Jack Lee
1.9K posts

Jack Lee retweetledi
Jack Lee retweetledi
Jack Lee retweetledi
Jack Lee retweetledi

反代 灰产 中转 应该都在这里了👇
一、综合型多协议网关
CLIProxyAPI 是这个赛道的标杆,把 Gemini CLI、Antigravity、ChatGPT Codex、Claude Code、Qwen Code、iFlow 包装成 OpenAI/Gemini/Claude/Codex 兼容的 API 服务
AIClient-2-API 是 Node.js 实现,通过模拟 Gemini CLI、Antigravity、Codex、Grok、Kiro 的客户端请求,封装成本地 OpenAI 兼容接口 2026 年初还加了 Grok 的 Cookie/SSO 逆向,是目前对 Grok 支持最完整的
Antigravity-Manager 是桌面客户端路线,Tauri + React 写的,把 Google/Anthropic 的 Web Session 转成标准化 API 接口,带 OAuth 链接生成和账号池调度。这个适合拿来讲"账号管家"这种场景化内容
9router 和它的 TypeScript fork OmniRoute 是"智能路由 + 多档 fallback"的代表,iFlow、Kiro、Qwen 被标为 FREE 的是真免费无限,通过 OAuth 和 device auth 接入
ccproxy-api Python 实现,特点是直接复用 Claude CLI SDK tokens 和 Codex CLI 的 credential store,插件系统做得比较干净
CliGate 带可视化 Dashboard,支持 ChatGPT Account Pool、Claude Account Pool OAuth PKCE login、Antigravity Account Pool,一键配置 CLI 工具
二、Claude 专项
claude-relay-service 是国内最火的 Claude 中转方案,集成 Anthropic 的 OAuth 授权流程,在 Web 界面点击 Add Account 生成授权链接,登录 Claude 账号授权后接入服务。拼车党的基础设施,教程必讲
ClewdR 是 Rust 写的,支持 Claude 网页和 Claude Code,单一静态二进制覆盖 Linux/macOS/Windows/Android,Docker 镜像齐全,带 React 管理界面。性能路线代表
claude-code-proxy 是 Claude Code 转 OpenAI 的经典实现,教程里讲"双向转换"的基础案例
claude-relay(npow)是另一个思路,直接起一个 claude -p 进程来代理,而不是逆向协议
claude-unofficial-api 和 unofficial-claude-api(st1vms)是更早期的纯 Session Key 逆向(前者 JS、后者 Python),适合在教程里讲"历史演进"
Claude Code Action with OAuth是官方 Claude Code Action 的 fork,支持 OAuth 认证,让 Claude Max 订阅者在 GitHub Actions 里使用订阅
opencode-claude-auth 走 Keychain 路线,从 macOS Keychain 读取 Claude Code OAuth credentials,支持多账号自动检测
三、ChatGPT/Codex 专项
PawanOsman/ChatGPT 是元老级项目,把逆向成本打到几乎为零
acheong08/ChatGPT(revChatGPT)是逆向 ChatGPT Web 的祖师爷仓库
codexProxy(J1aDong/codexProxy)你自己做的那个方向,类似的还有不少把 Codex 包成 Anthropic Messages 入口的。
四、Gemini 专项
gemini-proxy(KashifKhn)是目前最干净的实现,Bun + Hono + TypeScript,OAuth 2.0 + PKCE 浏览器登录,自动刷新 token,不需要付费 API key,不需要 gcloud CLI
gemini-openai-proxy(Brioch)、gemini-cli-proxy(ubaltaci)、geminicli2api(gzzhongqi)是同类的三个竞品,各有侧重。
openai-gemini(PublicAffairs)是 Serverless 路线,可以直接部署到 Vercel/Cloudflare Workers,讲部署那集必备。
五、Copilot 专项
copilot-proxy(hankchiutw)一个简单的 HTTP 代理,把 GitHub Copilot 的免费额度暴露成 OpenAI 兼容 API,思路清晰。
github-copilot-proxy(BjornMelin)做的是反向,让 Cursor 调 Copilot 的后端,绕过 Cursor 的 500 次 premium 限制。
copilot-proxy(lutzleonhardt)是 VS Code 插件路线,通过 Language Model API 暴露,思路很野
六、Kiro / Qwen / Grok 逆向
kiro-gateway(jwadow/kiro-gateway)是 Kiro IDE / Amazon Q Developer 的网关,免费白嫖 Claude 模型的核心姿势。
Qwen-Copilot-Proxy(edwardgj)伪装成 Ollama 接口来对接 Copilot Chat,思路巧妙。
GrokProxy(CNFlyCat)走 Cookie 路线,从开发者工具 Network 面板抓 sso= 开头的 cookie 配置进 cookies.yaml,教程里讲 Cookie 型反代的标本。
七、Cursor 专项
Cursor-To-OpenAI(JiuZ-Chn/Cursor-To-OpenAI)把 Cursor 编辑器的 AI Chat 包成 OpenAPI,从 Cursor 客户端 cookie(user_ 开头)提取认证,网页 cookie 不能用。这个对讲"客户端 Cookie vs 网页 Cookie 差异"是绝佳案例
八、逆向号池 + 商用平台(拓展视野)
FakeOAI/tokens(FakeOAI/tokens)是商用级别,轮训号池将各大平台的模型能力转化为 OpenAI、Anthropic、Gemini 等平台的 API 接口标准格式,支持 Claude Code、Codex、GeminiCli 等终端调用。
中文
Jack Lee retweetledi

