Sabitlenmiş Tweet
Neo
177 posts
Neo
@timerringX
Citius, Altius, Fortius @GeorgiaTech Once GitHub Trending #1 https://t.co/ONsBqSgoBb Blogger | Support me at timerring.sol
Hong Kong Katılım Aralık 2010
534 Takip Edilen99 Takipçiler
Very useful feature.
Pomelli By Google@PomelliByGoogle
Today, we're introducing Pomelli Catalog.📣✨ Add your products or services, and Pomelli will use them to generate personalized campaigns and high-quality photoshoots, tailored to your brand. Free of charge. Available everywhere. Try it at: labs.google.com/pomelli
English
Launched at the same time as Claude? Looks like there are more corn cases to be covered.
Perplexity@perplexity_ai
Today we're launching Perplexity Computer for Professional Finance. Finance teams can bring licensed data from providers like Morningstar, PitchBook, Daloopa, and Carbon Arc into Computer. We’ve also added 35 dedicated finance workflows for the work analysts repeat every week.
English
美军 C17 运输机，目前正以每天一架的频率，降落至北京首都国际机场。根据北京网友们的拍摄，昨晚再有两架 C17 降落北京。
似乎是在为特朗普 5 月 14 到 15 日，正式访问中国北京做准备。
Meguro-ku, Tokyo 🇯🇵 中文
This is incredibly exciting news! I have always considered Warp to be a top-tier tool among all the terminals I have used. I never expected it to actually go open-source now. A really respectable choice.
Warp@warpdotdev
Warp is now open-source.
English
Learned a great deal from this. Still need to reflect often.
Zara Zhang@zarazhangrui
1. What you find obvious is most likely not obvious to most people 2. What you think might be an extremely niche need might not be as niche as you think
English
github.com/anyspecs
大概一年前做过类似的事情，很好的方向，不过当时探索的最后结论是 Bottleneck 不在工程化实现上。而是在实际到底能否效果好，更包括了协作的一致性，如何从语义层面上回滚以及合并而尽可能避免偏移，如何在不同模型，不同上下文长度，不同 compact 机制 下保证 context 的还原，当时还甚至探索过 MemU 以及 Mem0的方案，总之 Effectiveness comes first，加油。🤠
中文
感谢 @leon7hao 分享 Lody 🤗🤗
喜欢 Lody 基于 story 做全量的 context 管理，而且可以团队共享!!（以前让搏哥干黑活是把电脑给他🤫以后可以共享 workspace 坐享其成，还能学习上下文
GitHub 是分布式代码仓库，我们会不会有分布式 context 的管理工具？
Dinah@dinah_zhang
本周日杭州 build 03(agent team 《Lody: 随时随地指挥你的 Agent 团队》@leon7hao 欢迎来现场和 agent team 一起 build something 报名链接: luma.com/dzrj8568
中文
If you want to launch iOS apps faster and with less App Store hassle, Blitz is an easy recommendation because it automates the painful submission workflow and works with MCP clients. It has saved me a great deal of time.
github.com/blitzdotdev/bl…
English
中文
@Solana_zh @matrixdock @Raydium 人声和背景音被过度限幅并且还削波，动态范围都拉到 1.2 LU 了。用 Meta 的 Demucs 模型跑了一遍分离，然后重新用 ffmpeg 压制了一遍视频。
大晚上的快睡着了让一个视频吓醒了，重制了一遍视频，后来的朋友们直接看这条重制后的视频吧，愿再没有人被吵醒。。。
中文
Furthermore, I believe that all the so-called skills or MCPs on the market designed to replicate web designs or mimic the styles of beautiful websites will be completely replaced by GPT image 2. Because the vast majority of AI models utilize very conventional components, it is difficult to expect a coding AI to replicate a webpage exactly without any additional media resources. This limitation exists because these tools cannot replace core, custom-designed media assets that are essential to the original design. However, image models can simultaneously handle all the tasks you need completed.
It is difficult to describe this feeling, but it is undeniable that if AI cannot fully understand a specific style you want or replicate it perfectly through traditional means, using image models for generation becomes an excellent solution.
I believe many developers will gradually come to realize this point.
Neo@timerringX
I am a little bit stunned. I believe the best workflow for frontend development right now involves: 1. Using Claude Design to help refine all the various requirements of your product. 2. Having it generate wireframes or mockups for the main user flows. 3. Using OpenAI's GPT Image 2 to complete all the remaining screens for every related process. 4. Leave it to coding to help you complete the restoration of the frontend page. The results are truly impressive. I think this could represent a genuine breakthrough in the field of front-end development. There have been some qualitative leaps that I could never have imagined just three or four years ago.
English
It is often hard to believe that you can generate such exquisite images using fewer than 100 words. This reminds me of three or four years ago when everyone was using products like Stable Diffusion UI or Midjourney. Back then, you had to write over 1k words of complex positive and negative prompts. Even with all those adjectives, you would usually only end up with an image that was barely passable. The evolution we have seen in just the last year or two is truly incredible.
All example prompts are here: github.com/wuyoscar/gpt_i…
English
I am a little bit stunned. I believe the best workflow for frontend development right now involves:
1. Using Claude Design to help refine all the various requirements of your product.
2. Having it generate wireframes or mockups for the main user flows.
3. Using OpenAI's GPT Image 2 to complete all the remaining screens for every related process.
4. Leave it to coding to help you complete the restoration of the frontend page.
The results are truly impressive. I think this could represent a genuine breakthrough in the field of front-end development. There have been some qualitative leaps that I could never have imagined just three or four years ago.
English
A design spec standardizes #design intent into a format that agents can consume without ambiguity. Looking at it from time to time still gives me a lot of inspiration.
designdotmd.directory
English
@blackanger 大概率是 AI Native Version 七八年前就已经是特征工程 机器学习版本了，在这个逆向都是 AI 自动的情况下，这个场景不上 AI 都是对模型的低估。。
中文
