Jesse Lau 遁一子

统计数据

948
文章
0
粉丝
0
获赞
33922
阅读
avatar
Gemini 3 Flash 在 AA-Omniscience 基准测试中幻觉率高达 91%🧐 当然3.0 Pro也很高，88%
avatar
还好，用的软管 看到这个想起来大学时候回家坐火车经过贵州，同学手伸出窗口，10块钱要买火车缓停区的东西，结果一个家伙跳起来把他钱就抓跑了。
avatar
20块Claude+Gemini，GPT辅助一下就差不多够一个月干5年以前100个project的我，真想不通为什么会有人用到1400刀一个月的情况
avatar
以前丁蟹变光头需要2步，这次测试一下一个prompt “Restore and enhance this photo as if it were captured today with a high-end modern camera. then make this man bald” 左：原图 中：gemini 右：GPT 感觉中间gemini更像一点
avatar
OpenAI内置的restore老照片效果还可以 还是用之前测试NBP的林徽因的老图 GPT自己生成了一大段prompt（图一） 生成效果（图二） 将OpenAI的prompt帖给gemini NBP（图三） 我感觉openai的更符合想象一些
avatar
OpenAI推出了GPT image 1.5 我不知道是还没推送到我（但我的images已经标识new），还是技术仍然落后 中文字不如NBP 用昨天作坐忘论的prompt试了一下（图一） OpenAI的文字错了不少（图二） NBP则表现良好（图三）
avatar
现在每天除了一个新project，再update一个老project claude code开2波英文初始，gemini web开几十波翻译，再为了省claude tokens，用codex-5.2high（不能用xhigh，会思考过度）审核django项目正确与否 三个月工资20刀的AI配合得从从容容游刃有余😊
avatar
短视频推送了一个背靠墙扭转深蹲，挺不错
avatar
x上推送了好些人生k线 我不去看人生了，不过启发了我来搞一个财运走势图 最近30天的干支对日主财运的影响。 以后作日内交易的时候先看看，运气差的时候控制好仓位或者空仓😄
avatar
Google translate这个AI学语言很好用啊😯 玩了好几轮了
avatar
antigravity号称将支持workspace用户
avatar
chatGPT写MT5交易程式要小心 其训练大概有错误，对布林线的取值不对（不排除其他函数也会有错） 我是让ta评价opus写的EA,ta说opus写的不对 我就让gemini评价两者的EA,再搜索MT5原文档，发现是chatGPT搞错了
avatar
今天周六，新开的《坐忘论》第一期视频公开了 时间轴： 00:00:13 现代人的矛盾：越想平静越焦虑 00:01:21 唐代藏宝图：司马承祯与《坐忘论》 00:02:41 修行的地基：为何「信」是根，「敬」是花托？ 00:04:33 卞和献玉与伍子胥：人类天性为何不信真理？ 00:06:26 内在的昏君：逻辑大脑如何砍断我们的直觉 00:08:24 核心功法拆解：庄子的「坐忘」三步骤 00:08:42 第一步「堕肢体」：卸下身体的盔甲 00:09:51 第二步「黜聪明」：关闭大脑的背景程式与噪音 00:11:49 第三步「同于大通」：打破你与世界的边界 00:13:06 最大的悲剧：怀宝求宝，骑驴找驴 00:14:21 思想实验：夜明珠与手电筒，你选哪个？ 00:16:55 实战应用一：停止将快乐的定价权外包 00:18:29 实战应用二：找回对生命的「敬畏心」 00:21:00 一分钟练习：随时随地找回内建的安宁
avatar
google ai builder黑客松截止了，总共有4000多个项目竞争😅 主要生成太容易，而且功能也还不错 我弄了个点菜辅助app，早做好了，前两天正好去餐馆，还用了来着 录了手机操作过程，本来想一镜到底，结果手机屏幕反射出来我的大头像，让我这i人有点不爽，所以，只用了前几秒 app全是google ai studio搞定，视频倒多用了几个AI工具 1. 前几秒的视频，用grok和veo3都作了，结果veo3单词写错了，所以用了grok 2. 静态图片用nano banano pro 3. 背景音乐是以前suno的track（版权无忧了） 所以app半小时，弄视频半小时 有这4000多参与者贡献idea开发，可以目测后面的app会出来很多很多。 感觉以后google可以用这个app buider专门搞一个市场，搭接其广告
avatar
刚传道德经第26章时发现油管多了个AI辅助 Ask Studio
avatar
我靠。又被奥特曼带进沟里一把 openai难道真tm完了？ 刚才试了一下django编程，其实就是项目的最后一步，检查python有无编译错误等等 gpt-5.2 xhigh思考了半天，把claude+gemini web原英文+翻译的py都改错了。
avatar
测试codex gpt-5.2 下来，感觉不太稳 --- 实际工程上最怕这种情况 本来第一个EA编译错误，已经通过告知其错误截图解决了 然后继续让ta作第二个EA的时候，无端端的又将第一个EA修改了（没有任何指令要求ta），再次出现编译错误
avatar
gpt-5.2 其他没测。 写EA目前看不及claude
avatar
gemini deep research又创一个新高分数 我有一阵子没作dr了，再试一把
avatar
我的web还没推送 codex推送了。 codex -m gpt-5.2 启动
avatar
大王旗又换了
avatar
用gemini web拆解翻译任务为10多个chat，比gemini cli稳定多了 遵循指令上强很多，再用claude check一下，一遍搞定 搞不懂为何gemini cli就不能处理好task的分步，非得我专属任务来写python调用gemini web
avatar
刚才想起来nana banana pro的玩预测 再玩个大饼。 nbp对走势不乐观啊 Prompt： “根据此图走势，画出未来10天的k线走势，以虚线在原图基础上扩展绘画” 属于纯娱乐，新增的坐标都不太对头
avatar
Google AI Studio黑客松，NBP预测前10名 看看你的项目是否跟NBP预测重复 Prompt：Google DeepMind - Vibe Code with Gemini 3 Pro in AI Studio Build with Gemini 3 and compete for $500,000 in credits Google DeepMind - Vibe Code with Gemini 3 Pro in AI Studio View Writeups Overview Join us for a Vibe Coding sprint. Use Gemini 3 Pro's advanced reasoning and native multimodality to turn natural prompts into working, complex applications in Google AI Studio. From December 5th to 12th, vibe code the next generation of AI-powered applications that solve complex real-world problems in Google AI Studio, share them with the community, and bring your ideas to life. Start 5 days ago Close 2 days to go Merger & Entry Description Imagine building an application that finally solves a daily frustration or addresses a global challenge