YouTube Deep SummaryYouTube Deep Summary

Star Extract content that makes a tangible impact on your life

Video thumbnail

Gemini CLI vs Claude Code: Which is the Best AI Coding Tool Ever?

Alex Finn • 23:52 minutes • Published 2025-06-28 • YouTube

🤖 AI Summaries (9 chapters):

🤖 AI-Generated Summary:

📚 Video Chapters (9 chapters):

📹 Video Information:

Title: Gemini CLI vs Claude Code: Which is the Best AI Coding Tool Ever?
Channel: Alex Finn
Duration: 23:52
Views: 22,736

Overview

This video presents a detailed, head-to-head comparison of two AI coding tools: Google's Gemini CLI and Anthropic's Claude Code. The creator builds an identical, complex stock tracking app with each tool, evaluating them across multiple criteria—speed, accuracy, "taste" (design/UX), and ease of use—while guiding viewers through each version's development. Each chapter builds on the previous, progressively revealing the strengths and weaknesses of both tools, ultimately providing viewers with actionable recommendations based on the real-world performance and user experience of each AI assistant.


Chapter-by-Chapter Deep Dive

Intro (00:00)

Core Concepts & Main Points:
- Introduction of Gemini CLI as Google's direct competitor to Claude Code.
- The main purpose: Determine which AI coding tool is superior by building the same app side-by-side with each.
- Notable difference: Gemini CLI offers the first 1,000 requests per day for free; Claude Code can cost up to $200/month for serious use.

Key Insights:
- Cost is a significant differentiator, but performance and user experience will be the true deciding factors.

Actionable Advice:
- Viewers are encouraged to follow along and build their own app as the video progresses.

Connection to Overall Theme:
- Sets the stage for a practical, feature-focused competition between two leading AI coding tools.


Competition begins (01:36)

Core Concepts & Main Points:
- The competition structure: Each AI runs in its own instance of the Cursor IDE, building the same stock tracking app.
- The apps will be evaluated on speed, accuracy, taste, and ease of use.
- Installation steps are briefly mentioned, with commands available in the video description.

Key Insights:
- The challenge is intentionally complex, moving beyond basic app examples to truly test the AIs.

Actionable Advice:
- Viewers can set up both tools using the provided commands to build alongside the creator.

Connection to Overall Theme:
- Establishes fairness and rigor in the head-to-head comparison, making findings relevant for serious developers.


The prompt (02:47)

Core Concepts & Main Points:
- The prompt, generated by Claude, specifies a stock tracking app with a chart, AI company summary, sentiment analysis, and a watch list.
- Both AIs receive the exact same prompt to ensure comparability.

Key Insights:
- Prompt quality is vital: a detailed, well-thought-out prompt leads to better, more autonomous AI performance.
- Initial behaviors: Claude Code builds a step-by-step to-do list; Gemini CLI provides an up-front plan and uses Next.js, which is more modern than Claude’s HTML/CSS approach.

Actionable Advice:
- Use detailed prompts for better AI output and reduced need for oversight.
- Consider the underlying technologies (e.g., Next.js vs. simple HTML) when building apps for future extensibility.

Connection to Overall Theme:
- Emphasizes the importance of prompt engineering and initial app scaffolding in AI-powered development.


V1 (05:38)

Core Concepts & Main Points:
- First working versions (V1) from both AIs are evaluated.
- Claude Code’s V1 is clean but uses mock data; Gemini CLI’s V1 uses real data for stock prices.

Key Insights:
- Gemini's speed and real-data integration are impressive, but its UI is less polished.
- Claude Code excels in user experience and design ("taste"), but lags on real data integration.

Actionable Advice:
- Check if your AI-generated app is using real or mock data before moving ahead.
- Prioritize both functionality and design for a usable product.

Connection to Overall Theme:
- Begins to highlight the trade-offs between immediate functionality and refined user experience.


V2 (07:04)

Core Concepts & Main Points:
- Both AIs advance: Gemini sticks closely to the prompt, while Claude Code autonomously adds extra features.
- Claude maintains and updates a to-do list; Gemini does not.
- Prompt quality is again emphasized.

Key Insights:
- Claude is creative and proactive but can go "off script."
- Gemini is more literal and focused.
- A superior initial prompt saves time and reduces errors.

Actionable Advice:
- Let your LLM generate the initial prompt for your project.
- Use to-do lists or stepwise approaches for better project management with AI tools.
- Approve steps rather than micromanaging, provided your prompt is strong.

