🔥 Search Hot Tweets
Search and analyze hot tweets from KOL accounts list (list: https://x.com/i/lists/1961235697677017443) within 6 hours. Use SoPilot plugin to quickly comment and occupy the comment section.
哈哈哈,自用的视频生成 Skill 终于做好了。 以后生产视频方便多了,只需要一句话! 公开技术方案: 1. Listenhub API实现声音克隆,合成,字幕时间轴控制 2. Seedream 4.5 生成背景封面 3. Manim库实现文本动画 4. FFmpeg合成视频。 同时支持16:9 和9:16 视频,抖音、小红书我来了!
Introducing slide-deck skill 🎨 Turn any article or content into professional slide decks with AI-generated images. 15 styles to choose from: • blueprint - technical diagrams • sketch-notes - hand-drawn feel • corporate - investor-ready • pixel-art - retro gaming vibes • watercolor - artistic warmth ...and 10 more Just run /baoyu-slide-deck https://t.co/Mr6DFuOZiM or pdf and get: ✓ Auto-generated outline ✓ Individual slide images ✓ Merged PPTX ready to share Install: npx add-skill jimliu/baoyu-skills https://t.co/pRDqxdiSrH
DeepSeek 最近发的三篇论文,全部都是带梁文锋署名的,论文讲的内容也很有意思。 尤其是那篇讲 Engram 的,DeepSeek 相当于给模型带了一本“字典”进考场,把死记硬背的脑容量腾出来,专门用来做难题。 但他们发现,如果模型全是记忆(Engram),那就是个死记硬背的书呆子,完全没法思考。 但如果全是推理(MOE),又要浪费大量算力去推导像“中国首都是哪”这种固定知识。 那要带多少记忆去考试呢? DeepSeek 研究出了一个模型“记忆”和“思考”的黄金比例。 最后测出来的最佳配比是:75% 给思考,25% 给记忆。 这个结论可能不仅适合于模型,也值得人类深思。 当一个人记住所有细节,就约等于没有空间思考。 逻辑思维,适当抽象,是人类进步的源泉。 当一个人完全没有知识,只会浪费脑力去思考最基本的东西,脑力空转。 DeepSeek 实测,加了字典减少思考后,模型多出了相当于 7 层网络的思考深度。 没想到,知识的广度以这种方式增加了思考的深度。 很有启发。
gitlab这个玩意现在用的还多么? 我记得我当年大一的时候,班里有人为了方便写项目有人弄了个gitlab。 刚看了一眼,这么多年了,PE还是负数,PS是6.23,57亿的市值,真的还有人在用这个东西么,github被微软收购之后,gitlab感觉已经逐渐淡出视野了,刚看主页也是拥抱AI了
很多人和我探讨过,为什么某某账号看似不错,怎么突然就断更了 我始终坚持:这属于哲学上的不可知论,我们需要承认我们无法知晓答案 就算原作者告诉你他断更的原因,这也不是真相 因为作者自己也不知道真正的原因 他可能会说「太忙了」「没灵感了」「生活变故」,但这些都只是他能意识到的表层理由,是他的理性系统事后给出的「合理化解释」 真实的原因可能是:童年某个被忽视的经历导致的完美主义焦虑、多巴胺奖励机制的阈值变化、某个他自己都没察觉到的人际关系张力、荷尔蒙水平的微妙波动、大脑神经连接模式在某个时刻的重组,以及其他成千上万个相互作用的因素 更深层的问题是:当你问「为什么断更」时,你预设了存在一个单一的、可被言说的「原因」。但人的行为不是这样运作的。断更不是一个「决定」,而是无数微小的、无意识的倾向累积到某个临界点后的涌现现象

MapleShaw
🔥没想到平安夜的美女九宫格那么受欢迎,这不得也给兄弟们整一套帅气逼人的海马体写真了!🥳 只需上传你的靓照,给你生成九款不一样的发型妆容😍直接跟托尼老师说,我就要这款! 提示词👇 { "project_type": "Nine-grid Trendy Star Portrait Collage", "aspect_ratio": "3:4", "visual_style": { "color_palette": "Black and white, Monochrome, High key, Bright grayscale, Clean whites, Light grays", "background": "Studio background, seamless white paper, light gray concrete wall, minimalist bright space, no dark voids", "lighting": [ "Soft frontal lighting", "Butterfly lighting", "Studio lighting", "Flattering beauty dish light", "No backlighting", "No harsh shadows on face" ], "mood": "Trendy, Cool, Confident, Star quality, Fashion editorial, Energetic, Edgy" }, "subject_description": { "identity_consistency": "Consistent facial features across all 9 panels (based on input reference)", "hair_and_grooming": [ "Varied trendy hairstyles", "Cool messy undercut", "Styled quiff", "Textured crop", "Slicked back modern", "Designer stubble", "Masculine scruff", "Well-groomed beard" ], "styling": [ "Fashion forward", "Streetwear vibe", "Leather jacket collar", "Designer hoodie", "Minimalist layers", "Statement accessories (e.g., single earring)" ], "expressions/poses": [ "Confident smirk", "Looking off-camera coolly", "Hand running through hair", "Slight jaw clench", "Direct confident gaze", "Dynamic poses" ] }, "composition": { "layout": "9-grid collage, Dynamic layout (not perfectly uniform), Mix of close-ups and medium shots", "style": "Fashion magazine contact sheet, Editorial spread" }, "technical_specs": { "camera_emulation": "Medium format fashion camera", "film_stock": "Kodak T-Max 400 (fine grain, sharp)", "resolution": "8k, masterpiece, sharp focus" }, "negative_prompt": [ "Dark background", "Black void background", "Backlit", "Silhouette", "Harsh shadows", "Underexposed", "Old fashioned", "Dull", "Uniform grid", "Same hairstyle in all", "Clean shaven (unless specified)" ] } https://t.co/FlnoShbgxn
Est. 100 views for your reply