that has been waiting for an AI breakthrough. This Vibe Coding with Gemini 3 Pro Hackathon challenges you to step up and use the latest advancements in AI to do exactly that. Harness Gemini's advanced reasoning capabilities and native multimodality to bridge the gap between idea and reality. You are tasked with building solutions that were impossible to build yesterday, like solving everyday personal friction points to tackling complex global challenges. Where will you make an impact? Science: Accelerate discovery and research Education: Reimagine learning Accessibility: Build tools that work for everyone Health: Improve lives and patient care Business: Reinvent workflows and goals Technology: Push the boundaries of code Accelerate Your Build in AI Studio The free tier is only available in AI Studio. To get started, access the Build tab of AI Studio to easily and quickly build advanced applications. You can also find existing templates to get started in our Gallery. Timeline Friday, December 5, 2025 (5:00 PM UTC) – Start date Friday, December 12, 2025 (11:59 PM UTC) – Final submission deadline December 13, 2025 - January 12, 2026 - Anticipated judging period* Early 2026 – Winners announcement *Time required to evaluate results is dependent on the number of submissions and subject to change. All deadlines are at 11:59 PM UTC on the corresponding day unless otherwise noted. The competition organizers reserve the right to update the contest timeline if they deem it necessary. Getting Started in AI Studio Google AI Studio is a platform for building with Google's Gemini models. The two main components are: Playground: Here, you can try different models and their capabilities Build: Here, you can vibe code products, mix-and-match different AI-powered apps, and more! For this hackathon, you'll bring your ideas to life in Google AI Studio Build building with Gemini 3. Once you enter your idea, an AI studio coding environment window will open. This is where you can preview your app, the code and make changes. If you want to learn more about AI Studio Build features, you can look at the following videos: Overview vibe coding in AI Studio Annotation Mode Vibe Code with your voice App Gallery in AI Studio for inspiration AI Studio Build uses Gemini 3 Pro Preview by default, so just go to the Build tab and get started! Submission Requirements Please read carefully. Projects that do not follow these formatting rules will not be eligible for prizes. A valid submission must contain a Kaggle Writeup, which includes: Kaggle WriteupAttached Video Demo Attached Public AI Studio App Link To create a new Writeup, click on the "New Writeup" button here. After you have saved your Writeup, you should see a "Submit" button in the top right corner. Your final Submission must be made prior to the deadline. Any un-submitted or draft Writeups by the competition deadline will not be considered by the Judges. Note: If you attach a private Kaggle Resource to your public Kaggle Writeup, your private Resource will automatically be made public after the deadline. All of the content on Kaggle is written in Markdown. If you are copy-pasting from another application, like Word or your browser, you may need to clean up the markdown or html for things to display properly. 