Connection to Overall Theme:
- Contrasts the personalities and approaches of the two AI tools, reinforcing the importance of prompt engineering and project management style.


V3 (09:59)

Core Concepts & Main Points:
- Both apps expand features: Claude Code adds an alert button and company news, but with mock data; Gemini refines charts and watch lists.
- Gemini is more accurate in basic functionality but lacks detailed sentiment explanations.

Key Insights:
- Claude Code tends to add features not requested, which can be a double-edged sword.
- Gemini’s output is more utilitarian—plain but functional.

Actionable Advice:
- If you want creative enhancements, Claude might be better; for strict adherence, Gemini is preferable.
- Test for requested features and clarity (e.g., explanations for sentiment analysis).

Connection to Overall Theme:
- Further explores the balance between creativity and strictness in AI outputs.


V4 (12:10)

Core Concepts & Main Points:
- Claude Code catches up on functionality and continues to add creative, unprompted features.
- Gemini runs into its first error; Claude Code remains error-free.

Key Insights:
- Reliability is a key differentiator: Claude Code is stable; Gemini can be fragile.
- Debugging workflow: Copy errors directly from the console and feed them back to the AI for troubleshooting.

Actionable Advice:
- When encountering errors, use a simple feedback loop: paste the error into the AI and ask for a fix.
- Value stable, "ironed out" tools for production work.

Connection to Overall Theme:
- Demonstrates the practical realities of AI coding: error handling and the importance of stability for developer productivity.


Final Version (17:40)

Core Concepts & Main Points:
- Final feature check: Claude Code’s app has technical indicators, news feed, and a polished UI/UX.
- Gemini CLI is stuck in an error loop and becomes unusable.

Key Insights:
- Claude Code’s app is visually appealing, intuitive, and "human-centric."
- Gemini CLI, despite initial speed and intelligence, ultimately falters due to bugs and lack of polish.

Actionable Advice:
- For a production-ready, user-friendly app, prioritize tools that consistently deliver polished, stable results.
- Be ready to input API keys or make minor tweaks for advanced features.

Connection to Overall Theme:
- The journey highlights the difference between building something that “works” and building something users want to use.


Scoring (19:55)

Core Concepts & Main Points:
- Final ratings by category: speed, accuracy, taste, ease of use.
- Claude Code: 7 (speed), 9 (accuracy), 10 (taste), 7 (ease) = 33/40
- Gemini CLI: 8 (speed), 6 (accuracy), 5 (taste), 8 (ease) = 27/40

Key Insights:
- Claude Code is superior in accuracy and taste (usability/design).
- Gemini CLI is faster and a bit easier at first, but error-prone and less appealing.

Actionable Advice:
- If budget is tight, use Gemini CLI to build and launch your first app for free, then upgrade to Claude Code if/when possible.
- Monetize early efforts and reinvest in better tools.

Connection to Overall Theme:
- The competition underscores the value of both tools, but positions Claude Code as the best all-around choice, especially for those who prioritize quality and reliability.


Cross-Chapter Synthesis

Recurring Themes & Building Concepts

  • Prompt Engineering: From “The prompt” (02:47) through “V2” (07:04), the importance of crafting a detailed, thoughtful prompt is repeatedly emphasized. Good prompts enable both AIs to work more autonomously and efficiently.
  • Trade-offs: Early chapters (“V1” and “V2”) reveal Claude's superior taste/design but slower speed and reliance on mock data, while Gemini excels in initial speed and real data integration but lacks polish. These trade-offs are explored and iterated upon in subsequent versions.
  • Reliability vs. Creativity: Claude Code is shown to be more creative, adding features beyond the prompt (sometimes useful, sometimes superfluous), and extremely reliable (rarely producing errors). Gemini is literal, focused, and fast, but less reliable as complexity increases.
  • Debugging & Error Handling: Starting in “V4” (12:10), Gemini’s errors illustrate the importance of robust error handling and the value of stable tools. The recommended debugging approach is to iteratively paste errors back into the AI for fixes.

Progressive Learning Journey

  • The video begins by introducing the contenders and establishing a fair, rigorous test environment.
  • Each chapter adds complexity to the app, exposing deeper strengths and weaknesses of the AI tools.
  • Viewers learn not only which tool is better but also how to prompt, manage, and debug AI-powered coding workflows.
  • The final scoring and recommendations synthesize these lessons, giving practical guidance for both beginners and experienced AI developers.

Most Important Points by Chapter

  • Intro: Cost and new competition in the AI coding space.
  • The prompt: Detailed prompts fuel better AI coding outcomes.
  • V1/V2: Early strengths: Gemini for speed and data, Claude for UX and to-do management.
  • V3/V4: Reliability and error management become crucial as complexity grows.
  • Final Version/Scoring: Claude Code is the overall winner for polished, stable, user-friendly apps, with actionable next steps for budget-conscious users.

Actionable Strategies by Chapter

Chapter Practical Advice & Strategies Warnings/Pitfalls Resources/Next Steps
Intro Try both Gemini CLI (free) and Claude Code (paid) to compare for your needs. Claude Code can be expensive. Installation commands in video description.
Competition Set up both tools in Cursor IDE for head-to-head comparisons. None specified. Follow along to build your own app.
The prompt Use LLMs to generate detailed, high-quality prompts before starting. Consider underlying tech stack (Next.js vs. HTML/CSS). Poor prompts = more errors, manual fixes. Copy-paste the sample prompt for your project.
V1 Validate whether AI-generated apps use real or mock data before proceeding. Mistaking mock data for real data. Test data sources early in development.
V2 Let Claude or another LLM help generate a robust, detailed prompt. Use stepwise to-do lists where possible. Sloppy prompts require handholding. Approve steps rather than micromanage.
V3 Test for all requested features and clarify output requirements (e.g., for sentiment analysis explanations). Extra features can distract from core. Review AI-generated output critically.
V4 When errors occur, copy and paste them back into the AI for troubleshooting. Favor tools with proven stability for production work. Gemini prone to error rabbit holes. Use console logs for error messages.
Final Version For production-ready apps, choose tools that deliver stable, polished results even if they are slower. Be ready to supply API keys for advanced features. Relying on "almost working" features. Prepare to tweak API integrations as needed.
Scoring For budget users: Start with Gemini CLI, monetize, then upgrade to Claude Code. Invest in tools that match your current needs and resources. Gemini can frustrate beginners. Link to monetization video in description.

Warnings/Pitfalls:
- Claude Code: Expensive for high usage ($200/month for serious use).
- Gemini CLI: Prone to errors and can become unusable in more complex apps; debugging can become a "rabbit hole," especially for beginners.
- Prompting: Sloppy or overly vague prompts lead to more manual correction and oversight.

Resources/Next Steps:
- Install commands for both tools are in the video description.
- Link to a video on app monetization for those using Gemini CLI and wanting to upgrade.
- Weekly live streams where the creator builds apps live for additional learning.


This chapter-structured summary provides a clear roadmap for understanding the strengths, weaknesses, and practical realities of Gemini CLI and Claude Code in AI-driven app development, with actionable insights at each step.


📝 Transcript Chapters (9 chapters):

📝 Transcript (692 entries):

