Today, we’re launching OpenUI Cloud. Generative UI in production needs more than rendering. Validation, error correction, fallbacks, versioning, audit trails - these are the parts teams shouldn’t have to rebuild from scratch. OpenUI handles rendering. OpenUI Cloud handles the production layer. Try it: npx @openuidev/cli@latest create

Trending on HF Diffusion models 😎

Agent observability is built for the wrong person. Traces, tool calls, evals - they prove your agent did exactly what you told it to do. Your users don’t care about any of that. Your evals can pass while your product still fails. OpenUI Observability changes the signal you look at. It measures what your agent actually did for the user: what they asked for, how well the agent responded, where they struggled, and where they gave up. We’re opening an early preview to a small group of design partners. If you run an agent in production and want early access, reply or DM us.

🚀🚀🚀

Agent Observability is built for the wrong person. Traces, tool calls, evals - proof that your agent did exactly what you told it to. Your user doesn't care about any of that. Your evals passed but your product failed OpenUI Observability changes that. It measures what your agent did for the user: what they asked, how well your agent responded and where they gave up Opening an early preview to a small group of design partners. If you run an agent in production, reply or DM

Traces are a terrible product signal. Your agent can execute exactly as expected. Your evals can pass. And users can still struggle, abandon the session, and never come back. Traces tell you how an agent ran. They tell you very little about whether the user actually got what they needed. We built OpenUI Observability to close that gap. With it, you can: → Replay sessions to understand the user experience → Find and triage the sessions that need attention → Discover user insights, emerging needs, and capability gaps → Annotate sessions with feedback → Close the eval loop by turning real user sessions into better evals Because “the agent ran successfully” and “the user achieved their goal” are two very different things. You need to see both. For the next month, we’re opening access to a small group of companies before the broader release. Only a few early access spots. Join the waitlist in the replies ↓

@angular devs, meet OpenUI from @thesysdev . The Angular Lang package I built for OpenUI is now live! You can bring OpenUI’s generative UI into your Angular apps, natively. Thanks to the #OpenUI team for building this together. Where would you use generative UI in your Angular app? github.com/thesysdev/openui/

AppLess is now powered by OUI-1! A phone with no apps. And now the full stack is open sourced with - OpenUI for Generative UI - OUI-1 powered inference. Open weights. Self host with vLLM - AppLess for a consumer interface AI should feel magical!

I love the smaller/task-focused Granite variants from @IBMResearch, OpenUI OUI-1 from @thesysdev, Jev from @typesafeai, the small models from @desertantlabs, etc.

Your AI assistant shouldn’t be stuck replying in plain text. OpenUI is a full-stack, renderer-agnostic Generative UI framework for builders who want models to generate structured interfaces. It helps you turn a component library into live UI by generating model instructions from the components you allow, then progressively rendering OpenUI Lang as the model streams output. Key features: • Component-driven output – define and register the components a model can generate • Streaming renderer – parse and render model output progressively in React as tokens arrive • Built-in UI libraries – start with charts, forms, tables, layouts, and more, then extend them • React support – use official React packages for rendering, headless chat state, or prebuilt chat interfaces • CLI scaffolding – create a starter app with streaming, built-in UI, and OpenUI Lang support It’s open-source (MIT license). Link in the reply 👇

⚡ OUI-1 is here — the first diffusion model built for generative UI 🎨 and it's completely rethinking how AI builds interfaces 🔹Generates full UI screens in ~1 second, not token-by-token 🔹Diffusion architecture: refines the whole screen from noise instead of streaming word by word 🔹Built on Google's DiffusionGemma, only 4B active params 🔹Runs fully local — tested it live on an H100 with vLLM 🔹Scores 71.7% on the Generative UI Benchmark (per model card) 🔹Native tool-calling — pulls real data before building the screen Watch the full hands-on video below 👇 piped.video/4ldVbgTpw_8?si=4vc8…

Amazing to see the community publish these quants so quickly 🙌 dicksondickson: huggingface.co/dicksondickso… Abiray: huggingface.co/Abiray/OUI-1-… The barrier to entry to Generative UI just got lower!

别再只卷组件库了。Generative UI 把瓶颈从「会不会写」换成「生成快不快、体验稳不稳」。OpenUI 这类协议在干的事：给 AI 动态出界面一套可落地的底座。下一波前端入口，大概在这儿。

前端开发可能真的要变天了，但不是因为 Copilot，而是因为 Generative UI。 刚才看完 OpenUI 刚发的 OUI-1，认知被刷新了： 过去大家总觉得让 AI 生成动态界面是天方夜谭——自回归大模型逐字吐 token 太慢，等界面画出来用户早跑了；而且 AI 写 UI 极容易“幻觉”，少个括号或漏挂一个组件整个页面就白屏。 OUI-1 直接换了底层打法： 不用自回归，改用 Diffusion 架构：一次并发写 256 个 token 块，速度直接飙到秒级。 拿编译器做奖励函数搞自蒸馏：让解析器去当“考官”，专门抓错和自动修复，把准确率从 13% 暴力拉到了 71.7%。 4B 激活参数：直接塞进单张 RTX 5090 本地跑，完全开源。 以后很可能根本没有固定的“App 界面”了。

🔸 مخزن OpenUI: ایجنت به‌جای متن، خودِ صفحه رو می‌سازه تا حالا مدل AI بیشتر جواب متنی می‌داد. این پروژه کمک می‌کنه مدل مستقیم UI بسازه؛ یعنی دکمه، کارت، فرم و چارت، همون لحظه روی صفحه ظاهر بشن. اسم این کار Generative UI هست و OpenUI یه استاندارد باز برای همینه. توی کار روزمره اینطوری به درد می‌خوره: تو می‌گی چه کامپوننت‌هایی مجازن، مدل فقط از همون‌ها استفاده می‌کنه، و خروجی‌ش هم‌زمان که می‌آد روی صفحه render می‌شه. برای چت ایجنت، نسخه‌ی آماده‌ی React داره. اگه با Cursor یا Claude Code کار می‌کنی، skill هم داره که راه‌اندازی رو ساده‌تر کنه. نظر شخصی: این ابزار طراحی توی Figma نیست. برای وقتیه که می‌خوای ایجنت واقعاً رابط کاربری بسازه، نه فقط توضیح بده. اگه داری یه chat هوشمند با خروجی بصری می‌سازی، این پروژه کاربرد داره. لینک GitHub: github.com/thesysdev/openui

AIの返答、「文章だけ」で終わってませんか？👀 OpenUI、⭐️約9千。AIの出力をグラフやフォームに変えるOSS。 何ができるかというと👇 ・チャート・表・フォームを生成 ・AIの返答に合わせてUIを順次表示 ・使わせたい部品からプロンプトを自動生成 ・専用言語でJSON比最大67%のトークン削減をうたう 「読むだけのチャット」から「操作できるAI」へ。 AIアプリの体験を一段上げたい開発者に。 github.com/thesysdev/openui

OUI-1 is the first diffusion model built for generative UI. Fine-tuned from Google DiffusionGemma 26B-A4B-it, it generates full UI screens in openui-lang in about one second instead of token-by-token. huggingface.co/thesysdev/OUI…

OUI-1: world's first model for Generative UI openui.com/blog/oui-1 (news.ycombinator.com/item?id…)

OpenUI: framework de UI generativa com linguagem compacta e streaming-first. Consume até 67% menos tokens que JSON para estruturar componentes. Menos tokens = interfaces generativas mais baratas e rápidas.