Frank He

453 posts

Frank He banner
Frank He

Frank He

@fshewl

Think, Build, Ship

San Francisco, CA Katılım Şubat 2011
292 Takip Edilen487 Takipçiler
Frank He retweetledi
Anthropic
Anthropic@AnthropicAI·
Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing
English
2K
6.7K
44K
30.8M
Frank He
Frank He@fshewl·
@Jiaxi_Cui 我之前也很难理解，但发现codex在opencode也用不了，因为openai从来就不允许subscription用在3rd party上，anthropic只是现在才封而已，3rd party用API的话一直没问题。感觉是opencode明知故犯，竖靶子而已……
中文
0
0
0
134
Panda
Panda@Jiaxi_Cui·
真的很难理解 Anthropic 的决策 因为 opencode，最近不仅明显加大封号力度，还断供 opencode (而这种事之前对 windsurf 也做过) Codex 的态度完全相反：拥抱 opencode 生态 期待 Codex 把 CLI 做得更快、更顺手 Claude 的 skills 就是舍本逐末，和半年多前的 MCP 一样： 试图把复杂度推给开发者，开发者需要的从来不是一堆配置，而是直接跑通需求
Tibo@thsottiaux

We are working with OpenCode to allow Codex users to use their Codex subscriptions and usage limits in OpenCode directly. Also exploring how to support other awesome actors in the space.

中文
22
2
123
44.6K
Frank He
Frank He@fshewl·
@SIGKITTEN Thanks for pointing this out. I didn’t even know this is the common practice for all. Anthropic seems just closed a loophole?
English
0
0
1
548
Frank He
Frank He@fshewl·
@thdxr Does it mean ChatGPT currently has exact restriction as Claude? I can’t use OpenAI subscription at open code, right? Then what is the problem of Anthropic blocking the loophole of using subscription credentials in 3rd party app?
English
1
0
1
96
dax
dax@thdxr·
we are working with openai to allow codex users to benefit from their subscription directly within OpenCode
English
188
98
3K
511.1K
Frank He
Frank He@fshewl·
NotebookLM is best productivity tool. It gives knowledge structure.
English
1
0
0
74
一阁
一阁@yegle·
我们组（Serverless SRE）有个L3 SWE-SRE的HC，base在SF，RTO 2天，不知道有没有人有兴趣？
中文
2
2
6
4.6K
Frank He
Frank He@fshewl·
@addyosmani Need agentic capability from Gemini in Chrome!
English
0
0
0
106
Addy Osmani
Addy Osmani@addyosmani·
🚀 Gemini in Chrome is here! US users will get AI-powered tab summaries, YouTube summaries, calendar integration & smart search in the Omnibox - no Google AI sub needed! Enhanced safety too. More agentic features soon! youtube.com/watch?v=WjOvZ8…
YouTube video
YouTube
English
39
54
408
76.6K
宝玉
宝玉@dotey·
Vibe Coding 是个糟糕的名词，很有误导性，它最大的意义是借助 AI 做原型开发，能帮助快速确定产品需求。在软件工程中，原型开发的代码通常都是抛弃型的，在正式开发产品时需要重新做系统设计，再编码和实现，Vibe Coding 的结果也是类似的，在确定完需求后，还是需要重新设计再开发。
铁锤人@lxfater

估计很多人不知道什么是 vibe coding？ 这个术语是由人工智能大神 Andre Karpathy 创造的， 当你向AI描述问题后，然后它就自己写好代码了。 只适用于那些只有自己会使用的简单周末项目 因为这是不重要的项目，他可以凭感觉行事，而不必真正计划或测试。 👇是梦开始的地方 x.com/karpathy/statu…

