Anthropic 上线新站点 https://t.co/UazfKYPKLO,面向用 Claude 做开发的人。站点会发布工程深度文章、Claude Code 和 API 使用指南、Claude 团队成员的经验分享,另外还藏了一些彩蛋。
原来的文档网站仍在 https://t.co/t4KD4eylrq,首页导航里的 DOCS 链接会直接跳转过去。https://t.co/UazfKYPKLO 更接近 Anthropic 工程团队的博客,文章分为智能体(Agents)、工程、实战手册(Playbooks)、技能(Skills)四类。技能指的是给 Claude 预先写好的一套操作说明,让它按固定流程完成某类任务。
首页现有的文章大多围绕 Claude 5 代模型的实际用法。例如《Building with Claude Sonnet 5.5》讲怎么用 Sonnet 5.5 开发,《What a task costs on Opus 5.5》算的是在 Opus 5.5 上跑一个任务要花多少钱,《The new rules of context engineering for Claude 5 generation models》讲怎样给新一代模型组织输入内容。也有团队自己的工程复盘,比如《How we made https://t.co/vefasn3rYt 3x faster in two weeks》,记录他们两周内把 https://t.co/vefasn3rYt 提速 3 倍的过程。
最新一篇是 9 月 28 日发布的《用 Claude 自动设计评测并逐步调优》。
对正在用 Claude API 或 Claude Code 的开发者,这些文章能直接看到 Anthropic 内部团队怎么用模型、怎么估算成本。
站点还收录了视频,包括 Claude Code 团队怎么用 Claude Code,以及 Ramp、DoorDash 等公司让员工用 Claude Code 和智能体工作的案例,完整视频放在 YouTube 上。
彩蛋之一是终端模式。点导航里的 TERMINAL,整个站点会变成命令行界面,文章列表像终端里的滚动输出,输入 /help 能看到可用命令。站点的主要入口都配了键盘快捷键,按 H 回首页,按 D 打开文档,按 T 进入终端。
Steam Next Fest Demo is coming! 🎮
#ArcheAgeChronicles is joining Steam Next Fest with a free playable demo! No registration required.
📅 Oct 19, 10:00 AM – Oct 26, 10:00 AM (PDT)
Learn more about the demo and how to participate 👉 https://t.co/ormRhT88iN
#SteamNextFest#SNF
President Trump stands alongside Elon Musk, Mark Zuckerberg, Jensen Huang, Speaker Mike Johnson, and other tech leaders at the White House as they meet with the press for a gaggle.
“I think it is worth highlighting the positive benefits of A.I. ... S.I., pardon me."
"Thank you.”
@elonmusk catches his own slip after calling the technology “A.I.” while speaking alongside President Trump and other tech leaders, quickly correcting himself to “super intelligence" after today's agreed-upon name change.
I know I'm a bit late to the STORM trend, but it's better late than never!
Created using Seedance 2.5
Prompt:
Use the uploaded reference VIDEO as the exact motion, choreography, camera, framing, timing, pacing, shot duration and cut-point reference.
Use <image_1> as the STRICT identity reference for the MAIN WOMAN.
The final video should feel like the SAME original video, but with the original central man completely replaced by the woman from <image_1>.
Do NOT recreate or preserve the original man's face or identity.
IDENTITY <image_1>
The central character is the exact same woman shown in <image_1>.
Preserve her:
- exact face
- facial structure
- jawline
- cheekbones
- nose
- eyes
- eyebrows
- lips
- ears
- skin tone
- natural facial proportions
- hairstyle
- hair length, volume and texture
- any facial features visible in the reference
- natural body proportions
Maximum identity preservation throughout the entire video.
She must remain recognizably the same person from the uploaded photograph in every shot and every angle.
NO face morphing.
NO identity drift.
NO changing facial structure.
NO different hairstyle.
NO unwanted hair changes.
NO beautification or artificial facial retouching.
NO age changes.
NO facial distortion.
CHARACTER REPLACEMENT
Replace ONLY the original central man with the woman from <image_1>.
Keep the exact position, scale, body movement, gestures, walking path, head movements, smoking actions and timing from the reference video.
Everything the original central character does in the reference video should be performed by the <image_1> woman at the same moment and with the same choreography.
Do not reinterpret or redesign the choreography simply because the replacement character is female.
WARDROBE
Dress <image_1> in the same character-reference outfit, adapted naturally for the woman:
Acid lime green single-breasted blazer, notch lapels, two buttons, buttoned closed, slightly oversized shoulders.
Plain white cotton shirt underneath.
Narrow black knitted tie, neatly knotted.
Straight black trousers.
Plain black leather ankle boots, low heel.
No jewellery.
No watch.
No belt.
No bag.
No phone.
No logos or patterns.
The lime blazer must remain exactly throughout the entire video.
IMPORTANT:
Do not let the wardrobe replacement change the original choreography, body positioning, composition or pacing.
SMOKING ACTION
Preserve the smoking sequence from the reference video exactly.
At the beginning, the main woman holds a cigarette naturally between her fingers and brings it to her mouth.
She lights the cigarette using the same hand movements, timing and body positioning as the reference video.
She takes a drag and naturally exhales smoke.
