与「PREVIEW」相关的搜索结果

おはようございます🙌 今週もViduで動画生成しました！ 元イラスト：Midjourney 映像：Vidu Q4 Preview BGM：ChatGPTで生成 #Vidu# #ViduCPP# #ViduQ4Preview# 招待コード：VIDUQ4CPP 【夕陽に輝く壮麗な宮殿都市】

Exciting Claude Duet update: every Pro, Max and Ultra subscriber in the US who was on the waitlist as of yesterday has now been invited. Thank you for your patience! If you haven’t joined yet, the waitlist is here: We also wanted to share a note on usage. To give you room to play with Claude Duet, the research preview comes with limit resets: 3 for Pro, 1 for Max and Ultra. Please check your inbox, give Claude Duet a try, and keep the feedback coming. We’re aiming to open up to more countries outside the US very soon.

A quick look around HyperHive. Five roles, one crew, built on @HyperliquidX's public API. AI writes the card, code sizes it, a sealed key trades it, and every fill is matched to a signed slip. Product preview for now. Trading perps can lose money.

Tired of walls of markdown? Try an /animatic. I'm working on two skills: - /setup-animatic, that chooses the tooling and sets it up on your machine (or a remote machine) - /animatic, that creates a video ready for upload I've been using them for a while. Here's a preview:

NAYEON SINGLE “NA-sty” SINGLE Preview (BABY ANGEL Ver. / NA-sty Ver.) 🖥️[NA-sty] Pre-save🖥️ ➮ Release 2026.11.20 FRI 2PM KST/0AM ET #TWICE# #트와이스# #NAYEON# #나연# #NA_sty# #BABY_ANGEL#

一块榴莲千层蛋糕，保质期居然长达 9 个月？？ 这两天董宇辉新品牌“禾与春时”首发的猫山王金枕榴莲千层（16 分钟卖光 15 万单）因为“9 个月保质期”吵上热搜（#董宇辉蛋糕检测报告多达60多页## 阅读破 1490 万）。有人拿它跟山姆的冷藏千层拉了张实拍配料表对比图，还用大红圈把底下一串化学名圈了出来，质疑是不是“科技与狠活”；品牌方则直接甩出了 62 页的 SGS 检测报告。 普通吃瓜群众谁有耐心去逐字核对糊图的配料表和 62 页报告？ 正好蚂蚁百灵刚发了原生多模态模型 Ling-3.2-flash（官方称已落地灵光 2.0，总参数 560B、单步激活 25B，支持 1M 原生长上下文，主打复杂图文理解与视觉反馈自检），我直接打开 Ling Studio 网页端，把这张打满红圈的网传配料表对比图和 SGS 核心指标丢给它，让它生成一张「深色系单文件 HTML 一图流鉴定卡」，顺便看看它会不会掉进红圈的节奏里。 【测试环境与配置】 • 测试入口：Ling Studio 在线网页端（左上角确认模型为 Ling-3.2-flash）。 • 工具配置：关闭联网搜索/时间/天气插件（纯靠模型原生多模态读图与推理），开启思考模式，最大推理生成长度 262144。 【实测第一轮：它从糊图里看出了什么？】 原本我以为第一版生成的 HTML 样式多少会有点错位，结果它第一把直出的深色系鉴定卡（直接内置在 Preview 预览窗里，自动分了 01~06 六个模块，右上角甚至自己写了个“🖨️ 打印/存为 PDF”按钮）排版居然一次成型，完全没崩。更让我意外的是这两点： 1️⃣ 一眼戳破“红圈添加剂”的伪命题：网友红圈重点圈出来的那个名字吓人的「羟丙基二淀粉磷酸酯」，它直接对比左右两栏配料表指出：“两边（禾与春时与山姆）都有『羟丙基二淀粉磷酸酯』→ 不是决策差异点！红圈成分是口感稳定剂（防冷冻解冻出水），不是防腐剂。” 所以这波还真不能算翻车，它本质上是用 -18℃ 冷冻链 + 杀菌工艺 + 隔氧包装做出来的速冻甜品（果肉 ≥50%，适合冷冻囤货），而山姆是 0~4℃ 冷藏即食（果肉 33%，适合大分量尽快吃完）。 2️⃣ 极强的“反幻觉”克制：原图右边山姆那一栏被网友画的大红圈挡住了净含量和保质期，而且最底下有几行压糊的重影错字。很多模型遇到这种图会自己脑补瞎编，但 Ling-3.2-flash 直接在表格里诚实标了「图中未显示」，并在表底主动写了一行小字：“右栏底部小字分辨率不足，未采信”，同时声明无联网条件下未核验 62 页报告原文，仅按摘要呈现。 【初版版面有什么不足 → 第二轮反馈怎么改？】 虽然第一版逻辑和样式都很稳，但从直接截图分享的角度看，初版 Preview 渲染画面还有两个可以打磨的细节： • 初版版面不足：一是初版把 01~06 六个模块做成了纵向长页面布局，高度超出一屏，手机端一屏截不完（见图3）；二是焦点成分的定性结论融在文字段落里，缺少一眼就能看懂的视觉标签。 • 反馈与微调：第二轮我直接针对渲染结果让它收紧版面，它用时 63 秒把原本 6 个纵向长模块重构成了三栏紧凑布局（控制在 100vh 一屏内看完），还主动把「羟丙基二淀粉磷酸酯」的国标分类（GB 2760 / INS E1442 变性淀粉）做成了显眼的红绿胶囊标签——「质构·口感稳定剂 ✓」「防腐剂 ✗」，并加上了一句金句总结：“名字长 ≠ 防腐剂，与保质期无因果关系”（见图4）。 【真实体验判断：值不值得用？边界在哪？】 • 能打的地方：据官方介绍，Ling-3.2-flash 采用 7∶1 混合注意力架构，在 DesignArena 的 Image-to-Frontend 前端生成评测拿了 1238 分（注：官方披露在 Hopper GPU、约 30K 上下文及 DSpark 投机推理加速特定条件下解码均值达 280 tokens/s）。实测下来，这个 1238 分确实没水分——直出单页 HTML 的审美和结构完成度很高，尤其是「看不清的糊字宁可标『未采信』也不瞎编」这点非常讨喜。 • 客观边界：一方面，无联网条件下模型仅对用户提供的图文与报告摘要做结构化呈现，不能替代专业检测机构或人工对 62 页原始报告的最终复核；另一方面，官方也坦诚指出，目前模型在长程跨应用任务状态维持、复杂三维空间推理及 3D 细节保真度上仍有客观差距，在线网页端也不直接内置官方演示里的 Blender 等外部软件环境。 搞了半天，所谓“9个月保质期千层蛋糕”根本不是加了防腐剂，而是把千层做成了「-18℃ 冷冻速食」。如果是你，你会买这种能冻大半年、果肉 ≥50% 的冷冻千层，还是买冷藏 3 天内必须吃完的现做千层？👇