1. Kaggle Writeup The Kaggle Writeup serves as your project report. This should include a summary of what you are submitting, how you created it with links to the AI Studio app, and why it's important and the impact you are creating. You must select a track for your Writeup and include a title, subtitle and thumbnail image in order to submit. The Project Description should not exceed 250 words. Submissions over this limit may be subject to penalty. The below assets must be attached to the Writeup to be eligible. a. Video Demo An engaging video demonstrating your project in action. Length: Maximum 2 minutes. Visibility: It must be posted publicly (YouTube, X/Twitter, Loom, etc.) and a direct link attached to the Writeup. Access: It must be viewable by the judges without requiring a login. Licensing: All materials used in your video must be permitted for re-distribution and publication and you warrant you have the right to use the content included in your video. Note: If the uploaded video is longer than 2min, only the first two minutes will be reviewed. Make your impact early. b. Public AI Studio App Link This is the working app, the interactive demo. This allows judges to experience your project firsthand. Instructions: Once you enter your idea in AI Studio apps under Build, an AI studio coding environment window will open. This is where you can preview your app, the code and make changes. When you're done vibe coding and ready to submit, go to the top right corner of the coding environment and click "Save". Then click "Share app". A pop up window will appear. Enable "Publish your app" and "Default to full screen". This step is mandatory. Copy the link (this is your interactive demo link) and attach it as a Project Link in your Writeup. Access: All links must be publicly accessible and not require a login or paywall. Evaluation Show, don't just tell. Your project will be judged primarily on your video demo and the app that you create. This is your chance to create something exciting and compelling. Judges want to see the "wow" factor. Your video should tell a story, demonstrate the real-world impact of your app and leave the judges inspired by the technical skill and implementation of Gemini's latest capabilities and native multimodality. Judging Criteria: Your video demo will be the primary lens through which judges evaluate your project, with your Writeup and application serving as verification and providing technical depth. Submissions will be scored based on the following criteria: Impact (40%): Does the application solve a real-world problem? Is the vision inspiring and does the solution have a tangible potential for positive change? Technical Depth & Execution (30%): Does the application work? How effectively does it utilize Gemini 3's advanced capabilities (multimodality, reasoning, context)? Is the technology real, functional, well-engineered, and not just faked for the demo? Creativity (20%): How original is the idea? Does it leverage Gemini 3 in a way that wasn't previously possible? How many of Gemini 3's capabilities are you using and how well do they work together? Did you use Gemini 3 in new ways? Presentation Quality (10%): How exciting, engaging, and well-produced is the video? Does it tell a powerful story that captures the viewer's imagination? Does it clearly and effectively demonstrate the product in action, showcasing a great user experience? Does it have viral potential? Tracks and Awards Overall Track · $500,000 Description The top 50 Submissions will receive $10,000 in Gemini API credits each Track Awards Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 Winner $10,000 “”“” 上面是黑客松规则，想象得奖项目的icon，画出前10名，有icon，项目名称等等
avatar
今天天气不错，去罗托鲁瓦红森林走一圈步道