Rick Ye

Building AI developer tools and games.

[Rick Ye @rickye26]

r3 v1.2 just released (https://t.co/NsipZpn1R1). r3 is like Claude Artifacts but runs locally for any agent. The latest version adds support for images in the feedback and replies.

Pasting images into the text input will upload it and leave a placeholder in the text, just like what you are used to in TUI coding agents. You can further annotate or crop it to make the intention clearer to the agent.

It's super handy for giving feedback to visual artifacts, like UI design mockup or scene rendering comparison. Try it out!

[Rick Ye @rickye26]

r3: like Claude Artifacts, for any agent. Open source. Runs fully locally.

You can:
- 🎨 Compare UI designs and pin feedback to HTML elements
- 📄 Review richly formatted design docs
- 🧪 Learn how things work via interactive demos

https://t.co/XaQSthHpwb
#AIAgents #codex

[Rick Ye @rickye26]

@Alex_ybuild 虽然没看懂规则，但我知道不停重置就对了，画面特效和动作反馈看着特别爽

[Rick Ye @rickye26]

I still love TUI coding agents for its simplicity but I can see where it comes short. I build two tools to fill the gap.

https://t.co/hFgLIi52fa as a mission control. It removes the mental burden of managing multiple terminal tabs. Remote sessions are supported too.

https://t.co/NsipZpn1R1 to host html artifacts (design doc, explainer, etc.). It lets user provide feedback pinned to the exact positions on the page.

[Rick Ye @rickye26]

@DashHuang 让我想起了以前在宿舍和室友刷刷刷的时光

[Rick Ye @rickye26]

@karpathy I've been using html design doc and explainer for a while. The efficiency can be felt instantly.

I also made a local tool (https://t.co/XaQSthHpwb) for commenting directly on the html elements. It feels so smooth to communicate with LLM in this way!

https://t.co/U999UAL5oh

[Rick Ye @rickye26]

r3 v1.2 just released (https://t.co/NsipZpn1R1). r3 is like Claude Artifacts but runs locally for any agent. The latest version adds support for images in the feedback and replies.

Pasting images into the text input will upload it and leave a placeholder in the text, just like what you are used to in TUI coding agents. You can further annotate or crop it to make the intention clearer to the agent.

It's super handy for giving feedback to visual artifacts, like UI design mockup or scene rendering comparison. Try it out!

[Rick Ye @rickye26]

My favorite part of building r3: shipping the full stack web app and cli as one executable, bundled by Bun.

It uses `https://t.co/2fk4GC2AS8()` with the `compile` option to bundle the React UI and its assets into the executable, together with the server, CLI, SQLite and Bun runtime. `Bun.serve()` serves the UI and HTTP API.

The end result? Users can simply download the executable from https://t.co/rsoY6r5bre and hand it over to the agent and start a claude artifacts equivalent that runs fully locally.

The only downside is the executable size, which is 80 MB for MacOS and 110 MB for Linux.

Thanks @bunjavascript for making this way of shipping local web apps possible!

[Rick Ye @rickye26]

r3 v1.2 just released (https://t.co/NsipZpn1R1). r3 is like Claude Artifacts but runs locally for any agent. The latest version adds support for images in the feedback and replies.

Pasting images into the text input will upload it and leave a placeholder in the text, just like what you are used to in TUI coding agents. You can further annotate or crop it to make the intention clearer to the agent.

It's super handy for giving feedback to visual artifacts, like UI design mockup or scene rendering comparison. Try it out!

[Rick Ye @rickye26]

@_FORAB 这种技术用来做恶简直是有无限可能性。同样的技术如果���制只能生成二次元人物的话好像能规避掉所有的道德和法律风险？

[Rick Ye @rickye26]

@trq212 This looks cool! Although there are many cool new game demos made with opus 5.5 in one or few shots, tuning and polishing the feel of game mechanism are still something human need to play test and give feedback to AI. And that's the fun part of making games.

[Rick Ye @rickye26]

@999toba Looks like a distilled video gen that can get quality output with only few steps of denoising, paired with some duplex A2A model. Something like https://t.co/uwDuVPWBHS

[Ailing Zeng @AilingZeng81332]

Introducing LPM 1.0 — a video-based character performance model that speaks, sings, listens, reacts, and emotes in real time.

- Generating full-duplex conversation, identity-consistent infinite-length generation, and nuanced human-like performance.
- Building across a co-designed data pipeline, Base model, Online model, and streaming inference optimization.
- Key advantages over other video generation models: performance quality, emotional conversation, precise lip-sync, identity preservation, and lifelike naturalness.

Turning an image into a performance video, LPM 1.0 serves as a visual engine for conversational agents, live streaming characters, and game NPCs.

Page: https://t.co/Ve2c2YNuqj
Arxiv: https://t.co/cM54T3KSPs

[Rick Ye @rickye26]

每次建 tailnet 都会想抽个好名字，花了很多时间不停抽卡，最后学会了很多动物的英文单词 🤣

[Rick Ye @rickye26]

过去几个月我的订阅在 openai 和 anthropic 之间反复横跳，下个月不会要开始订阅 gemini 了吧。不过现在的模型，跑分和实际使用体验差异有点大

[Google @Google]

Today we’re introducing Gemini 4 Argon.

It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit.

[Rick Ye @rickye26]

@xiaohu 现在最新的独立 TTS 模型感情表达上我觉得差不多了，就是单独生成音轨的话同步问题没法解决。之前考虑过一个给番剧做中文配音的想法，就是卡在这个问题上没去做

[Rick Ye @rickye26]

@ZacharyZhang 很棒！我正好在做一个类似 live2d 的模型编辑器，看看 desktop app 能不能直接用上这个

[Rick Ye @rickye26]

真的有苹果发布会的感觉

[Department of State @StateDept]

SECRETARY RUBIO on https://t.co/pt9rQXC44r: Starting next year, Americans will be able to apply for their first passport entirely online.

[Rick Ye @rickye26]

What happens when your gaming buddy becomes an in-game purchase? Tencent is testing an AI companion that watches your game play and talks with you as you play. Based on a screenshot, it uses usage-based pricing.

Tencent is no stranger to the freemium model. But on AI-heavy product, they are still not comfortable enough to do a freemium pricing. Charging per-minute price for a companion will certainly affect the user engagement. Not sure how this conflict can be resolved in the short term, assuming we won't have a significant drop on inference cost.

https://t.co/HZcRuETOkt

[Rick Ye @rickye26]

If a game lets you talk with an anime character the way you would on a video call, what would you want to play? I am thinking of a text adventure game where you experience the story by talking freely with different characters.

#AIVTuber #AIAvatar

[Rick Ye @rickye26]

前几天家里停电，恢复了以后 zigbee gateway 就用不了了，试了重启和重新插拔都没能恢复，导致两个房间晚上都开不了灯。

今天终于有空了，在 codex 的帮助下重新刷了固件才恢复正常。

系统智能化了以后，出问题的概率也高不少。要是没有 AI 自己修的话还会很花时间。

[Rick Ye @rickye26]

@0xInk_ The art style looks beautiful! I used the same self-play simulation in my text adventure game too. As long as the game is turn-based, it's quite trivial to let AI self play and validate ideas.

[Rick Ye @rickye26]

@liangzhu_AI 为自己工作是会上瘾的