Qwen4 is coming. Deeper knowledge. Richer world understanding. Stronger coding and agentic capabilities. Built for real-world tasks — from complex coding and knowledge work to front-end builds and professional software workflows. Qwen4-Max is now in preview on Alibaba’s Token Plan. You can also get early access to the new model on QwenWork, Qoder, Qwen APP, and QwenStudio. Put it to the test on a real problem. Show us what you build. 👀 Token Plan:

Omarchy Mobile just passed 400 stars on GitHub. ⭐ We're now just 30 days old. Here's where we're at: • 1,048 commits • 791 pull requests opened, 547 merged • 36 people have sent code • 76 forks Real Omarchy on your phone: Arch, Hyprland, a mainline kernel, no Android underneath. The first developer preview shipped this week for the Pixel 7 Pro, and it updates itself with `omarchy update`. 16 phones are in progress: Pixel 9a, 6 Pro, 7, 7a, 8 Pro, 8a, Fold, plus OnePlus and Xiaomi devices. On the 9a, calls and texts already work. Today the 9a's modem and the 6 Pro's display both landed. Today alone: 46 pull requests reviewed, 25+ merged, every shared change tested on my own 7 Pro before it went in. Next up: the installer. The core is now public, written in Rust: one codebase for a desktop app on Linux, Windows and macOS, and a web flasher. Plug in, click a few buttons, done. It's an early prototype, and there's more work coming on this soon. Want in? Grab a supported Pixel, read the contributor guide, open an issue for your phone. Thank you to everyone who has starred, forked, tested and built. We're just getting started. 🙏

Apple is working on a new app called Harbor Harbor turns your Mac into a private AI inference server. Run open-weight models locally on Apple silicon—your data never leaves your hardware, and nothing reaches the cloud. Connect multiple Mac computers over Thunderbolt 5 to pool unified memory, enabling large models and higher concurrency that a single computer can handle alone. Set up in minutes, not hours Harbor automatically detects your devices over Thunderbolt 5 to form a cluster. Plug in, start, and you’re running—no manual configuration, no command line, no infrastructure expertise required. Download and run models directly Browse, download, and load open-weight models from within the app. Harbor handles model management so you can focus on the work, not the setup. Cluster for scale Connect two or more Thunderbolt 5-capable Mac computers and Harbor pools their unified memory into a single inference resource. Run models that exceed any single computer’s capacity, or serve more concurrent requests across your cluster. Harbor is currently in a private preview under the Claris brand.

Potential 2028 hopeful Sen. Ruben Gallego previewed his new economic blueprint that would create a New Deal-style public service jobs program funded by new taxes on large AI companies, data centers and tokens.👇 Read more in Decoded: