Building https://xitter.gq/t.co/Tiv6hh0Ur0 · AI product notes, vibe coding tips, and generative art experiments
Joined August 2014
- Tweets3.2K
- Following765
- Followers12K
- Likes10.6K
Kimi K3 实测，一句话复刻动森风格游戏，美学和可玩性都很好，太惊艳了🤩
我是直接粘贴了 X 别人的视频推文，启用多模态分析自己完成，买的是 199/月 档位的订阅，消耗还是挺快的：
- 周消耗 5%，4小时消耗 27%
这个任务耗时情况：
- 游戏从零到可玩（含分析视频、写代码、浏览器实测）：11:10–11:55 ≈ 45 分钟
- 二刷视频 + 细节打磨：≈ 5 分钟
我把「用 AI 一个晚上复刻一款爆火小游戏」打包成了开源 Skill 🐾
对你的 Agent 说一句"复刻 Water Sort"，它会：
→ 调研游戏规则，和你确认范围
→ 先写求解器再写生成器，每关可解性是证明出来的
→ 三消的目标分由机器人构建期实跑标定
→ 自己开浏览器玩通一局才算验收
产出是成品，不是 demo。仓库里有两款完整案例，全部开源：
github.com/hellokaton/game-c…
我把「用 AI 一个晚上复刻一款爆火小游戏」打包成了开源 Skill 🐾
对你的 Agent 说一句"复刻 Water Sort"，它会：
→ 调研游戏规则，和你确认范围
→ 先写求解器再写生成器，每关可解性是证明出来的
→ 三消的目标分由机器人构建期实跑标定
→ 自己开浏览器玩通一局才算验收
产出是成品，不是 demo。仓库里有两款完整案例，全部开源：
github.com/hellokaton/game-c…
腾讯刚刚发布了 HunyuanImage 3.0-Instruct，基于混元MOE结构的新一代视觉语言旗舰大模型。
我刚刚用官方的示例测了一下，中文有点拉啊😅
Today, we introduce HunyuanImage 3.0-Instruct, a native multimodal model focusing on image-editing by integrating visual understanding with precise image synthesis! 🚀
It understands input images and reasons before generating images. Built on an 80B-parameter MoE architecture (13B activated), it natively unifies deep multimodal comprehension and high-fidelity generation.
🧠 A "Thinking" Model with Native CoT & MixGRPO: The model doesn’t just execute commands, it processes them through a Native Chain-of-Thought (CoT) schema. Enhanced by our self-developed MixGRPO algorithm, it reasons through complex instructions to achieve flawless intent alignment and human-preference consistency.
🎨 Precise Editing & Multi-Image Fusion: The model enables accurate image editing by adding, removing, or modifying elements while keeping non-target areas perfectly intact. It also excels at seamless multi-image fusion, synthesizing complex scenes by extracting and blending elements from multiple sources into a unified, consistent output.
🏆 SOTA Performance: HunyuanImage 3.0-Instruct sets a new benchmark in visual quality and alignment, delivering performance that matches leading proprietary models.
We aim to enable the community to explore new ideas with a state-of-the-art foundation model, fostering a dynamic and vibrant image generation ecosystem. 🛠️🎨
💻Try it at (PC only): hunyuan.tencent.com/chat/Hun…
感谢 @umesh_ai 分享的混合媒介极简线稿 + 真实物体的创意。
我做了几组和 AI/科技行业相关的图玩了一下，提示词就不放了。
下面帖文里详细说一下我是如何构建提示词的方法👇
Nano Banana Pro gives wings to your imagination!
Prompt : Create a minimalist mixed-media illustration that combines a simple black line drawing of a detailed human eye with eyelashes and eyebrow with a real-world object, a blue glass marble with swirling patterns, strategically placed to represent a key element in the artwork. The marble becomes the eye's vivid iris and pupil. The object should be seamlessly integrated and visually meaningful in a mesmerizing way. Use a clean, white or off-white background for contrast, and ensure the lines are clear and the object stands out naturally.
冬日搞怪九宫格人像（同脸九连拍）
Prompt:
Create a 3x3 grid of playful winter portraits featuring the person from the attached image in all nine frames. Each frame shows a different expression, hairstyle, and outfit — funny, awkward, joyful, bored, mischievous, over-the-top. The vibe is childlike, spontaneous, and imperfect, like a creative winter photoshoot done just for fun. Solid winter-toned backgrounds (cool blue, grey, off-white, muted beige). Styling varies between frames: messy beanies, scarves, ear warmers, oversized coats, fuzzy sweaters, layered winter textures, quirky glasses, exaggerated bed-hair, slightly wet hair from snow. Some frames include tongue out, wide smile, surprised face, mock-serious look, squinting from cold. Lighting is soft studio with a cool-neutral tone, visible grain, slight texture, no polish. The grid feels cohesive but intentionally chaotic and playful — winter mood without holiday symbolism. The face must remain 100% identical in all 9 frames.
对 Agent 开发感兴趣的可以看看 @miantiao 开源的 AIGC 周刊，绝对让你大有收获。
这个项目基于 claude-agent-sdk 实现的多 Agent 协作+MCP 使用，实战型选手必看，不过 cloudflare 部分代码不是核心可以不看。
你甚至可以直接用任意熟悉的 AI Coding 工具来帮你了解一下。
github.com/miantiao-me/aigc-…
如何从零用Claude agent SDK搭建一个智能体。
有160多万阅读。
作者的substack也发了，先Mark。
除了Skill，下一个很值得研究的就是Agent SDK。
nader.substack.com/p/the-com…
在 X 上每天都充斥大量的信息，实在接收不过来了，尤其 AI 相关。
可以试试 Tech Twitter：把 AI / 创业 / 产品相关领域的高价值推文、Threads 和文章筛出来，刷屏内容压缩成 5 分钟精华阅读。
想要更少噪音、更高信号可以直接订阅 Newsletter
👉 techtwitter.com
