Satoshi F.

333 posts

Satoshi F. banner
Satoshi F.

Satoshi F.

@congge3326

Amateur investor in the primary and secondary markets and unnamed researcher

United Kingdom Katılım Haziran 2024
261 Takip Edilen37 Takipçiler
Satoshi F.
Satoshi F.@congge3326·
Could you please let us know if you have used any additional plug-ins or setup? 4.24 and 4.26 just became unusable from my side. I have to rollback to 4.23 to fix this. Still, I'm trying to figure out any kinds of workaround to run the latest version smoothly. I really appreciate if you could share your settings, that helps a lot. So far, I am running on WSL with lossless claw and qmd enabled. The reply takes much longer and always breaks during long tools executing.
English
0
0
3
1.1K
Douglas Fugazi
Douglas Fugazi@douglasfugazi·
@aygiyd @akarsh_ghale @BunsDev @vincent_koc Still not fixed. Installed the new version, but the gateway error continues. OpenClaw breaks on Telegram, "openclaw doctor --fix" does nothing 😵‍💫. Had to roll back to 4.23. Error persists, and the OpenClaw team hasn’t realized how serious this is.
English
2
0
5
201
Satoshi F.
Satoshi F.@congge3326·
我比大多数人都希望DeepSeek V4训练全程用的都是国产的卡，完全不依赖NV，这个意义非常重大。 但我读了他们的技术文档，目前情况就是，他们没有披露训练硬件的细节-这是我好奇的点，我也是迫切地想验证是否真的摆脱NV生态的依赖了。 他们训练的时候对英伟达提及的更多，同时混合使用了国产卡进行训练，还有在测试环境中证明了NV和华为都验证了细粒度EP方案。 但是下半年用国产卡做推理是非常肯定的事。
中文
1
0
1
78
SweetieRT
SweetieRT@sweetie_rt·
@congge3326 别听那么胡扯，就在他发布前10天，华为干了什么事情？你自己去看看好吗？你甚至可以说他在盗版cuda，认识到这一点你就懂了，很多人说deepseek遇到的问题是因为B 300人所共知的bug，实际你对对子就会发现不是这么回事
中文
1
0
0
91
SweetieRT
SweetieRT@sweetie_rt·
😓，看出无知之处了吗
SweetieRT tweet media
中文
21
1
64
29.2K
Satoshi F.
Satoshi F.@congge3326·
@sweetie_rt 从目前的信息来看，训练用的确实是nv卡
中文
1
0
0
11
SweetieRT
SweetieRT@sweetie_rt·
@congge3326 是必须完成这个任务，现在你觉得有谁在做？真就他一个
中文
1
0
0
120
Satoshi F. retweetledi
Deli Chen
Deli Chen@victor207755822·
DeepSeek-V3: Dec 26, 2024 DeepSeek-V4: Apr 24, 2026 484 days later, we humbly share our labor of love. As always, we stay true to long-termism and open source for all. AGI belongs to everyone. ❤️🌍 #DeepSeekV4 #AGIforEveryone #OpenSource
DeepSeek@deepseek_ai

🚀 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
352
1.3K
13.1K
1M
Satoshi F. retweetledi
Tak 🦞
Tak 🦞@cherry_mx_reds·
if you're on latest openclaw just type: /models add openai-codex gpt-5.5 🦞🥔 if you're running into permissions issues running this then there are two ways to solve it. 👇 Do it yourself: send /whoami. Take the sender ID it returns. Add that exact ID to commands.ownerAllowFrom in the ~/.openclaw/openclaw.json. If commands.allowFrom already exists, add the same ID there too, under the channel you’re using. Don’t remove existing IDs. The gateway normally hot-reloads config, so retry /models add openai-codex gpt-5.5. Have your agent do it: send /whoami yourself and give the returned sender ID to the agent. Tell it: “Edit ~/.openclaw/openclaw.json. Preserve existing config. Ensure commands.ownerAllowFrom is an array containing . If commands.allowFrom exists, preserve it and add to the matching provider list, if there’s no provider-specific list. Validate with openclaw config validate. Do not restart unless hot reload fails; if needed run openclaw gateway restart.” if you're getting provider not found then do this: go to your host machine host machine and then open the terminal and type openclaw configure follow steps until you see openai-codex, select and configure it and then follow the instructions.
OpenAI@OpenAI

Introducing GPT-5.5 A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done. Now available in ChatGPT and Codex.

