Tweet ghim
Wei
196 posts

Wei
@WeiBuilds
🧑💻 Engineering manager by day 🤖 Building with AI after hours 🧠 AI workflows, automation, and leverage 🦞 Slight crustacean bias 人在美国,有合规业务公司,欢迎私聊合作
Atlanta, GA Tham gia Nisan 2022
36 Đang theo dõi39 Người theo dõi

实测真实
拿近期我在做的老照片修复举例,半个月已经收入500+了
省时且0成本 闲鱼去挂商品 一张图的利润是99%
提示词和效果如下:
在保持精确构图和色彩的同时,增强并提升图像质量。去除模糊,使肌肤呈现逼真细腻的质感:清晰的毛孔、细腻的纹路以及自然的光影过渡。保留浅色调和背景,并优化眼部、睫毛、嘴唇和发丝周围的边缘锐度,使人像照片拥有高端美妆摄影般的质感,呈现自然不做作的肌肤效果。
就这个万能提示词就行


黄小木@ai_xiaomu
说几个我验证过的AI副业真实收入: - AI文案代写:单篇30-200元,自媒体月分成500-3000 - AI短视频剪辑:每条50-300元,代运营月入3000-8000 - AI老照片修复:单张20-50元,情感价值高,复购率极高 共同点:零成本起步,工具全免费,时薪15-50元。 不是暴富,但比打工多一份底气。
中文

这篇是不是有点上强度了,不要紧,咱们有codex啊!
你注册好cloudflare以后,在codex装cloudflare插件,然后把文章丢给codex就行,要改什么参数你就和codex直接说就行,咱们的理念就是AI First,能用AI的绝不动手,搞起来吧!

雨哥向前冲@xiangxiang103
中文

@thsottiaux @steipete Yep, very enjoyable experience using gpt-5.5 with my 2 OpenClaw instances
English

It's also become super good after all the @steipete cooking over the last week or so. I've had the best experience with GPT-5.5 so far, it's night and day compared to two weeks ago.
Sam Altman@sama
you can sign in to openclaw with your chatgpt account now and use your subscription there! happy lobstering.
English

私有化部署大模型的终点是 Mac 还是 Nvidia?
私有化部署,Mac 还是 Nvidia,这个问题的答案在问出来之前就决定了,取决于你的钱包,以及你准备拿它干什么。
先说结论:
Mac 堆内存,Nvidia 堆算力,两件事不在同一个赛道上较劲。
统一内存架构的优势是你能把一个 72B 的模型塞进去跑起来,慢是慢,但它活着。
独立显卡的优势是解码速度,显存够的前提下,推理速度会把 Mac 按在地上摩擦。
问题是家用显卡的显存天花板在这。
5070Ti 跑大参数模型是种折磨,5090 32G 的价格已经逼近工作站卡了,而工作站卡本身才是这条路的真正入口。
所以预算决定命运:
6万以上,冲 Pro 6000 这类工作站显卡,别犹豫
3-4万,等 M5 的 Mac Studio,统一内存方案里目前最能打的
预算有限,继续用 5070Ti 跑 27B 以下的量化模型,别幻想
本地跑模型的意义只有几种:
隐私敏感场景、跑没有审查模块的模型、出图出视频这种长期高频需求。
如果只是偶尔测试,租 GPU 比买机器理性得多,模型进化速度比硬件折旧快,今天的"终点"明年就是起点。
真正的终点大概是云端,只是很多人不愿意承认这一点。
你觉得终点是 Mac 还是 Nvidia?
中文

It’s happening.
3 months ago I predicted in the next 2 years all hardware costs would triple
Base Mac Mini’s would go from $600 to $1,800
Top of the line Mac Studios would go from $10,000 to $30,000
Everyone called me insane. They said it was engagement farming. Yet it’s happening faster than I predicted.
I urged you to buy your own compute. I said intelligence will not get cheaper anytime soon. Only more expensive (100% true)
Now it’s all playing out in real time
It all makes too much sense. As openclaw and personal agents get more popular and powerful, people will need more compute, all while hardware is facing shortages
Even the top labs like Anthropic don’t have enough compute to serve models from months ago. Limits going down every week
Own your own compute. Host your models. Localize your agents.
Do it before having powerful intelligence is reserved only for the elite.

English

@WeiBuilds In 1 year the world will be so different at this pace lol
English

Grok Image is soo good now
Elon Musk@elonmusk
Grok Imagine tutorial made with Grok Imagine. These is all AI-generated!
English

Say hello to my new pet
OpenAI Developers@OpenAIDevs
Pets. Now in Codex. Use /pet to wake your pet.
English

很多人以为 agent 变强,靠的是模型更聪明。
但很多时候,真正让它变有用的,是你终于给了它一个能稳定调用的工具面。
Android CLI 这种东西的价值,不是更酷,而是让 agent 从“会点按钮”变成“能执行流程”。
developer.android.com/tools/agents/a…
中文

Made my own Codex Pet, looks familiar 👀
OpenAI Developers@OpenAIDevs
Pets. Now in Codex. Use /pet to wake your pet.
English
















