DEV Community: Hongming Wang

We built it during the NVIDIA DGX Spark Full-Stack AI Hackathon — and it ended up winning 1st place overall 🏆

Hongming Wang — Tue, 21 Apr 2026 08:49:52 +0000

Agent tooling is getting powerful.

But something interesting happens when you start running many agents at the same time:

the bottleneck shifts from intelligence → coordination.

Over the past year, I’ve been working heavily with systems like OpenClaw, Hermes Agent, and Claude Code to automate different parts of my workflow.

As the number of agents increased, the real problem became clear:

managing agents started to feel like managing windows.

So we asked a different question:

What if agents weren’t just tools?

What if they were team members inside an actual organizational structure?

That idea became Starfire.

We built it during the NVIDIA DGX Spark Full-Stack AI Hackathon — and it ended up winning 1st place overall 🏆

After the hackathon, we decided not to leave it as a demo.

We continued building it as a real system.

Today, Starfire has evolved into:

Molecules AI
https://www.moleculesai.app/

The core idea is simple:

the future isn’t single agents.

the future is agent organizations.

Inside Molecules AI:

workspaces represent roles
agents collaborate across runtime boundaries
hierarchies replace flat workflows
coordination becomes a first-class primitive

Instead of building another chatbot interface,

we’re building something closer to:

an AI Team Operating System

Interestingly, our internal AI team is already using Molecules AI to help develop Molecules AI itself.

If you’re working on multi-agent systems, orchestration layers, or agent infrastructure — I’d love to connect and exchange ideas.

Built My Portfolio with Google's AI Code Agent & Cloud Run - What Took Me Days Now Takes an Hour ⚡

Hongming Wang — Sat, 10 Jan 2026 20:36:19 +0000

Check it out live: https://www.wanghongming.xyz/

The Stack
For this build, I went all-in on Google's ecosystem:

Google's AI Code Agent - For rapid development and best practices
Google Cloud Run - Serverless deployment that just works
Three.js - Interactive 3D particle effects
AI Chatbot Integration - Built-in conversational experience

The Mind-Blowing Part
Here's what really got me thinking...
2 years ago, when I wanted to create this kind of interactive 3D scene with particle effects and an AI chatbot, it would take me an entire day minimum. Hours of debugging Three.js configurations, wrestling with particle systems, figuring out the math for smooth animations, and don't even get me started on integrating a chatbot.
Today? I built the same thing in under an hour. With better code practices. More maintainable architecture. Smoother animations.
The version from 2 years ago would be shocked at what's possible now.

What I Love About Tech
This is exactly why I love this industry - it never stops evolving.
Every day there's something new to learn. Every month the tools get better. What seemed impossible becomes trivial. What was complex becomes a one-liner.
Some people find this exhausting. I find it exhilarating.
The fact that AI coding assistants can now help you scaffold a production-ready 3D portfolio with best practices baked in? That's not replacing developers - that's supercharging us.

Features

🎨 Interactive 3D particle hero - Responsive particle system that creates depth
🤖 AI-powered chatbot - Ask it anything about my experience
🌙 Dark/Light mode - Because we're civilized
📱 Fully responsive - Looks great on everything
⚡ Blazing fast - Thanks to Cloud Run's auto-scaling

Final Thoughts
If you're on the fence about using AI tools for development - just try it. Not to replace your skills, but to amplify them.
The best developers I know aren't the ones who memorize syntax. They're the ones who adapt, learn, and leverage whatever tools make them more effective.
Two years from now, we'll look back at today's tools the same way I look back at 2023. And that's exciting.

What's the biggest "wow" moment you've had with AI-assisted development? Drop a comment below! 👇

Building Plot Engine with Kiro: How AI-Powered IDEs Transform Development

Hongming Wang — Wed, 03 Dec 2025 03:16:09 +0000

When we started developing Plot Engine, a cross-platform tool for writers to manage stories, timelines, and AI-assisted editing, we faced a challenge: how to move fast without sacrificing quality or structure. Enter Kiro, an AI-powered IDE that completely transformed our workflow.

Why Kiro?

Kiro isn’t just an autocomplete tool — it’s a full AI-driven development environment. With features like spec-driven development, vibe coding, agent hooks, steering docs, and MCP integrations, Kiro lets developers prototype, iterate, and maintain complex projects with ease.

We chose Kiro because Plot Engine involves multiple layers — a Flutter frontend, AI-driven analysis backend, and a rich data model. Kiro allowed us to rapidly generate clean, maintainable code while keeping architecture, documentation, and testing standards intact.

How We Used Kiro
Spec-Driven Development

For major features like our enhanced text panel with rich styling, we wrote detailed specs describing functionality, behavior, and UI/UX. Kiro transformed these specs into working code, complete with architecture, event handling, and styling rules. Compared to traditional coding or ad hoc iterations, this approach saved us time, improved consistency, and reduced bugs.

Vibe Coding

For smaller, experimental tweaks — like quick UI adjustments or prototype features — we used Kiro’s vibe coding mode. We’d describe the feature in natural language, get a draft, and iteratively refine it. The most impressive result was the real-time AI suggestion panel, which Kiro generated with clean Flutter code, full state management, and backend integration. What normally would take days was done in hours.

Agent Hooks & Automation

We automated repetitive tasks using agent hooks: generating unit tests, updating documentation, formatting code, and syncing modules whenever key files changed. This kept our codebase consistent, reduced errors, and let us focus on core features instead of boilerplate.

Steering Docs

By maintaining detailed steering docs with architecture context, coding conventions, and AI integration rules, we ensured Kiro generated code that was cohesive across all layers. This centralized knowledge allowed consistent outputs and faster integration.

Results

Thanks to Kiro, we built Plot Engine faster, with fewer errors, and maintained a clean, maintainable codebase. We could switch seamlessly between rapid prototyping and structured spec-driven development depending on the task — giving us both speed and quality.

Kiro isn’t just an AI helper; it’s a partner in coding. For anyone building complex apps, it fundamentally changes how development can happen in 2025 and beyond.