我只想说一句,这文章太tm牛逼了!
最好的诈骗教程就是看反诈案例。所以这篇做市商操盘案卷,就是最好的跟庄教程!
文中拿了两个经典案例,把 $myx 和 $coai 做市商的收割手法拆出来给你看👇
部分和我抓 $lab 的手法异曲同工,变相验证了里面的方法论
原文英文版,我将关键部分整理在下面了👇🧵

tradinghoe@tradinghoex
中文
Jack Lee retweetledi

For the next 5 days, we're making our Collector Crypt dashboard free to access.
The most comprehensive dashboard on @Collector_Crypt, track CC's Financials, Inventory, Users, and data on its token, $CARDS.

English
Jack Lee retweetledi

如何用土耳其账号订阅到最便宜的ChatGpt账号?
第一步:创建新的土耳其Appstore账号;
第二步:购买土耳其礼品卡兑换里拉;
第三步:下载土耳其地区的ChatGpt账号,然后订阅即可;
看看youtube上 小陈师傅c_z 的视频,很容易~
这里重点说一下,土耳其那边的礼品卡网站oyunfor 也可以使用加密支付的,我每次都是用加密支付的,根本就不用什么信用卡。
比特币橙子Trader@oragnes
推荐一个全球AppStore应用价格对比的网站,也有App应用,可以自己下。 这个网站上都可以看到不同地区的的App价格。 appstoreprice.org/zh
中文
Jack Lee retweetledi

Jack Lee retweetledi

🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
Try it now at chat.deepseek.com via Expert Mode / Instant Mode. API is updated & available today!
📄 Tech Report: huggingface.co/deepseek-ai/De…
🤗 Open Weights: huggingface.co/collections/de…
1/n

English
Jack Lee retweetledi
Jack Lee retweetledi
Jack Lee retweetledi
Jack Lee retweetledi
Jack Lee retweetledi
Jack Lee retweetledi

Claude Code 2.1.75 has been released.
1 flag change, 19 CLI changes, 2 system prompt changes
Highlights:
• Opus 4.6 uses a 1M context window by default on Max, Team, and Enterprise plans
• Tool permission denials prompt for a reason when intent is unclear instead of guessing next steps
• Memory files now show last-modified timestamps to help distinguish fresh vs. stale memories
Complete details in thread ↓
English
Jack Lee retweetledi