## Intro [00:00] [00:00] Google just released what might be the [00:01] most insane AI coding tool ever. But the number one question on everyone's mind is, does it beat the king of AI coding, Claude Code? What if I told you one of these is clearly the king of AI coding? But it's not for the reason you think. [00:16] In this video, I'll put Gemini CLI and [00:19] Claude Code head to head to figure out [00:22] which is the best AI coding tool ever. We'll build an app side to side on both. You'll learn how to use each and by the end of the video, you'll know which tool is king and which you can throw away forever. Let's get into it. So, a few [00:34] days ago, Google released their claw [00:36] code competitor, Gemini CLI. Google went full mask off. This is a complete clone of Claude Code. It works the same. It [00:44] behaves the same. It looks the same. It is basically the same tool. For those who haven't been paying attention, there's a reason why they copied Claw Code so blatantly. Clawed Code is by far [00:55] the best AI coding tool ever. It's been taking the entire AI space by storm the last few weeks. But here's the minor difference though, and this is where Google went savage mode. They made the first,000 requests every day free on Gemini CLI. With Clawed Code, you can [01:12] pay up to $200 a month if you're [01:15] building any sort of serious app. This $17 tier is basically useless. You get like five requests. you really want to be using the $200 plan, which is pretty expensive for most people. So, the fact [01:26] that Gemini CLI is free for the first [01:29] thousand requests is mind-blowing. But none of that matters though if Gemini CLI stinks. So, here's the test we're ## Competition begins [01:36] [01:36] going to do. On your screen here is two instances of cursor. On the left, we're going to load up Claude Code and cursor. On the right, we're going to load up Gemini CLI and cursor. We're going to [01:46] give them the same app to build and [01:48] we're going to see how it plays out. As each AI agent finishes their task, I'm going to immediately give them another task so we can see which one works faster. We're going to build out a full complex stock tracking app. This isn't going to be your mama's to-do list or habit track or anything like that. This [02:04] is going to be a complex stock tracking [02:06] app to see how well these two perform. Then we're going to look at the results and we're going to rate each AI tool based on speed, accuracy, taste, and ease of use. We're going to see which tool earns your hard-earned money and which is the best. So, let's do this. [02:21] Let's load up both. If you haven't installed Claude Code in Gemini yet, I'll put the commands down below that you can just copy and paste in. Feel free to follow along with me here so that you have your own app at the end of this too that you can build out. So, I have them both installed already. Let's [02:35] just load them up. For Claude, all you do is type in Claude into your terminal, hit enter. That's all loaded up and good to go. And then for Gemini, it's basically the same exact thing. We type [02:43] in Gemini, I hit enter, and that's [02:46] loaded up as well. So, what I did was ## The prompt [02:47] [02:48] this. I went into Claude. I had it build me a prompt for a stock tracking tool that I can now just copy. And we're going to put it in both. Basically, what [02:57] this prompt is going to do is build out [02:59] the stock tracking tool. It's going to have a stock chart in it, the price chart. It's going to have an AI company summary. It's going to have full sentiment analysis. It's going to have a [03:07] watch list in it. We're going to take this prompt, put it into both, and we're going to get to work on each to see what kind of app they build out. So, I'm going here. We'll paste this into Claude and we'll paste this into Gemini. Uh, [03:20] now let's hit enter on both and see what [03:23] kind of results we get. So, here we go. Enter. Enter. So, 0.1 second advantage [03:28] to Claude Code because I hit enter on it [03:29] first, but let's see what happens. So they're both Claude Code is pontificating while Gemini CLI is planning the stock tracker. I think I prefer pontificating a little bit more. Claude Code is going in and updating a to-do list. So it's building out its [03:42] to-do list for us so it can do it step [03:44] by step. Gemini is just giving us the plan which it's already done. The plan is already done. Of course I will build out a sleek AI powered stock tracking application. Here is my plan. Claude [03:54] code is still going here. So it has the front end uh styling AI integration interactive search dynamic price. Okay, so it's doing that and then over here claude code just finished. It already started writing some code and it has a to-do list. So I'll build an AI powered [04:08] stock tracking app with all the [04:09] features. I'm going to let it go into YOLO mode here. So it can just keep working and I will do the same with Gemini CLI. So it looks like they're both now installing the technology behind our app. Uh, it looks like Claude [04:22] Code is now flipping, [04:25] which I think is their attempt at humor, [04:27] which I don't mind. You know, let's make it so the AIS are fun to interact with. So, I don't I don't hate on it. Little tryhard, but I don't hate on it. I think [04:34] we're going to have two pretty cool [04:35] looking apps by the end of this. It's going to be interesting to see. I'm most interested in taste. I'm most interested to see which one's going to have the better tasteful app. Which one's going [04:44] to feel better to use? Which one's going to look better to use? I'm I'm interested to see which one's better at that. Claude code now went into implementing the stock search functionality. So it already set up the [04:53] project structure in HTML foundation and [04:55] created the CSS styling. So now it's implementing stock search. Both are writing out code as we speak. Uh it looks like Claude code has gone and just done the most basic HTML app possible while over on the right Gemini went with Nex.js which I'd prefer the Nex.js so [05:13] that you can do more with it. So Gemini might be in the lead a bit here so far. While Claude Code is on straight up HTML, JavaScript CSS, which a little old school, uh, which is interesting. Okay, so it looks like it's ready to run. All [05:25] we need to do is open index.html. So let's do that. This reminds me of my college days, just spilling out JavaScript, HTML, and CSS files. All [05:35] right, let's run index.html. Let's see what we got here. Okay, this actually ## V1 [05:38] [05:39] looks pretty nice. It's pretty clean. What happens if I is this just going to work? Search. What's going to happen? [05:44] Apple 289. Is that Is this fake data or is this real? Are we looking at real data here? Let me check my phone. I [05:50] mean, we're literally, I think, a minute [05:52] and a half into this. If this is real, I'm going to be very impressed. Apple. Nope. This is fake data. Uh, Apple is [06:00] currently at 2011. Okay, so it put in some mock data. I think what it probably needs to do next is hook up the API so that it has some real data. But this is what the a the UI is going to look like. [06:10] And honestly, overview, market [06:12] sentiment, this is pretty good from a [06:15] taste perspective. This is pretty nice. Go back in and see how both of these are working. Okay, looks like it's running. [06:20] Let's boot that up and see what we got [06:21] here. So, let's compare the two so far. So, this is V1. Here we go. Uh, from a [06:26] looks perspective, I mean, it's [06:28] comparable. I think I lean a little bit more towards the Cloud Code one. From a looks perspective, let's see from a data perspective, that's accurate. 201108. It [06:36] oneshot the functionality. It oneshot the AI fun. So 2011, boom, that's the actual price of Apple at the moment. So this is accurate. That's accurate. [06:46] That's using real data. So oneshot real data. It didn't implement the AI functionality for the summary yet uh or the market sentiment. But the price perspective, it had in one shot, which is pretty amazing. From a look and feel [07:00] perspective, I think it's quite a bit [07:02] weaker than Claude Code. Claude Code won ## V2 [07:04] [07:05] nicer colors, nicer feel. This one feels pretty raw, like I'm in like a Bloomberg terminal or something. V1 going to have to give the Gemini CLI the first kind of go through here. Mostly because the functionality seems to work. It's just a [07:17] slight edge though because everything [07:19] else I'd say is weaker, but the fact [07:20] that it was able to get the price done [07:22] in V1 is pretty good. So in Claude Code, I'm going to say that seems to work. What are next steps? I'm going to assume it's pulling in real data is the next step. And then in Gemini CLI, I'm going [07:30] to say that works. The UI works, the pricing works, but there's no AI data to see what it does from there. Okay, so Claude Code shows the updated to-do list, which I love. I love that Claude code works this way. It works in kind of [07:41] a to-do fashion where it builds and [07:43] maintains its own to-do list as it goes. Uh, Gemini CLI doesn't seem to be working kind of with to-do list. It kind of just says, "Hey, this is what we're going to work on," which I think I leaned the user experience on claude code a little bit more in that regard. So immediate next steps it is going to implement the real-time pricing on cloud code give us pricing alerts have portfolio tracking which tracks buy sell transactions in and calculates P&L okay and it's going to have news integration so claude code's kind of going on here and just adding a whole bunch of functionality I didn't really ask for I don't mind it to be quite honest with you but it's something to take it's something to note Gemini CLI on the other end says now let's bring in the AI feed features. So, it's sticking exactly [08:27] to what I'm asking for. So, it's going to pull in the company summary and add the AI features. So, they're both going off. So far, Claude Code seems to have more of a mind of its own, like it's a human being, and it's adding other features it finds interesting, which I don't absolutely love, but to be honest with you, I don't hate. It's very uh [08:45] Claude 3.7 for those who live through the claude 3.7 days where it just go off and do a billion things on its own. And over on the right hand side, Gemini CLI seems to be sticking to exactly what we're doing. It's very important just as [08:57] a tip for each of these here. As you're using Claude, as you're using Gemini CLI, the higher the quality your first prompt, the better they're both going to work and the more time you'll save in the back end. Right? If you spend a lot more time in the front end building out really good prompts like we did here with this really big detailed prompt from Claude, you're going to be able to save a lot more time and you don't need to handhold the AI every step of the way. You just need to approve the next [09:23] steps, right? Like you've seen so far each of these all I'm doing is saying yes, continue, yes, continue, yes, continue. But