中文
28
10
137
49.3K
Frank He retweetledi
Karl Weinmeister
Karl Weinmeister@kweinmeister·
Want to take your AI project from a cool demo to a real-world application? 🚀 Get ready in "Accelerate AI with Cloud Run," a free, full-day, hands-on developer workshop! 👉 goo.gle/accelerate-ai Registration is now open for our first tour in North America. Seats are limited!
Hudson Bend, TX 🇺🇸 English
2
7
13
1.9K
Frank He
Frank He@fshewl·
Run Docker Compose seemlessly on Cloud Run, including your ai model!
Steren@steren

Announcing @Docker compose support for @GoogleCloud Run: - Define your app in compose.yaml - Run locally with docker compose up - Deploy with gcloud run compose up

English
0
0
3
111
yetone
yetone@yetone·
/compact 再继续就好了，效果一模一样，还不用复制 summary（
周尔复@cholf5

使用 GeminiCLI 的一个技巧：如果对话太长，AI的幻觉越来越多，问题越来越难以解决，不妨开一个新的会话。在退出前让它做一下工作交接，Prompt 可以这样说：本次对话的上下文已经太长了，我打算关掉并重新开一个新的会话。你有什么想对你的继任者说的，以便它能更好的理解你当前的工作并顺利继续？

中文
9
20
215
62.4K
Ammaar Reshi
Ammaar Reshi@ammaar·
We're cooking up a fresh new design for AI Studio! ✨ And we'd absolutely love your input. If want to test an early prototype and share your thoughts on a call, just drop a reply below! Your feedback will help us shape it.
Ammaar Reshi tweet media
English
950
153
2.8K
485.9K
Frank He
Frank He@fshewl·
@tualatrix 上次我从湾区飞Vegas，到降落了我都没连上WIFI。只能放弃了。
中文
0
0
1
201
图拉鼎
图拉鼎@tualatrix·
美联航的 Wi-Fi 让我回到了曾经拨号上网的年代：又慢又贵。看到界面上写着“Network by Panasonic”，然后查询了一下，速度是垫底的。据说现在开始向 Starlink 服务升级了，最高能到达 100M，希望早点覆盖到全部航班。
图拉鼎 tweet media图拉鼎 tweet media
中文
5
0
16
10K
Frank He
Frank He@fshewl·
I saw the live demo at GCP Next in April. It's really amazing to see how to use Nvidia GPU on Cloud Run with Ollama and Gemma open models to build agents!
NVIDIA AI Developer@NVIDIAAIDev

Smart Health Agent uses multi-agent workflows for real-time health metrics 🌟 Check out Jay’s demo using Gemma 3 on accelerated GPUs and all your favorite technologies like LangGraph Ollama deployed to Google Cloud Run. Download or clone to see how you can use Gemma in action ➡️ nvda.ws/44aw6iR

English
0
0
2
129
Frank He retweetledi
NVIDIA AI Developer
NVIDIA AI Developer@NVIDIAAIDev·
Smart Health Agent uses multi-agent workflows for real-time health metrics 🌟 Check out Jay’s demo using Gemma 3 on accelerated GPUs and all your favorite technologies like LangGraph Ollama deployed to Google Cloud Run. Download or clone to see how you can use Gemma in action ➡️ nvda.ws/44aw6iR
NVIDIA AI Developer tweet media
English
7
31
114
21.1K
Frank He retweetledi
Steren
Steren@steren·
🚀Today at #GoogleIO, we're making AI deployments easier by introducing 3 new ways to deploy your apps to Cloud Run: 1⃣ Deploy AI Studio apps to Cloud Run in a click 2⃣ Deploy Gemma 3 to Cloud Run in a click from AI Studio 3⃣ Deploy code to Cloud Run from any MCP client with a new MCP server
Steren tweet media
English
5
110
600
84.7K
Frank He retweetledi
Steren
Steren@steren·
In this @theCUBE interview, @YunongX and I talked about the origins of Cloud Run, Serverless GPUs,... and even fashion!
Steren tweet media
English
1
3
6
642

Keşfet

@Jiaxi_Cui @SIGKITTEN @thdxr @yegle @addyosmani @dotey @yetone @ammaar