Preserve the original:
- hand position
- cigarette position
- lighter movement
- timing
- head movement
- facial expression
- smoke timing
- arm movement
The smoking action should look physically natural and cinematic.
Do not exaggerate the smoke.
Do not create excessive smoke.
No smoking action should occur from any other character.
All other characters remain unchanged.
EXACT ORIGINAL ACTION / TIMING
0–4 SECONDS:
Match the reference video exactly.
The central woman is shown smoking.
She brings the cigarette toward her mouth, lights it, takes a drag and exhales.
Match the original close framing, facial positioning, hand movements and timing.
4–11 SECONDS:
Match the reference video exactly.
The woman turns away from camera and moves forward into the crowd using the exact same movement and timing as the original central character.
Preserve her body position, walking speed, direction and camera framing.
11–17 SECONDS:
Match the original choreography exactly.
The woman moves back toward the central position and performs the same gestures and movements as the original central character.
Do not invent new actions.
17–22 SECONDS:
The woman reaches and holds the same central position as the reference.
She becomes calm and composed.
Maintain the exact same camera composition.
22–30 SECONDS:
The crowd begins the same synchronized chaotic movement from the reference.
The central woman remains calm and controlled in the center while the surrounding crowd moves violently and rhythmically around her.
Match the exact timing, intensity and choreography of the original crowd movement.
The central woman must remain clearly identifiable and visually separated from the crowd.
CAMERA
COPY THE REFERENCE VIDEO'S CAMERA LANGUAGE EXACTLY.
Preserve:
- camera position
- camera height
- camera distance
- lens perspective
- focal length feel
- framing
- depth of field
- camera movement
- push/pull movement
- tracking
- camera shake if present
- zoom
- subject scale
- transitions
- shot duration
- exact cut points
Do NOT create new shots.
Do NOT add new camera movements.
Do NOT change the original video's pacing.
The output should have the same visual rhythm and edit structure as the reference video.
CROWD
Keep the original crowd concept and choreography from the reference video.
Large group of young men surrounding the central woman.
Preserve their:
- positions
- density
- formation
- movements
- synchronized choreography
- timing
- background behavior
Do not replace the crowd with women.
Do not introduce new characters.
Do not allow anyone else to become visually dominant over the main woman.
Only the central character is replaced.
LOCATION
Preserve the same environment and architecture as the reference video.
Keep the same:
- concrete/architectural setting
- stairs
- background structure
- crowd arrangement
- environmental depth
- overall composition
Do not redesign the location.
Do not introduce new buildings or props.
LIGHTING / COLOR
Preserve the reference video's overall cinematic lighting and color grading.
Keep the dark, moody atmosphere.
The acid lime blazer should be the distinctive color of the central character.
Keep natural skin tones.
Realistic shadows and highlights.
No artificial glow.
No haze.
No bloom.
No excessive smoke.
No fantasy lighting.
VISUAL STYLE
Photorealistic cinematic live-action footage.
Natural human motion.
Natural body physics.
Realistic fabric.
Realistic hair.
Realistic skin.
Realistic cigarette smoke.
Realistic lighting.
No CGI appearance.
No animation.
No cartoon.
No plastic skin.
No face distortion.
No duplicated main character.
CRITICAL INSTRUCTION
REFERENCE VIDEO = controls EVERYTHING about the original video's motion.
<image_1> = controls ONLY the identity and appearance of the replacement central woman.
Replace the original central man with the woman from <image_1> while preserving the original video's actions, choreography, camera, timing, pacing, framing and cuts as closely as possible.
The result should look like the original video was filmed with <image_1> instead of the original actor.
ONE MAIN CHARACTER ONLY.
NO IDENTITY DRIFT.
NO EXTRA VERSION OF <image_1>.
STORM with a twist
Seedance 2.5
Prompt :
Create an ultra-realistic cinematic fashion/music video using the uploaded reference image as the exact visual reference for the woman, crowd, wardrobe, composition, lighting, and atmosphere.
A young woman with a short sleek dark bob, realistic facial features, natural skin texture, and subtle makeup wears a fitted black turtleneck, black trousers, black shoes, and minimal jewelry. Keep her face, hairstyle, outfit, and proportions perfectly consistent.
She walks slowly and confidently toward a large, tightly packed crowd, maintaining a calm and emotionless expression while the camera smoothly tracks backward and keeps her centered.
She reaches the exact center of the crowd and stops completely, standing perfectly still with her arms naturally at her sides and looking directly into the camera.
The crowd remains completely frozen for a brief moment.
Then, without the woman moving, the people behind her slowly begin turning toward her. The head turns start subtly and spread through the entire crowd until everyone is facing and looking directly at her.
The woman remains completely motionless and expressionless, never reacting to the crowd.
The camera slowly pushes closer while the crowd stays turned toward her, creating an unsettling synchronized effect. Their final movement is one sharp, simultaneous turn toward her while she continues staring into the lens.
Dark, mysterious, photorealistic live-action aesthetic, soft cinematic lighting, deep shadows, realistic skin, hair and clothing physics, shallow depth of field, natural human movement.
No animation, anime, cartoon, CGI look, duplicate woman, distorted faces, morphing, random crowd movement, text, subtitles, logos, watermark, excessive camera movement, artificial fog, glow, or exaggerated effects.