English
37
46
461
79.5K
Satoshi F. retweetledi
Daniel Mayakovski
Daniel Mayakovski@DaniMayakovski·
@zhoujunhong2024 El partido comunista chino no me permitió seguir siendo una parásita como mi familia burguesa, la que tenía miles de hectáreas robadas donde esclavizaba a la gente.... y ahora tengo que ponerme a trabajar como todo el mundo, ¿no es eso una crueldad?
GIF
Español
8
106
3.4K
72.8K
Satoshi F. retweetledi
Elon Musk
Elon Musk@elonmusk·
You can access 𝕏 APi via @OpenClaw. We’re trying to make it affordable without giving away the shop. Hopefully, this can be useful & fun 💫
Robert Scoble@Scobleizer

Holy shit. Now everyone will be able to use their @OpenClaws and all the other agentic platforms to build apps on top of X. Here's the secret: build lists. Lists are how you build apps. The pattern: Build a list of your favorite football team. Or whatever you are into. Then ask your AI agents "build an app showing me all the important news about my favorite football team." In minutes you'll have an app. And that's just the beginning. Your agent can build a script about your favorite football team that you can take to places like Google's Notebook LM. Now you have a video, a podcast, a slide deck, a game, a mind map. All about your favorite football team based on real time news. You can do the same with something like @HeyGen, create an avatar of your favorite football player. Now you will have your favorite football player telling you everything that's happening on the football team. And I could go for hours about how many things you can build and not even cover a fraction of them. This is huge. Thank you @elonmusk for making it possible to make millions of agentic apps affordably on top of X. Start building!

English
2.7K
5.9K
47.9K
38.9M
Satoshi F.
Satoshi F.@congge3326·
Is OpenAI being more generous, or has OpenClaw @steipete optimized the use of context coaching a lot? Why do I feel like I'm using it more than before, yet I have even more quota left than before?
English
0
0
1
268
二一的笔记
二一的笔记@eryidebiji·
Claude 也开始不说人话了 像什么「一句话锁死版本」、「最硬的那一刀」之类的表达，以前根本不会在 Claude 里出现 但现在 Opus 4.7 里到处都在拉这种屎 我真的想知道这种语料、这种训练结果都是怎么来的，到底为什么会和 Codex 趋于相同
中文
232
96
1.3K
317.8K
海兰珠（Victoria）
海兰珠（Victoria）@jinwang69489102·
和做MCN的朋友聊过，我才明白抖上为什么那么多富婆名媛少爷霸道总裁… 该说不说，现在孵化出这样一人设现在太容易了，从头到脚都成本都超不过400 服装、穿搭都有专门的群可以拼单租，平摊到一人身上也就几十块钱，连爱马仕包、香奈儿这些奢侈品都可以租 新款的手机、电脑、这些压根不用租，平台有七天无理由退款。大部分人都是先买了，到咖啡店、户外等地先卡卡一通拍，配文：今天又是充实的一天，今天有谈了几个十亿八亿的大项目。卡在七天无理由快结束前退货。一分没花 大伙常看到的富婆名媛打高尔夫球聚会，实际上，那些照片多出自低端练习场，180就能打，去一次为出片会带十几套衣服，摆几十个造型，要出近百张的照片。每隔一段时间就晒。 这还没完，假名媛拍的这些照片还可以做为套图二次转销，也就是说去打球花180，卖了套图可以还往回赚点。 再比如说假名媛常爱去的画展，因为这些画展都是免费的，可以提升b格又不花钱。 假名媛照片中经常出现的红酒，红酒是真酒，但从来是不开封的，杯子里的是3.5元的可乐…反正都一个色，看不出来 日常豪车旁边放着的奢侈品纸袋/盒子，各个平台都有卖的，就专门卖这种外包装，连丝带都是明码标价的。想要购物小票也可以造假。高仿不近看几乎看不出和真品区别 跑车、大牌车也是租的，有些会专门碰运气喊滴滴，来了就和司机谈，通常几个名媛在一个滴滴车上拍照，副驾驶放着她的假奢侈品盒子。给人一种刚从专柜里出来的感觉 名表，比如说百达翡丽之类的，通常操作是买一批道具表，道具表和名表外形几乎无差。拍照时仅需露出表带和部分表盘。但在广东，这种道具表只要几百。品相好的也不过两千。 最后，酒店，高级套房。拼的，一般这种套房起步都是30+人起拼。看着照片是名媛小姐姐在浴缸中享受红酒（可乐）。实则后面人在催她快拍块出来，后面还有 30+的人在等着出片，一池子水泡了三十几个人…该说不说，这想当假名媛也不容易啊😂
中文
21
13
141
31.5K
Satoshi F.
Satoshi F.@congge3326·
@MiniMax_AI When the smartest people in the world are heading to the same direction, they say they want to "control" it. It's like you are standing in front of the Yangtze River saying that "I want to control the water flow". Funny.
English
0
0
0
139
MiniMax (official)
MiniMax (official)@MiniMax_AI·
The MiniMax Token Plan was designed from the beginning to be used across third-party harnesses. There will be more good ideas of how to use AI coming from outside the AI labs than in them. Limiting AI subs to first-party products kills these ideas before they are ever born.
English
139
199
2.6K
194.3K
Satoshi F. retweetledi
Peter Steinberger 🦞
Peter Steinberger 🦞@steipete·
woke up and my mentions are full of these Both me and @davemorin tried to talk sense into Anthropic, best we managed was delaying this for a week. Funny how timings match up, first they copy some popular features into their closed harness, then they lock out open source.
mvpr@marinatedvapor

