Building a web-based personal knowledge management tool sounded like the perfect challenge for today’s most accomplished coding agents. It wasn’t just another landing page or main dashboard.
I wanted a fully functional app that could manage notes, tags, note previews, local storage, search, and visual links between related ideas.
I wanted to see how Claude Code, Codex, and Google Antigravity handled the project beyond just generating code. Could they plan the architecture, make sound product decisions, identify problems early, and keep the whole build running smoothly? Let’s find out.
Setting rules
Learn about demand
To keep the comparison fair, I used the most powerful models available in each tool: Gemini 3.5 in Antigravity, Opus 4.8 in Claude Code, and GPT-5.5 with High Grounding in Codex. Each agent received exactly the same request without additional prompts, follow-up instructions, or adjusted requirements.
The task was ambitious. Instead of creating a simple prototype, each agent had to create a complete web-based one personal knowledge management software.
The application needed a system of notes, organized tools, linking notes and a visual knowledge graph to show the connections between ideas. Task management and quick-grab functionality were also required to keep thoughts efficient.
The requirement also included engineering requirements covering architecture, data processing, responsiveness and reliability. The test was also mandatory.
Codex
Packed with features, but with a below-average UI
Codex it didn’t come close to meeting my expectations. At first glance, the app looked polished and the overall design gave me a good first impression. Unfortunately, that feeling disappeared as soon as I started using it.
I could create notes, but there was no proper way to open and view them in the full screen workspace. Several buttons overlapped with other interface elements, and the layout often felt crowded and poorly thought out.
The biggest problem was that the Codex was trying to do too many things. Instead of prioritizing the basic note-taking experience, he packed the interface with features without making sure the basic workflows were smooth.
Iconography also felt key. A few menus used generic icons based on the first word of the section name, which made the interface look clumsy rather than thoughtful.
I was also surprised by how well Codex ran on a Mac. While generating the code I could feel a noticeable heat under the car. Claude Code and Antigravity did not have the same effect during my testing.
Claude Code
A strong effort, but a few losses
Claude Code was up there with Google Antigravity in terms of features. It understood the brief well and delivered almost everything I would expect from a web-based PKM tool. Notes, organization, linking, tags, tasks, and the knowledge graph were all present, and the overall experience felt much smoother than what Codex produced.
However, the interface could have made better use of the available space. Some areas felt unnecessarily cramped while others felt like a very empty room.
The layout was usable, but it didn’t feel as balanced or efficient as Antigravity’s implementation.
My biggest disappointment was Quick Capture. Instead of opening a small, lightweight login window, it launched a new note in the full editor. This technically allowed me to capture the idea, but it missed the point of the feature.
Despite this flaw, Claude Code produced a solid and highly competent result. He came close, but just missed the crown.
Google Antigravity
It ticks all the right boxes and takes the crown
Google Antigravity easily the biggest surprise of this comparison. Based on my previous experience with Gemini 3.1, I was expecting a functional but fairly basic app. However, Gemini 3.5 changed the impression.
The first thing that stood out was the interface. It was aesthetically pleasing, balanced and more refined than I expected from the first attempt. Antigravity also made excellent use of icons within the app.
The record management experience was strong. I could pin important notes, star them for quick access, and browse labels without digging through multiple menus. The knowledge graph was easy to find and gave me a clear visual overview of how the various notes were connected.
My favorite feature was Quick Notes. Antigravity added a small, window-style menu that lets you grab ideas without opening the full editor.
It sounds small on paper, but it can make a big difference when I’m trying to remember an idea before I lose it. There were a few areas that could have been better. The favicon looked basic and didn’t match the quality of the main interface. The tasks section could offer more functionality and better organization.
Still, these are minor niggles compared to everything Antigravity does right. This one definitely takes the crown for me.
Three agents entered and one took possession
A skilled coding agent must understand the product, prioritize the right features, and make thoughtful design decisions. Codex provided an attractive interface, but poor UX, overlapping controls and careless execution made it the weakest result.
Claude Code fared better and came close to winning, but his inefficient use of space and a broken Quick Capture implementation held him back.
Google Antigravity is the clear winner. Gemini 3.5 understood both the technical information and the practical purpose of the application. Although it wasn’t flawless, Antigravity got the basics right and it was the only tool that behaved like a tech lead.
