Parth, a seasoned tech writer, wields the keyboard (or pen) with finesse to unravel the intricacies of both Windows and Mac operating systems. He has covered evergreen content on mobile devices and computers for multiple publications over the last six years. You can find his work on AndroidPolice, GuidingTech and TechWiser. Whether it’s demystifying system updates, deciphering error codes, or exploring hidden features, Parth’s prose guides readers through the binary maze. When not immersed in tech jargon, you’ll find him sipping chai, pondering the next software review, and occasionally indulging in a friendly debate about mechanical keyboards.
AI coding tools have reached a point where building a functional web application no longer requires weeks of development or years of programming experience. However, to find out which AI coding tool could handle a real challenge, I put Claude Code, OpenAI Codex, and Google Antigravity through the same demanding test.
I asked all three to build a personal finance management app, complete with interactive charts, expense tracking, budget management, and multiple functional pages. Let’s see who executed like an experienced developer, handled the finer details, and delivered something I would actually want to use every day.
A word on the prompt
Let’s set the rules for fair competition
For this experiment, I wanted to build something I could actually use to manage my personal finances, rather than another throwaway project. I settled on a comprehensive finance dashboard with expense tracking, budget management, interactive charts, savings goals, and detailed spending insights.
I prepared a single detailed prompt outlining the design, functionality, and features I expected from a finished app. To keep things fair, I decided to stick with the default offerings in three popular AI coding agents: Opus 5.5 in Claude Code, Gemini 3.8 in Google Antigravity, and GPT-6 Astra in OpenAI Codex.
I entered the exact same prompt without additional instructions. My goal was simple: find out which tool could deliver a polished, functional application with the least manual intervention.
OpenAI Codex
It followed my instructions but lacked polish
I started my experiment with OpenAI Codex, powered by GPT-6 Astra, and my first impressions were largely positive. I liked how Codex used a subtle green background throughout the application. It’s a small design choice, but it fits perfectly with the financial theme and gives the dashboard a distinct personality.
Codex followed my instructions closely and delivered all the major sections I had requested. However, once I started exploring the interface, some familiar shortcomings became obvious. I have been complaining about OpenAI’s handling of typography since GPT-5 models, and unfortunately, GPT-6 is no different.
Most of the text throughout the application was unnecessarily small, which affects readability and makes the interface feel less refined than it should. The transactions page was another disappointment. While the essential elements were present, Codex relied on generic icons that lacked personality.
The overall layout felt functional but unfinished, almost as though a developer had implemented the core features without giving the interface a final design review.
The Analytics page didn’t fare much better. Some elements were oversized, while others appeared small. Overall, Codex understood the assignment and delivered a respectable foundation. However, I couldn’t shake the feeling that the app needed another round of refinement.
Google Antigravity
It delivered a much more robust setup
Moving on to Google Antigravity, powered by Gemini 3.8, the difference was noticeable. It was a huge improvement over Codex, especially in terms of UI design, typography, and overall presentation. The text was appropriately sized, the spacing felt natural, and the entire dashboard looked much closer to a professional financial application.
I liked the dedicated greeting box at the top. However, Antigravity wasn’t without its quirks. For some reason, it placed two Add Transaction buttons on the main dashboard. I couldn’t understand the need for this duplication.
The transactions page also could have used more attention. Like Codex, Antigravity went with rather bland black-and-white icons that didn’t complement the otherwise gradient-heavy, attractive interface.
The Savings Goals page was beautifully designed, with tasteful gradients, attractive progress indicators, and a layout that made tracking financial milestones feel engaging. The Accounts page followed a similar design philosophy.
Overall, Antigravity did a solid job and delivered a more refined experience than Codex. I am already looking forward to trying the latest Gemini 4 Argon model in Antigravity.
Claude Code
Handled the task like a professional developer
Finally, I moved on to Claude Code, and it didn’t take long to realize that I had a clear winner. Unlike Codex and Antigravity, Claude Code managed to avoid practically every UI and UX issue I had faced with the other two tools. The typography was spot-on, the spacing felt natural, and every section looked like it had been carefully reviewed by an experienced designer.
Instead of limiting graphs to the main dashboard and Analytics page, it thoughtfully added data visualizations throughout the application. The Budgets and Accounts sections featured beautiful gradient themes that added personality without making the interface look flashy.
I also appreciated the colorful category icons in the Transactions menu, which were a refreshing change from the bland, generic icons used by GPT-6 and Gemini 3.8.
Even the animations felt smooth. These little touches become especially noticeable when you are dealing with dozens of transaction entries every day. The use of space, typography, icons, charts, and visual hierarchy was picture-perfect.
If I had to nitpick, I would have preferred a slightly greener background, similar to what Codex implemented. Claude Code didn’t just follow my instructions. It understood the kind of experience I was trying to create and delivered an application that felt ready for everyday use.
One contender was in another league
After putting Claude Code, Codex, and Antigravity through the same challenge, one thing became clear: generating a functional application is only half the battle. The real difference lies in the finer details that make an app feel polished and professional.
While Codex followed my instructions and Antigravity was a major improvement, Claude Code was in a league of its own. It delivered the most convincing experience, and for my next project, I know exactly which coding agent I will turn to first.
Claude is a cloud AI that rivals ChatGPT and Gemini.