the current code base which those HTML file and then it will create some kind of instruction that will be using every time that you are interacting with uh the the tool.

[00:24:01]So, it's good to keep it very like light. I don't know, less than 100 line, let's say.

[00:24:09]Because it is every time on every prompt, it will load that 100 lines. So, if as long as you're you're having So, if you have a long MD file, it will using more token. All right? So, let's just Okay, both of them created. If I go back to the um the code, you can see it creates a folder for the Copilot and then it puts Copilot instruction.

[00:24:44]Like it is just trying to understand what it is this current project and different things. So, it generates 64 line. How about the the cloud.md?

[00:24:57]Again, the same thing, same structure, and parity line, which is way better.

[00:25:03]Here, Cloud Code is doing better because that is very important. Every time you're going to load this one. So, 60 is just like double usage for your tool.

[00:25:17]Okay?

[00:25:18]And what else I wanted to highlight probably in the next videos, we will talk about you have also different options like MCP servers. Here I am using Context 7 and same here in the Copilot. This is the Microsoft layer which is coming from plugin and the Context 7.

[00:25:42]Also, you can use skills which is a new one.

[00:25:47]Not new actually, but recently Anthropic introduced it. You can create a skills for your agents. Same for here. I mean, as I said, all of the command mostly are the same.

[00:26:03]And you can add skills to your agent and it will be using on demand based on the changes you're doing. So, in later videos, I'll try to create more. Let me know in the comments if you want something else based on the features or what tools you're using. Are you using Cloud Code or Copilot? Yeah, share with us.

[00:26:30]Um That was it. I tried to compare them as much as I I could.

[00:26:36]So, hopefully this video will give you better picture if you want to choose between these two tools which is probably best one in the the market.

[00:26:47]Thank you for your watching. Goodbye.