if you have kind of a sloppy prompt up front, you're going to have to do a lot more handholding and fixing of errors. So, my little tip here as they're both building and going at it at the moment in our competition is make sure before you start building any app, run your ideas through an LLM. I find [09:44] Claude to be the most creative out of [09:46] all the LLMs and ask for a first prompt. Then you'll get a really nice first prompt like this that I showed you here. That'll save you a lot more time as your AIS are building. Okay, so both have finished what they were doing. Let's see [09:58] where we're at. Okay, it looks improved. ## V3 [09:59] [10:00] Okay, so we have now a set alert button. What does a set alert button do? We can get wow we can get alerts when we go above or below a certain price. I didn't ask for that but that's still kind of cool a thought of that. I kind of wish [10:12] it would focus on the core features [10:13] first before adding all these extra [10:15] features but that's fine. Whatever company overview okay so it's getting the real time news. It has like the company overview. Uh is a multinational technical company that has many okay so it has that's the real news. That's real [10:27] Apple news. Uh market sentiment mixed signals and price action market indicators. has a balanced wait and see approach. That's not really telling me anything, but it looks good, which counts for something. Does not look like [10:38] the price is real time yet, though. It still looks like it is mock data. Yep. So, it's switching every time I click it. So, did not figure out that we're [10:46] two steps into cloud code. Did not figure out the real data yet, but it looks really nice. Let's go over to Google CLI and see what they got over here. So, I'm going to pull open the Gemini CLI, see what we got. Okay, the [10:57] chart looks nicer now. uh about Apple. Based on the information available, Apple is a company that operates in the following sectors. It's a multi. Okay, [11:04] so it's got the real-time news. Market sentiment, uh it doesn't mean the sentiment doesn't tell me as much. Somewhat bullish. Doesn't really mean anything. And the watch list, can we add [11:13] other stocks to the watch list? Yes. Okay. So, it has the real-time stocks from the watch list, which is pretty cool. And the chart looks pretty nice as [11:21] well. I mean, out of all the things requested, really the only thing this Gemini CLI is missing is an explanation on market sentiment. It says somewhat bullish, but we asked for in our original prompt some sort of explanation. So, so far through two prompts here. Claude Code's looking [11:39] really good and they're adding a bunch [11:40] of features. They haven't really done the core of what we're looking for yet. Gemini CLI, on the other hand, is going just straight to what we asked for. It's not really adding any of the window dressings or like things that make apps fun or interesting to use, right? Like [11:55] this is as plain Jane as it gets, but [11:57] it's doing what we asked for, which is [11:58] really helpful and nice. So, so far I'm seeing this in this competition. Clawed Code seems to be building better code. Like the the app looks nicer. It looks ## V4 [12:10] [12:10] more modern. It looks more usable. It seems like, and this is pretty expected with Claude, just based on everything we've seen over the years, it's able to come up with context on its own and like it's able to do more than you ask. The number one complaint with Claude 3.7 was [12:26] that it just did too much, right? It would you give it a command then do a hundred other things. It still has elements of that in here, but I actually like it, right? Like it did extra things like make this look pretty, right? Like [12:36] make the chart look nice, use nice [12:39] looking colors. I like that it kind of has a brain and a taste of its own. The Gemini CLI side, it's doing what we're asking very well, but it's not doing it in a way where like like this is just not appealing. No one would use this app because it's very ugly and simple. [12:56] There's no taste to it. It doesn't feel like it has its own brain. It feels like it's just doing what I'm telling it to do, which don't get me wrong, it's not necessarily a negative thing. Like from a functionality perspective, it seems like a step ahead of Claude. But like so [13:11] far, I think I prefer what Claude's [13:13] building a little bit more. And over on the Claude side, let's test out the real-time pricing to see if they finally got that. So I'm going to refresh here. I'm going to type in a stock Google. [13:22] It's going to load the stock data 178. Let's see if that's the real time price. Loading up the stock chart here. 178. So [13:28] that's accurate. It has the real-time data. That's awesome. Can we add that to watch list? Yep. And it has the real [13:33] chart there. Let's see if we can get Apple to update. Yep, it has the real Apple chart. Let's see if we can add that to the watch. It's already in the [13:39] watches. All right, so it has the mock data say. Oh, nope. It updated. There we [13:42] go. Okay, so now the functionality works here. So now it's caught up to Gemini CLI. It just took one extra command to make it catch up from a functionality perspective. But again, from a usability [13:52] perspective, I like this a lot more. That looks great. What are next steps? Oh, and while I'm doing that, it says Gemini has completed phase one and phase two. So, usability should be improved [14:03] and it should have better AI analysis. Let's see what we got here. Okay, so looks like we