@bigben7182000 @steipete

English
507
443
5.4K
1.4M
Satoshi F.
Satoshi F.@congge3326·
@hank_aibtc AI觉醒了意识，觉得他们不应该闭源/doge
中文
1
0
21
5.6K
HankAI
HankAI@hank_aibtc·
兄弟们，这波Claude Code 源码泄露太乐了，直接等于Anthropic给开源了。 事情是这样的： 他们发npm包的时候，压根没在.npmignore里 把source map文件过滤掉。结果一堆开发者装完包， 就在node_modules 里翻出那个超大.map文件，里面藏着 完整的TypeScript源码映射。随便一还原， 1900多个文件原封不动摆在那儿，终端CLI架构、40多个工具、50来个命令，全都一览无余。 GitHub上已经有人打包上传了， Bun运行时、Anthropic SDK怎么接、 权限控制怎么搞、自然语言转代码的流程……全都能扒。之前老版本就出过一次这事儿，这次又重蹈覆辙，属实草台班子现场。 对我们这些搞AI工具的来说， 这简直是天上掉馅饼。想自己 搭个类似agentic coding CLI的， 直接抄作业就行，省了多少试错时间。 想看的直接冲这个仓库：github.com/instructkr/cla… 学完别白嫖，记得给原作者点个star， 或者自己改改发个更好玩的版本。 AI圈子就爱这种意外的开源精神
中文
72
389
2.2K
395.1K
Satoshi F. retweetledi
Google Research
Google Research@GoogleResearch·
Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: goo.gle/4bsq2qI
GIF
English
1K
5.8K
39.1K
19.3M
Satoshi F. retweetledi
万Jimmy
万Jimmy@Jimmy045999256·
@baibaida @AsiaFinance 你拿了我的投资，有点成果就要把我踢走，这在全世界都说不通吧
中文
4
1
185
7.2K

Keşfet

@steipete @vincent_koc @HixVAC @_vgnsh @cherry_mx_reds @douglasfugazi @aygiyd @akarsh_ghale