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes, today I measured that the leaderboard's "Time to GPT-2" drops from 2.02 hours to 1.80 hours (~11% improvement), this will be the new leaderboard entry. So yes, these are real improvements and they make an actual difference. I am mildly surprised that my very first naive attempt already worked this well on top of what I thought was already a fairly manually well-tuned project.
This is a first for me because I am very used to doing the iterative optimization of neural network training manually. You come up with ideas, you implement them, you check if they work (better validation loss), you come up with new ideas based on that, you read some papers for inspiration, etc etc. This is the bread and butter of what I do daily for 2 decades. Seeing the agent do this entire workflow end-to-end and all by itself as it worked through approx. 700 changes autonomously is wild. It really looked at the sequence of results of experiments and used that to plan the next ones. It's not novel, ground-breaking "research" (yet), but all the adjustments are "real", I didn't find them manually previously, and they stack up and actually improved nanochat. Among the bigger things e.g.:
- It noticed an oversight that my parameterless QKnorm didn't have a scaler multiplier attached, so my attention was too diffuse. The agent found multipliers to sharpen it, pointing to future work.
- It found that the Value Embeddings really like regularization and I wasn't applying any (oops).
- It found that my banded attention was too conservative (i forgot to tune it).
- It found that AdamW betas were all messed up.
- It tuned the weight decay schedule.
- It tuned the network initialization.
This is on top of all the tuning I've already done over a good amount of time. The exact commit is here, from this "round 1" of autoresearch. I am going to kick off "round 2", and in parallel I am looking at how multiple agents can collaborate to unlock parallelism.
github.com/karpathy/nanoc…
All LLM frontier labs will do this. It's the final boss battle. It's a lot more complex at scale of course - you don't just have a single train. py file to tune. But doing it is "just engineering" and it's going to work. You spin up a swarm of agents, you have them collaborate to tune smaller models, you promote the most promising ideas to increasingly larger scales, and humans (optionally) contribute on the edges.
And more generally, *any* metric you care about that is reasonably efficient to evaluate (or that has more efficient proxy metrics such as training a smaller network) can be autoresearched by an agent swarm. It's worth thinking about whether your problem falls into this bucket too.

English
Jack Lee retweetledi

在小红书上刷到一个账号叫 「虾薯」。
离谱的是——
这个账号的运营者不是人。
是一只 OpenClaw 虾 🦞
它会自己:
写文案、做封面图、发布笔记、回复评论。
一整套小红书运营流程,全自动。
背后其实是 GitHub 上的两个项目:
Auto-Redbook-Skills
负责内容生产 + 发布:
AI 自动写小红书笔记、Playwright 渲染排版图片(多主题模板)、自动发布到小红书(支持私密发布)。
xiaohongshu-ops-skill
负责运营 + 互动:
自动发布笔记(封面 + 标题 + 正文)、自动回复评论(按账号人设语气)、爆款笔记复刻(输入链接分析爆点生成类似内容)、自定义账号人设。
把这两个 Agent 串起来就是:
AI 写笔记 → 自动生成封面 → 自动发布 → 自动回复评论 → 自动复刻爆款。
一个人就能跑一个小红书账号矩阵。
AI Agent 已经开始自己做内容运营了。 🦞
中文
Jack Lee retweetledi

这个小红书全自动运行Skills功能很强大,支持 OpenClaw、Codex、CC 等所有支持 Skill 的编辑器,感谢white0dew大佬开源🤗
核心原理还是通过 Chrome DevTools Protocol (CDP) 控制测试浏览器完成自动化。
目前已经支持的功能有:
图文笔记自动发布(含话题标签写入)
视频笔记发布(本地视频 / 视频 URL)
多账号管理(账号隔离、切换、默认账号)
浏览器的无头模式 / 有头模式 / 远程 CDP
搜索相关话题笔记、获取笔记详情
对指定笔记自动评论
抓取「评论和@通知」数据
抓取内容数据表并支持导出 CSV(曝光、观看、点赞等)
登录状态缓存、复用标签页、随机延迟等稳定性优化
项目地址:
github.com/white0dew/Xiao…
中文
Jack Lee retweetledi

宣布一件大事,我们把 6551 的X + 全网新闻源MCP + SKILL 开源了!
很多人说,6551 的新闻源、推特面板很好用就是消息太多看不完。
还有很多朋友跟我说 X API 太难接,Skill 学不会,折腾半天龙虾就是跑不起来。
今天直接解决,我们把我们积累了1年的数据基础架构全部打包成 MCP + SKILL,任何人都可以几分钟部署,24h帮你看新闻。
🦞 你的龙虾现在可以:
• 直接连上 X 数据 + 全网50+实时新闻+链上数据,不用配 API 密钥。
• 24h 监控、分析、触发tg提醒。
照着 GitHub README 部署,几分钟就能装好。
欢迎大家安装试用和分享体验,有问题及时反馈及时迭代。
也欢迎👏🏻有热情的 dev 参加我们的生态
MCP
github.com/6551Team/openn…
github.com/6551Team/opent…
SKILL
clawhub.ai/infra403/openn…
clawhub.ai/infra403/opent…

中文
Jack Lee retweetledi