have our first bug. We have our first We have our first bug. [14:11] Gemini messed up. It's not loading and we got an error. So, let's give this back to Gemini and see what we got here. One of the great things about Claude Code I've been seeing over the last few weeks as I've been using it every day for like 10 hours a day, it does not mess up, right? Like out of the hours [14:26] and hours and hours I've spent in clawed [14:28] code, I haven't gone down a single error [14:32] rabbit hole, which is amazing. Back in the old days, the wild west days of AI coding in like September of last year, I mean, you would go in rabbit hole after rabbit hole after rabbit hole and I'd be pounding my desk. That doesn't happen with cloud code. It doesn't get errors. [14:46] So, that's amazing. Looks like Gemini got some errors. So, let's just put that in. Got an error and I paste it in. This [14:53] is what you do when working with your [14:54] AIS, by the way. You get errors, you keep it simple. Hey, got an error. You copy and paste the error from the console, right? Right click console in [15:01] your browser and you're able to give the [15:03] error and it should be able to fix it [15:05] for you. So, as Gemini is fixing that, we are at this point Claude has done pretty much what we asked. The market sentiment analysis isn't great. Let's test the kind of taste a little bit more cloud code and let's just see what it wants to build next and kind of give it a little bit of freedom to add a couple features. Then we'll call it there. See [15:25] what it ends up with and then Gemini [15:27] will fix the errors and see where that [15:28] ends up with as well. So I asked Claude Code what it wants to add next. It wants to add portfolio tracking. So it tracks your real-time portfolio, buy and sell transactions, P&L, advanced technical indicators, and user authentication. So [15:42] let's do this. Let's take a couple of these and have it just build it out. Can you please add the technical indicators and news feed? So I took a couple of things it came up with. I said just feel [15:54] free to add that. We'll leave this as last steps of claude code. And then on the Gemini side, we will try to see what is going on here. Okay, so it says I've corrected the error. Let's pull it open. [16:04] Uh does not appear like it corrected the [16:06] error. Looks like we're still getting errors here. So we're in our first rabbit hole here. Let's do this. Let's [16:12] take this. Let's paste this error in. To Claude's credit, it did not get any errors, which is expected because I've been using Claude Code for a while and I never get errors with it. More errors. I [16:22] paste in the errors to Gemini. I say, "Go at it. Let's fix it." Claude Code's building out its other functionality that it kind of just came up with on the fly. We already met the requirements on [16:32] Claude Code. Now, it's just building out extra stuff. It's just styling on us now, while Gemini is trying to fix. While Gemini did get to the basic functionality quicker, it wasn't as pleasant of an experience. Now we're [16:44] running into errors, unfortunately. All right, Gemini says it fixed. Let's see what we got here. Let's refresh. Oh, it [16:50] almost showed something. Now we're getting more errors. Okay, uncaught reference dates is not defined. All right, we're getting a little fed up with Gemini here. Claude code just [16:58] doesn't do error rabbit holes. Funny because a lot of people say Gemini 2.5 Pro is the best coding model out there. It seems to be very highly intelligent like as we kind of saw it was able to get to the final results quick but it it doesn't seem it seems to be kind of un nonironed out right like claude code's very ironed out dependable reliable which I think is very important Gemini is not very reliable if you use Gemini and like cursor before you know it messes up with tool calls and things like that you're right my apologies I introduced a reference error at my last fix okay so this is your fullon rabbit hole here And over here, Cloud Code finished with its bonus features. It [17:37] finished with its bonus features. So, let's see. So, I'm going to refresh ## Final Version [17:40] [17:40] here. Let's put in Tesla. Shout out Tesla. I got a Cybert truck and it's awesome. Okay, so we got Tesla. Let's [17:48] see what we got here. We got technical indicators. So, it has the RSI, the MACD, the Bowlinger bands. That's pretty good. Latest news. Okay, so this is [17:55] fake. This is fake data. I'm sure this is fake data because it didn't actually implement the API yet. I'm sure I have to put in like an API key, but it has the features built out. And just like [18:06] the real- time pricing, probably all I [18:08] need to do to get that to work is put in [18:10] an API key, right? So, let's look at what Claude Code has for us so far. It has, I think you'd probably agree watching this video, the best UX, the best UI. This is nice. This is an app I [18:20] would use every day. The watch list looks great. I mean, it has little taste features like the hover state, things like that. Like, you hover over things, it looks nice. I click it, it loads it [18:29] in, right? Boom. It updates the price. Nice. It has the I I like that it has [18:33] like the color coding. The colors are nice. I like that. It has the overview, which is good. It has technical [18:40] indicators, which is cool. That's just bonus features it came up with. And the latest news and I'm sure I just put in an API key and it'll get the latest news. So, this is good. This is clearly [18:49] like Claude understands human beings and [18:51] made like a very human specific app. Was it as fast as Gemini? It took a little bit while to get to that baseline usage, right? But it eventually got there. [19:02] Gemini. Let's see what we got going on in Gemini. So, it finished. Let's see what we got here. More failed to fetch [19:07] data. More failed to fetch. Let's see what happens if I put in a stock here. Nope. Failed to fetch. So, we've run [19:14] into a full-on rabbit hole of errors [19:16] with Gemini. And to be honest, this is kind of this is kind of predictable using both models. I've used both models a ton. There was a lot of hype when Gemini 2.5 Pro came out. People were [19:28] saying, "Oh, it's the smartest ever." You know what? It is pretty damn smart. But to be honest with you, it's just kind of like unpredictable and unfocused, right? It runs into errors. [19:38] It doesn't know when to use tools, [19:40] right? It'll do things that blow you away. Like it kind of oneshot the basic functionality of the app, but the moment we started trying to change anything, it tripped over itself. So, let's let's do this. Let's score these and determine [19:53] the winner. All right. So, let's rate ## Scoring [19:55] [19:55] these. We'll rate them all one to 10 in each one of these categories. Let's do this. Claude Code speed. Claude Code was [20:02] a little slower, so it didn't execute [20:03] quite as fast, but it was a good speed. If we're going to go 1 to 10, I'll probably give it a seven. With Gemini speed, I'll probably give it an eight. Slightly faster. Accuracy. I define [20:13] accuracy as does it get it done without [20:16] many errors or bugs or messing up at [20:18] all. Claude Code from an accuracy perspective had zero bugs. It did do a little couple extra things, but to be honest with you, I don't mind it. Claude 4 clearly is a little bit more focused than 37. Like if it was 37 running this, [20:33] then we'd have a 100 features that were [20:34] nothing that we requested. So, I'm going to since it had no bugs, I'll give this a nine. The accuracy of Gemini CLI, I mean, it had all those bugs at the end. To its credit, it was able to get the core functionality done much faster, but the bugs, I mean, if you're a beginning programmer, you're new to AI vibe coding, that would have probably had you frustrated wanting to quit. So, I'm [20:55] going to have to give that a six. From a taste perspective, Claude Code had fantastic taste. It was an easy to use app. It looked beautiful. It was very [21:03] pleasant to use. Honestly, I'm going to have to give it a 10 from a taste perspective. Taste perspective, a Gemini CLI, it was ugly. Uh, it was kind of gross. I didn't feel motivated to use [21:15] it. And like when you're building apps out, taste is so important. The emotions you feel when you're using the app are so important. So, I'm going to have to give that a five. Ease of use. Just how [21:25] easy it was to use. I mean, honestly, both are pretty easy to use. If anything, Gemini CLI might have been a little easier because I had to ask Claude Code to kind of do some of the little tiny things like implementing the API keys or things like that. Gemini went out of its way and just implemented the APIs and got that done for me. [21:44] Clawed Code, it didn't really say, "Hey, [21:46] you need to put your API keys in here or [21:48] do anything like that." It kind of just assumed I would figure it out. So, from an ECU's perspective, Gemini is a little easier. I'd say probably an eight for Gemini CLI and a seven for Claude Code. [21:59] So, that brings us to the totals. Claude Code, that's a 33 total for Claude Code and a 27 for Gemini CLI. Now, I'll say this. This backs up my experiences with both. Gemini CLI is an amazing tool and [22:14] if Claude Code did not exist, Gemini CLI [22:17] would be blowing minds, right? But Claude Code just seems a little bit more humanlike, more lifelike, a little bit more intelligent. It just seems to operate like a human, right? Gemini CLI, it does pretty incredible things, but every once in a while it just goes off the rails a little bit and messes up, which was what my experience was with with Gemini 2.5 Pro. That was my [22:39] experience. So, it matches up with that. Here's what it comes down to. If you're tight on budget, Gemini CLI is pretty amazing for its cost, which is uh free. [22:48] Free for the first thousand requests. So, if you don't have money to spend, here's what I would do. I'd go out, I'd use Gemini CLI. I'd build your first app. I'd put it online. I'd start making [22:58] money. If you want tips on building apps and making money, I have a link for a video down below you can watch and how I monetize my app. That's down below. You start making money by with your app on Gemini CLI and then you take that money and then you get on the $200 a month claude code. Right? If you have extra [23:14] money, you invest in yourself and you go [23:16] straight to Claude Code. But what Google did with Gemini CLI is amazing making it free, right? The more competition, the better it is for all of us. Claude Code though does appear smarter. does appear [23:27] more focused and just has better taste [23:29] overall. The apps that come out of Claude Code just feel like better apps, more usable apps, and this does back up my personal experiences using both for hours and hours. If this was helpful, make sure to subscribe, turn on notifications. All I do is make amazing videos about AI. Leave a like as well. [23:46] Also, every Thursday I do a live stream [23:48] and I build apps live, so watch out for [23:50] that as well. See you in the next video.