🚨 Big Announcement 🚨
We’ve officially completed our rebranding! MGX is now Atoms (@atoms_dev) 🎉
But that’s not all
for the latest product details, case studies, AMA sessions, and event updates, head over to @atoms_dev!
👇 Follow us now to stay in the loop!
Introducing Atoms: the first AI team that builds real businesses.
From research to build, launch, and scale, all autonomous.
Don’t Vibe Code. Vibe Business.
→ https://t.co/g0L4NRU7xE
Your first Atoms project starts here.
Pick a template prompt. Follow the guided tour. Publish your site, connect
your domain, and set up Google Analytics.
Watch the walkthrough, including how to earn free credits for your next build
👇
https://t.co/xWWYYIz20V
• Your first Atoms project starts here.
Templates → guided tour → publish → custom domain + Google Analytics.
Watch the demo and learn how to earn free credits 👇
#Atoms #AIAgents #BuildWithAI #VibeCoding #NoCode #AIBuilder #WebDev
#BuildInPublic #GPT6 #Astra
What would you actually do with GPT-6 Astra?
Skip the hype. Watch the demos: 3D worlds, browser games, and real workflows—all collected in one place.
Find an idea worth building.
Make every idea count.
https://t.co/I9E3I5d73L
#AIAgents#BuildWithAI#Atoms#GPT6#Astra
Make every idea count. A portfolio you can play. A world you can explore. A game you wish existed. Explore our GPT-6 Astra case library. Find inspiration, sources, and starter prompts. Make something yours. https://t.co/A8AlsSro8N #Atoms#GPT6#CreativeCoding#GameDev#Astra
https://t.co/q4VAqzEdHn
Grateful to be featured by MIT Technology Review @techreview 🙏
Open source is driving the frontier forward. DeepSeek, Kimi, and Qwen prove it.
We’re proud that @MetaGPT_ (now @atoms_dev) is part of that movement, building AI Agents as core infrastructure in the open ecosystem.
Complexity breaks models: performance drops from single charts → multi-view → dashboards. If your product generates dashboards, you need evaluation that understands composition and cross-view semantics, not just “nice colors”.
Paper: https://t.co/ZtgnXftnTw
Code & benchmark: https://t.co/p3nx2vIgUM
Special thanks to @xie_yupeng87927, @didiforx, @EvanWu50020, @ZhaoyangYu22356, @BangL93, and @AlexanderWu0 for collaboration and feedback.
Let’s discuss.
5/5
LLMs can generate charts. Can they judge them?
We introduce VisJudgeBench: 3,090 expert-scored real-world visualizations across 32 chart types (single, multi-view, dashboards).
VisJudge improves to MAE 0.421 (−23.9%), corr 0.687 (+60.5%) vs GPT-5.
1/5
A key failure mode: score inflation.
Many MLLMs rate almost everything as “pretty good”, shifting right compared to human experts (μ=3.13).
VisJudge is calibrated: its distribution aligns closely with humans (μ=3.11), improving discrimination on bad charts.
4/5
💡From launch to leadership.
MGX is the ranked #1 product of week on @ProductHunt .
Global creators are building, testing, and launching their first vibe coding products.
The ecosystem is alive.
💡 From framework to platform.
On February 19, 2025,
MGX (MetaGPT X) officially launched as the world’s first multi-agent development team.
A complete environment where AI agents collaborate, build, and ship products.
The next chapter of product creation has begun 🔥
💡A milestone at the intersection of science and engineering.
MetaGPT was selected for an Oral Presentation at ICLR 2024,
ranked #1 among LLM-based agent frameworks.
From code to conference, AI collaboration became a science.
The paper has now been cited 1,783 times.
🌊From an experiment to a global movement.
By August 2023,
MetaGPT had topped GitHub’s Monthly Trending list 17 times.
More than 50,000 developers joined along the way.
Open-source intelligence had found its spark.
Today, MetaGPT stands at 62.6k stars 🙌
Are you one of them?
It feels like a platform built to protect leadership credibility directly.
🧐How does it work?
It evaluates risk at the source, starting with emails, meetings, and everyday communication, before issues escalate.