Hey dev community 👋
I'm hudy9x, a web app developer in Hanoi, and for the past few years I've been grinding on enterprise projects — the kind where budgets are tight, licenses are forbidden, and "just use VS Code preview" is the default answer to everything.
But in February 2026, I decided to stop complaining and start building. Using Google Antigravity (the new agentic AI IDE powered by Gemini 3), I shipped three custom desktop tools in just three days that actually make my daily work better.
No massive codebases. No weeks of setup. Just prompts, a bit of polish, and publish.
Here's exactly what I built — and why this changes everything for solo devs and teams on a budget.
Quick context: My day job pain points
I work on a big web app project. Every sprint includes:
- Writing PlantUML and Mermaid diagrams for basic + detailed design docs
- Managing Docker containers (but company banned licensed Docker Desktop)
- Letting testers browse S3 buckets without giving them full AWS console access
Paid tools? Nope. VS Code extensions? They suck for real-time preview/editing. So I took matters into my own hands.
App 1: PlantUML + Mermaid Viewer/Editor (Day 1)
Why I needed it:
Our team draws sequence/component diagrams daily. VS Code's PlantUML preview is laggy, no drag/drop, no easy zoom, and editing feels clunky. No budget for paid diagramming software like Lucidchart or Draw.io desktop.
What I did:
I opened Antigravity, described the app in one prompt:
"Build a lightweight desktop app using Tauri + React that renders PlantUML and Mermaid diagrams from local .puml / .md files. Support live preview, drag-and-drop to rearrange elements, zoom in/out, pan, and inline editing of diagram text with instant re-render. Offline-first, no cloud."
The agent planned the stack, generated boilerplate, wired up rendering libs (like @mermaid-js/mermaid + plantuml-encoder), and even added nice-to-haves like theme toggle and export to PNG.
Because I already have solid experience with Monaco Editor (for the inline editing), implementing search, and integrating those main packages, the initial running version came together surprisingly fast.
It took me 12 hours to complete a functional first version on Day 1 — but that version only included:
- Monaco Editor for text editing
- Basic tab manager
- Simple PlantUML preview
The fancy stuff (Mermaid support, drag-and-drop rearranging, zoom/pan, inline editing with instant re-render, export, etc.) came later — I kept iterating and adding features over the next 30+ days.
Still, having that usable core in just 12 hours felt like magic. My PM now asks me to share the evolving version with the team 😂. Here it is https://depdok.com
App 2: Custom Docker UI (Day 2)
The problem:
Company policy bans any licensed software — including Docker Desktop. We all use Docker in WSL2 on Windows, but no nice GUI = constant terminal pain.
Prompt to Antigravity:
"Create a desktop Docker management UI that connects to a local Docker daemon (via exposed socket or TCP). Show running containers, images, volumes. Buttons to start/stop/restart, view logs, exec into shell. Handle WSL2 setup where daemon is inside WSL."
It built a Tauri frontend + Rust backend to talk to Docker API. Generated nice tables, real-time stats, dark mode...
But... it was buggy on Windows at first (connection refused because WSL port wasn't exposed). After a few hours of debugging and prompting fixes ("add port forwarding from WSL to host via netsh interface portproxy"), it worked perfectly.
Now I have my own "Docker Desktop" that's 100% allowed and super lightweight.
App 3: Simple AWS S3 Viewer (Day 3)
Super simple need:
Testers need to check uploaded files in S3 without logging into AWS console (security reasons + they don't have creds).
I just cloned a sleek file-browser UI I saw on Instagram, took a screenshot + quick description, and told Antigravity to start building the app for me with that exact look and feel.
One prompt:
"Build a desktop app that lets users input AWS credentials (access key + secret) and browse S3 buckets/files. List objects, preview images/text, download files. Securely store creds in OS keychain. No upload/delete for safety."
Antigravity handled AWS SDK integration, file listing, presigned URLs for preview/download, and even added a clean file tree view.
Polished in under an hour. Shared with QA team — instant productivity boost.
The real magic: Prompt → Polish → Publish
This whole process feels like cheating in 2026:
- Prompt — Describe what you want in natural language (be specific about stack, features, constraints)
- Polish — Review code, fix bugs, tweak UI/UX (agents help here too)
- Publish — Build binaries with Tauri, share via GitHub / itch.io / your site
No more "I don't have time to build a tool." If it takes 3 days max for a working version, why not build exactly what your workflow needs?
How the last 2 years flipped my career thinking
I used to think indie/dev tools were for "full-time makers" with endless time.
Then AI coding tools exploded: Cursor, Windsurf, now Antigravity...
Suddenly, a solo dev in Vietnam can prototype production-grade desktop apps faster than waiting for company approval or budget.
It shifted my mindset from "I'm just a web dev" to "I can build whatever tool solves my pain — and maybe others' too."
(This same energy helped me ship DepDok, my offline-first Markdown editor with Mermaid/Kanban support. If you're into lightweight notes tools, check it out!)
What about you?
What small (or big) desktop/web/tool have you been wishing for at work?
If it only took 3 days with Antigravity, would you build it?
Drop a comment — I'd love to hear your ideas. Maybe I'll prompt one next weekend 😄
Follow me on X @hudy09 for more build-in-public chaos.
Happy building!
hudy9x 🚀
Top comments (0)